Latest / INSPIRE STORY / #063_ STORY INSPIRE : They’re Lying About AI: 2027 Could Change Everything

Transcript

#063_ STORY INSPIRE : They’re Lying About AI: 2027 Could Change Everything · INSPIRE STORY

• 0:00
  Speaker 1: I think generative AI is at its heart con. And
• 0:03
  Speaker 1: seeing these ultra-rich, ultra-powerful people, life is a turns my stomach.
• 0:08
  Speaker 2: The word con is a strong word well.
• 0:10
  Speaker 1: What do you call something where from the very beginning
• 0:12
  Speaker 1: they've sold it in the terms of magic but it's
• 0:15
  Speaker 1: just a half-arsery machine they are misleading the entire world.
• 0:18
  Speaker 2: You are the first person that i've spoken to that
• 0:19
  Speaker 2: has that opinion well.
• 0:21
  Speaker 1: The fact that this is happening is insane and the
• 0:23
  Speaker 1: fact it's not a scandal is insane and i've been
• 0:25
  Speaker 1: in the tech industry for 16 years now and i
• 0:27
  Speaker 1: love technology and i'm enthusiastic about it but i don't
• 0:31
  Speaker 1: like being misled and this is the largest non-consensual push
• 0:35
  Speaker 1: of technology in.
• 0:36
  Speaker 2: History So we're going to play a game, Ed. I
• 0:38
  Speaker 2: have the things that you consider to be myths about
• 0:40
  Speaker 2: the AI industry.
• 0:42
  Speaker 1: Let's play it. The AI industry is creating enormous economic growth. No,
• 0:46
  Speaker 1: it's not. All of these companies run at a horrifying loss.
• 0:49
  Speaker 1: OpenAI lost $ 20. 9 billion last year. None of these people
• 0:52
  Speaker 1: can just say, yeah, we're on the path to making
• 0:54
  Speaker 1: this profitable because they can't.
• 0:55
  Speaker 2: Next one.
• 0:56
  Speaker 1: AI will replace all human jobs. That just isn't happening
• 0:59
  Speaker 1: and there's no economic data to support it. Next, the
• 1:02
  Speaker 1: United States need to spend trillions to beat China in
• 1:04
  Speaker 1: the AI race. What's the race to do? For us
• 1:06
  Speaker 1: to constantly piss our pants worrying about China? But people
• 1:09
  Speaker 1: keep saying, what if these models fall into the wrong hands?
• 1:11
  Speaker 1: They're already in the wrong hands. Mark Zuckerberg, Sam Altman,
• 1:15
  Speaker 1: Dario Amadei.
• 1:16
  Speaker 2: Mark Zuckerberg says, we'll continue to invest aggressively in infrastructure
• 1:19
  Speaker 2: to meet the demand. God matters a monstrosity.
• 1:22
  Speaker 1: Makes me think of Shrek with Lord Farquaad. Some of
• 1:25
  Speaker 1: you may die, but that's a risk I'm willing to accept.
• 1:27
  Speaker 1: If only these people gave a fuck about poverty or
• 1:30
  Speaker 1: actual problems in the world versus are we buying enough GPUs?
• 1:33
  Speaker 2: If this continues, what does the future look like? Fuck.
• 1:41
  Speaker 2: This is super interesting to me. My team gave me
• 1:43
  Speaker 2: this report to show me how many of you that
• 1:45
  Speaker 2: watch this show subscribe. And some of you have told us,
• 1:47
  Speaker 2: according to this, that you are unsubscribed from the channel randomly. So,
• 1:51
  Speaker 2: favour to ask all of you. please could you check
• 1:53
  Speaker 2: right now if you've hit the subscribe button, if you
• 1:55
  Speaker 2: are a regular viewer of the show and you like
• 1:56
  Speaker 2: what we do here. We're approaching quite a significant landmark
• 1:59
  Speaker 2: on this show in terms of a subscriber number. So
• 2:01
  Speaker 2: if there was one simple free thing that you could
• 2:04
  Speaker 2: do to help us, my team, everyone here, to keep
• 2:06
  Speaker 2: this show free, to keep it improving year over year
• 2:09
  Speaker 2: and week over week, it is just to hit that
• 2:11
  Speaker 2: subscribe button and to double check if you've hit it.
• 2:13
  Speaker 2: Only thing I'll ever ask of you, do we have
• 2:15
  Speaker 2: a deal? If you do it, I'll tell you what
• 2:16
  Speaker 2: I'll do. I'll make sure Every single week, every single month,
• 2:20
  Speaker 2: we fight harder and harder and harder and harder to
• 2:22
  Speaker 2: bring you the guests and conversations that you want to hear.
• 2:24
  Speaker 2: I've stayed true to that promise since the very beginning
• 2:26
  Speaker 2: of the Diary of a CEO, and I will not
• 2:28
  Speaker 2: let you down. Please help us. Really appreciate it. Let's
• 2:31
  Speaker 2: get on with the show. Ed Zitron. There are a
• 2:38
  Speaker 2: number of things that you believe that a lot of
• 2:40
  Speaker 2: other people don't believe. You have, I think, a couple
• 2:44
  Speaker 2: of controversial opinions and opinions that are in contrast to
• 2:48
  Speaker 2: the other guests that I've sat here with. What exactly
• 2:51
  Speaker 2: are those opinions, Ed?
• 2:53
  Speaker 1: I think generative AI is at its heart con. I
• 2:57
  Speaker 1: don't think it is sold as honest software. I think
• 3:00
  Speaker 1: that they overstate both what it can do, what it
• 3:03
  Speaker 1: will do, and the underlying financials to the point that
• 3:06
  Speaker 1: they are misleading the entire world, and they're actively exploiting
• 3:09
  Speaker 1: the weaknesses in journalism, in our economies, and indeed within
• 3:13
  Speaker 1: the Responsible parties with sell-side analysts, governments, and all over
• 3:17
  Speaker 1: the shop.
• 3:17
  Speaker 2: The word con is a strong word.
• 3:19
  Speaker 1: Yeah, I mean, what do you call something where from
• 3:22
  Speaker 1: the very beginning they've sold it in the terms of
• 3:24
  Speaker 1: magic as this thing that will replace all jobs, that
• 3:27
  Speaker 1: will cure cancer, as all of these things. And when
• 3:29
  Speaker 1: you look at it, it's boring cloud software that's extremely
• 3:33
  Speaker 1: expensive and unprofitable and also unreliable at its core.
• 3:37
  Speaker 2: People will be asking, where are you drawing from in
• 3:39
  Speaker 2: terms of your references, your personal experiences, where will you educate,
• 3:42
  Speaker 2: what you study, what you write about, what you do, Ed?
• 3:44
  Speaker 1: So that's the funny thing is people say, he's not
• 3:46
  Speaker 1: going to finance experience. He's not going to tech. I've
• 3:49
  Speaker 1: been in the tech industry 15, 16 years now.
• 3:51
  Speaker 2: In PR, but still had practical experience.
• 3:54
  Speaker 1: And I love technology and I'm enthusiastic about it. And
• 3:57
  Speaker 1: this thing just comes along. Everyone is telling me it's
• 3:59
  Speaker 1: the best thing since sliced bread. It can't even do
• 4:01
  Speaker 1: the basics. It can't even do search. Well, whenever you
• 4:04
  Speaker 1: ask an AI person, well, what's your setup? They describe
• 4:07
  Speaker 1: this peewee's playhouse thing of like, well, you've got a
• 4:09
  Speaker 1: harness here and you've got to use the right prompt. Well,
• 4:11
  Speaker 1: you don't want to use that prompt. You want to
• 4:13
  Speaker 1: use this prompt here with this model, but don't use
• 4:15
  Speaker 1: this model for the beginning. But at the end, you're
• 4:17
  Speaker 1: going to want to use this model. And this is
• 4:19
  Speaker 1: meant to be artificial intelligence. It's meant to be smart.
• 4:23
  Speaker 1: It's meant to be autonomous. It's meant to be something
• 4:25
  Speaker 1: that you set and forget.
• 4:26
  Speaker 2: We have the sort of six leading AI companies on
• 4:29
  Speaker 2: the table here. Anthropic, Amazon, NVIDIA, Microsoft, OpenAI, Google. You're
• 4:34
  Speaker 2: saying that their fundamental business model is.
• 4:37
  Speaker 1: Well, their revenues are not really coming from AI. Up
• 4:40
  Speaker 1: until fairly recently, none of their revenues were coming from AI,
• 4:43
  Speaker 1: like dribbles a bit. Right now, 70% of all AI
• 4:47
  Speaker 1: revenues across those three companies are from OpenAI and Anthropic,
• 4:50
  Speaker 1: two unprofitable, unsustainable companies that literally cannot afford to exist
• 4:54
  Speaker 1: without these very same companies giving them money. Amazon sent
• 4:58
  Speaker 1: $ 50 billion to OpenAI this year. They sent $ 5 billion
• 5:02
  Speaker 1: to Anthropic. Google sent $ 10 billion to Anthropic. And in
• 5:06
  Speaker 1: the next three and a half years, OpenAI and Anthropic,
• 5:09
  Speaker 1: based on actual sell-side analyst evaluations, their estimates that inform
• 5:13
  Speaker 1: whether stock is going to go up or down after earnings,
• 5:16
  Speaker 1: they are expecting $ 400 or more billion of revenue, 30
• 5:21
  Speaker 1: or something percent of cloud growth, just from these two
• 5:24
  Speaker 1: unprofitable companies that will need to be given the money
• 5:26
  Speaker 1: from somewhere. And on top of that, these companies have
• 5:30
  Speaker 1: such low respect for the average investor, for the analysts,
• 5:34
  Speaker 1: for everyone really, that they don't even disclose their AI revenues.
• 5:37
  Speaker 1: The few times they deign us worthy, they use something
• 5:40
  Speaker 1: called a run rate, an annualized run rate, which means, well, nothing.
• 5:44
  Speaker 1: They never define it. It can mean month times 12.
• 5:47
  Speaker 1: It can mean month times 13. It can mean last
• 5:49
  Speaker 1: four weeks times 13. It's different every time and they
• 5:52
  Speaker 1: never define it. And then they sometimes just don't mention it.
• 5:54
  Speaker 1: So you've got this big thing that is meant to
• 5:56
  Speaker 1: be the biggest, most influential change to software ever. And
• 6:00
  Speaker 1: whenever you ask them about it, when you say, how
• 6:02
  Speaker 1: much are you making from this? They go, oh, I
• 6:04
  Speaker 1: couldn't possibly say. I'm too shy. These are public companies,
• 6:07
  Speaker 1: or at least the ones that aren't Anthropic and OpenAI.
• 6:10
  Speaker 1: When they have good news, they'll tell you. And when
• 6:11
  Speaker 1: they don't tell you something, well, that actually speaks volumes.
• 6:15
  Speaker 2: Have you used these tools? Yes. AI tools, Gemini, Anthropic, ChatGPT,
• 6:20
  Speaker 2: et cetera. And you found no value in them?
• 6:22
  Speaker 1: There's some value, but it's not. They've spent over a
• 6:25
  Speaker 1: trillion dollars in capex.
• 6:27
  Speaker 2: What does capex mean?
• 6:28
  Speaker 1: Capital expenditures. So When you are a business and you
• 6:31
  Speaker 1: have operating expenses like electricity, for example, those come right
• 6:34
  Speaker 1: off immediately. Capital expenditures are long-term investments that are theoretically one-off.
• 6:38
  Speaker 1: So a data center or indeed the GPUs you put
• 6:41
  Speaker 1: inside an AI data center.
• 6:43
  Speaker 2: Okay. So you've got a data center and then you
• 6:45
  Speaker 2: have these GPUs, which are like computer chips.
• 6:48
  Speaker 1: So AI GPUs are much bigger, much more power intensive.
• 6:52
  Speaker 1: They take a bunch of high bandwidth memory and they
• 6:55
  Speaker 1: Because of how many of them you need, you need
• 6:57
  Speaker 1: thousands of them, tens of thousands, hundreds of thousands in
• 6:59
  Speaker 1: some case, you need a bunch of power. So an example,
• 7:03
  Speaker 1: OpenAI and Oracle are building a data center in Texas,
• 7:06
  Speaker 1: in Abilene, Texas, 1.2 gigawatts called Stargate Abilene. Within that,
• 7:12
  Speaker 1: with each one of the eight buildings, there'll be 50,000
• 7:14
  Speaker 1: NVIDIA GB200 GPUs. So city of Bristol takes about 780,
• 7:19
  Speaker 1: 800 megawatts of power a year, right? Well, Stargate Abilene
• 7:25
  Speaker 1: is condensing more power than that, 1.2 gigawatts, into a
• 7:29
  Speaker 1: space around 1,172 times smaller. The city of Bristol is
• 7:33
  Speaker 1: about 1.2 billion square feet. Stargate Abilene is about 998,000.
• 7:38
  Speaker 1: So you're condensing all of this power, all of this money,
• 7:40
  Speaker 1: all of this labor into this one spot. And all
• 7:44
  Speaker 1: of these data centers cost billions of dollars. All of
• 7:46
  Speaker 1: these companies other than Microsoft are now to take out debt.
• 7:49
  Speaker 1: And the thing is, they've spent over a trillion dollars
• 7:52
  Speaker 1: so far, and they want to spend another trillion dollars
• 7:53
  Speaker 1: next year. And for what? To make tens of billions
• 7:56
  Speaker 1: of dollars, most of which comes from two unprofitable companies,
• 8:00
  Speaker 1: Anthropic and OpenAI.
• 8:02
  Speaker 2: One of the rebuttals to that would be that the adoption,
• 8:05
  Speaker 2: the customer adoption of people using OpenAI and Anthropic has
• 8:09
  Speaker 2: been absolutely insane. These are the fastest growing products in
• 8:12
  Speaker 2: all of history, especially as it relates to sort of technology.
• 8:15
  Speaker 2: If we just focus in on technology, there are, you know,
• 8:17
  Speaker 2: hundreds and hundreds of millions of people, billions of people
• 8:20
  Speaker 2: are using these tools every single day. for things that
• 8:23
  Speaker 2: they have subjectively decided are problems they need solving. So,
• 8:28
  Speaker 2: you know, money is a lagging indicator of value. So
• 8:32
  Speaker 2: one would argue that they're just investing ahead of the
• 8:35
  Speaker 2: monetization options.
• 8:36
  Speaker 1: The first, let's start with this adoption. Is it honest
• 8:40
  Speaker 1: adoption when you are forced to use generative AI when
• 8:43
  Speaker 1: you load Google?
• 8:43
  Speaker 2: When you load Google.
• 8:45
  Speaker 1: Docs, Gemini screams in your ear. When you load Word,
• 8:48
  Speaker 1: Copilot's bugging you. When you use Amazon, whatever Rufus AI is,
• 8:52
  Speaker 1: wants to have opinions on what socks you're buying. This
• 8:56
  Speaker 1: is the largest non-consensual push of technology in history. ChatGPT,
• 9:00
  Speaker 1: for example. Every single media outlet has been screaming about
• 9:03
  Speaker 1: this for three years. They've been saying this will take
• 9:06
  Speaker 1: your job. You must use this. If you don't use this,
• 9:09
  Speaker 1: you're going to be falling behind. So people are using
• 9:12
  Speaker 1: it because they've been told to use it constantly. And
• 9:14
  Speaker 1: they're using it like search predominantly. And that's partly because
• 9:17
  Speaker 1: Google fell behind search and also because Better ingesting queries sometimes.
• 9:21
  Speaker 1: Sometimes if you use a generative search, it's like a
• 9:23
  Speaker 1: trawling vessel. It's not very good at specifics. But if
• 9:25
  Speaker 1: you're like, does this thing exist? Has this person ever
• 9:28
  Speaker 1: said anything like this? It'll still probably get it wrong,
• 9:31
  Speaker 1: but it'll scour the ocean for you. Nevertheless, that's not
• 9:34
  Speaker 1: worth a trillion dollars. None of it is. The amount
• 9:37
  Speaker 1: of money being sunk into this is just incomparable to anything. Railways,
• 9:42
  Speaker 1: it blows everything out of the water because there is
• 9:46
  Speaker 1: no post-bubble story even for this. AI GPUs are not
• 9:50
  Speaker 1: useful for other things either. It's a directionless egregore of capitalism,
• 9:55
  Speaker 1: this headless beast that lumbers around desperately to seek out
• 10:00
  Speaker 1: growth everywhere in the hopes that if it harasses people
• 10:03
  Speaker 1: and scares people and demonizes labor enough, people will be
• 10:08
  Speaker 1: forced to use it.
• 10:10
  Speaker 2: The reason I pause is because I think about my
• 10:13
  Speaker 2: own company. Obviously, everybody thinks about their own personal situation.
• 10:16
  Speaker 2: So you have people listening now that don't use any
• 10:17
  Speaker 2: AI tools. Then you have people that are using it
• 10:20
  Speaker 2: for everything from coding new software tools to everything they
• 10:24
  Speaker 2: write to, you know, images, whatever. And when you look
• 10:26
  Speaker 2: at the stats around enterprise adoption, it says 88% of
• 10:29
  Speaker 2: organizations regularly use AI at least once for one particular
• 10:33
  Speaker 2: business function. And I'd say in our company, 95% of
• 10:38
  Speaker 2: people use one of these AI tools like Anthropic or
• 10:41
  Speaker 2: ChatGPT or Gemini every day. Right. And that exists on
• 10:44
  Speaker 2: some kind of spectrum of like the super users that
• 10:47
  Speaker 2: that are using it probably every hour of every day
• 10:50
  Speaker 2: for almost everything, to someone maybe hiring the executive team
• 10:54
  Speaker 2: that's using it less because their job doesn't require of
• 10:57
  Speaker 2: it as much. And when you look out into the
• 10:59
  Speaker 2: world at how the world is changing from a content perspective,
• 11:04
  Speaker 2: if we're looking at generative AI, it is obvious that
• 11:07
  Speaker 2: these tools are being widely adopted. Part of the symptom
• 11:10
  Speaker 2: is the AI slop you see all over the internet.
• 11:12
  Speaker 1: Right.
• 11:13
  Speaker 2: So this idea that it's not being used, I struggle with.
• 11:19
  Speaker 1: It's being used. Here's the thing with the slop. Before
• 11:22
  Speaker 1: we had AI slop, we had SEO slop. Because Google
• 11:26
  Speaker 1: incentivized doing the lowest common denominator that would rank well
• 11:29
  Speaker 1: on search. There's a whole story about how they pulled
• 11:31
  Speaker 1: back spam guards, thanks to Prabhagar Raghavan, which we can
• 11:34
  Speaker 1: get into, where they made the internet worse by allowing
• 11:37
  Speaker 1: worse content to rank higher. It's why when you used
• 11:40
  Speaker 1: to Google... our.
• 11:42
  Speaker 2: Best washing machine.
• 11:43
  Speaker 1: There's 11 different horrible blogs that read like somebody got
• 11:46
  Speaker 1: a concussion. They are built to rank rather than be
• 11:49
  Speaker 1: read by humans or built to be made good. So
• 11:53
  Speaker 1: AI helps weaponize that scale. Yeah, you can make a
• 11:56
  Speaker 1: bunch of generic slop. We've had slop for years. We've
• 11:59
  Speaker 1: just found a slop machine. But then also there's the
• 12:02
  Speaker 1: problem of cost. So when you use AI services, you
• 12:05
  Speaker 1: burn tokens and it's per million tokens.
• 12:07
  Speaker 2: What's a token?
• 12:08
  Speaker 1: So it's around three quarters of a word.
• 12:11
  Speaker 2: So it's characters so the ai companies have a currency
• 12:14
  Speaker 2: in which they charge you like a taxi in new
• 12:16
  Speaker 2: york has a meter yeah and they call it tokens
• 12:19
  Speaker 2: yeah and every word let's just say for ease it's
• 12:22
  Speaker 2: a word yeah.
• 12:23
  Speaker 1: About a word yeah and it's per million tokens so
• 12:26
  Speaker 1: you'll be charged per million input tokens the stuff you
• 12:29
  Speaker 1: feed into it like a document or a code base
• 12:32
  Speaker 1: and the output tokens are both the stuff it spits
• 12:35
  Speaker 1: out at the end, but also when it thinks. So, okay,
• 12:37
  Speaker 1: you've asked me to give you the best restaurants in
• 12:40
  Speaker 1: this area of New York. I should find the best
• 12:41
  Speaker 1: restaurants in New York. All of that's output tokens as well. However,
• 12:45
  Speaker 1: when you're paying for a monthly service, you don't see
• 12:47
  Speaker 1: any of that. Put all that crap to the side.
• 12:50
  Speaker 1: They just have rate limits. So you can use them
• 12:52
  Speaker 1: a certain amount and then when you run out, but
• 12:53
  Speaker 1: they kind of obfuscate what that was. Now, someone recently found,
• 12:57
  Speaker 1: Semi Analysis actually found this, a big analyst group, They
• 13:00
  Speaker 1: found that on a $ 200 a month ChatGPT subscription, you
• 13:04
  Speaker 1: can burn $ 14, 000 worth of tokens. And on Anthropix, you
• 13:09
  Speaker 1: can burn $ 8, 000 for $ 200. That is how most, and
• 13:14
  Speaker 1: even on the $ 20 a month service, you can.
• 13:16
  Speaker 2: Burn $ 400.
• 13:16
  Speaker 1: Now, most people don't realize that. Most people have no
• 13:21
  Speaker 1: idea what AI costs. Most people just think, oh, it's
• 13:24
  Speaker 1: $ 20 a month. No, all of these companies run at
• 13:26
  Speaker 1: a horrifying loss. OpenAI lost $ 20. 9 billion last year because
• 13:31
  Speaker 1: people can burn as many tokens as they want. And
• 13:33
  Speaker 1: when they tried to move everybody on the enterprise side,
• 13:36
  Speaker 1: so companies bigger than 150, onto actually paying the cost
• 13:40
  Speaker 1: of AI in around March of 2026, to quote Sam Orman,
• 13:44
  Speaker 1: they said, people have a big problem with it. I
• 13:46
  Speaker 1: think it's a huge issue, which is not really what
• 13:49
  Speaker 1: the heir apparent to tech's history is meant to be saying.
• 13:52
  Speaker 1: But the point is, enterprises immediately started freaking out uber
• 13:56
  Speaker 1: burned through their entire annual token budget in three months
• 13:59
  Speaker 1: so suddenly after everyone's saying ai is the most productive
• 14:02
  Speaker 1: thing ever it's amazing it's changing everything the moment people
• 14:05
  Speaker 1: actually had to pay for it they go oh i
• 14:08
  Speaker 1: don't know actually um maybe it's it's obviously we all
• 14:11
  Speaker 1: love it it's all great right But it's costing too much,
• 14:15
  Speaker 1: so we now need to reduce the cost. Because people
• 14:17
  Speaker 1: are just dumping stuff into it, being like, what do
• 14:19
  Speaker 1: I do here? And getting whatever the median is out.
• 14:22
  Speaker 1: Because that's what these things do. They provide the median answer.
• 14:26
  Speaker 2: So essentially, someone like me who's a power user of
• 14:29
  Speaker 2: these tools, I could be costing Anthropic or OpenAI $ 1, 000,
• 14:35
  Speaker 2: but they're only charging me $ 100, let's say. So they
• 14:38
  Speaker 2: are having to subsidize $ 900 of my usage because of
• 14:41
  Speaker 2: the electricity costs and the costs at their data centers.
• 14:44
  Speaker 2: And so your assertion here is that that is unsustainable. Yes.
• 14:47
  Speaker 1: And just to be clear, they're probably not one for $ 1.
• 14:50
  Speaker 2: It might be $ 34.
• 14:51
  Speaker 1: We don't know. I think it's unprofitable. These companies don't
• 14:55
  Speaker 1: disclose them. Even in their auditive financials, they play funny
• 14:57
  Speaker 1: games with how they categorize things. But nevertheless, yes. And
• 15:01
  Speaker 1: on top of that, the way that you stand up inference,
• 15:03
  Speaker 1: which is the thing that creates the output within these
• 15:06
  Speaker 1: data centers, you're not just saying, okay, turn the inference
• 15:09
  Speaker 1: machine on, let's go. you are standing up the GPUs
• 15:13
  Speaker 1: necessary to take in the demand. And if you buy
• 15:16
  Speaker 1: too much, you've wasted the money. You have to pay
• 15:19
  Speaker 1: for the hourly GPU use regardless. If you buy too few,
• 15:23
  Speaker 1: your customers can't use it. They get pissed off at you.
• 15:25
  Speaker 1: They cancel the go with someone else. But nevertheless, yeah,
• 15:27
  Speaker 1: they would get demand selling $ 20 or $ 40 for a dollar.
• 15:31
  Speaker 1: And that's what these services do. And really the simplest
• 15:33
  Speaker 1: way to explain it is they were actually profitable if
• 15:36
  Speaker 1: they were actually just, they believe that these services were
• 15:39
  Speaker 1: worthwhile and that they were worthy of the cost. They'd
• 15:41
  Speaker 1: charge it. Regular people wouldn't be able to get a
• 15:43
  Speaker 1: monthly subscription. They'd just be paying what it's worth. Unless,
• 15:48
  Speaker 1: of course, there was an economic problem. And it's very simple.
• 15:52
  Speaker 2: You pay when you use an LLM, regardless of whether
• 15:55
  Speaker 2: you get what you want.
• 15:56
  Speaker 1: When these things hallucinate, say you're doing something, you're coding something,
• 15:59
  Speaker 1: and they go through a code base and they fuck
• 16:02
  Speaker 1: up a bunch of stuff, they break a bunch of stuff,
• 16:04
  Speaker 1: you're paying for that. You're paying for it whether it
• 16:05
  Speaker 1: works or not. Unless, of course, you're using one of
• 16:07
  Speaker 1: these subscriptions.
• 16:09
  Speaker 2: I think the really interesting point is, are they spending
• 16:13
  Speaker 2: ahead of the value showing up, which is I imagine
• 16:17
  Speaker 2: what they would argue, or are they spending all of
• 16:21
  Speaker 2: this money and subsidizing all of their users in a
• 16:24
  Speaker 2: way that's unsustainable and that will never be justified? Like,
• 16:27
  Speaker 2: does it, you know, because you think back through the
• 16:29
  Speaker 2: history of technology, you often get people losing money to
• 16:33
  Speaker 2: grab market share. Right. And they're also focusing on bringing
• 16:37
  Speaker 2: the costs down and making it more profitable for them
• 16:41
  Speaker 2: as well. But they can't afford to underinvest.
• 16:44
  Speaker 1: If they were bringing the cost down, they would have
• 16:46
  Speaker 1: brought the cost down, which they have not. It seems
• 16:49
  Speaker 1: to be getting more expensive. In fact, everyone, inference providers
• 16:52
  Speaker 1: don't seem to be profitable. Even the companies renting out
• 16:55
  Speaker 1: GPUs don't seem to be profitable. I imagine that it
• 16:58
  Speaker 1: wasn't like they started out and they were like, shit,
• 17:00
  Speaker 1: this is unprofitable. At the beginning, we know it, screw it.
• 17:03
  Speaker 1: We'll keep doing it anyway. I don't think it's some
• 17:05
  Speaker 1: big conspiracy. They probably thought at some point, yeah, this
• 17:09
  Speaker 1: will go profitable. The chips will catch up. Customers will
• 17:12
  Speaker 1: pay for the overwhelming value because you don't know in
• 17:14
  Speaker 1: 2023 where it's going to be in 2026. You assume
• 17:17
  Speaker 1: it's going to go up. That's the nature of venture capital.
• 17:19
  Speaker 1: They should have stopped in 2024 when OpenAI lost over
• 17:23
  Speaker 1: $ 5 billion. They should have been like, yep, this is
• 17:25
  Speaker 1: not going to work. But they kept going because it
• 17:28
  Speaker 1: helped number go up so much. It helped stock values pump.
• 17:31
  Speaker 1: It helped everyone pump. It helped Nvidia pump, Microsoft, everyone.
• 17:35
  Speaker 1: And not from the revenues. Because here's the funny thing
• 17:38
  Speaker 1: about Google, Microsoft, and Amazon. People for years have been
• 17:42
  Speaker 1: saying their AI bets have paid off. Wow, their AI
• 17:45
  Speaker 1: bets have paid off. As these companies refuse to say
• 17:47
  Speaker 1: how much they're making from AI. But because their existing
• 17:50
  Speaker 1: businesses continue to grow, and did so, by the way,
• 17:53
  Speaker 1: through price increases, changes to how Google and Meta did
• 17:57
  Speaker 1: advertising amazon bumped up prices and changed how they did
• 18:00
  Speaker 1: that actually amazon started a remarkable ad business during this
• 18:03
  Speaker 1: whole time as well and the selling through amazon platform
• 18:05
  Speaker 1: anyway nothing to do with ai but because number go
• 18:08
  Speaker 1: up because revenue go up everyone it's ai because these
• 18:11
  Speaker 1: companies wouldn't spend a trillion dollars for for no reason
• 18:14
  Speaker 1: right except in fiscal year 2026 which just ended for
• 18:18
  Speaker 1: microsoft annoying i know they made total according to bloomberg
• 18:23
  Speaker 1: about 34 $ 4. 33 billion, $ 24. 1 billion of that was from OpenAI.
• 18:29
  Speaker 1: So that leaves them with about $ 10 billion in a
• 18:31
  Speaker 1: year when they spent $ 115 billion on capital expenditures and
• 18:34
  Speaker 1: intend to spend $ 175 billion next year. The math does
• 18:38
  Speaker 1: not make sense. I imagine their plan was, okay, this
• 18:41
  Speaker 1: is just going to get exponentially more valuable. And at
• 18:43
  Speaker 1: some point, the costs will be outpaced by the return.
• 18:47
  Speaker 1: Problem is that large language models need a bunch of
• 18:49
  Speaker 1: money to train them. They need constant data flows through.
• 18:52
  Speaker 1: They need customized data. It's just this big, expensive monster.
• 18:56
  Speaker 1: And when you try and talk to people about it
• 18:59
  Speaker 1: and you try and say, hey, look, this is really bad.
• 19:02
  Speaker 1: NVIDIA has sold $ 215. 9 billion in their last fiscal year
• 19:07
  Speaker 1: worth of GPUs mostly. And you try and go, yeah,
• 19:10
  Speaker 1: that's to support like $ 22 billion of revenue total in
• 19:16
  Speaker 1: the entire world outside of these two companies that literally
• 19:19
  Speaker 1: require money being fed into them, sometimes by NVIDIA to
• 19:22
  Speaker 1: keep alive. When you tell people that, they go, well,
• 19:25
  Speaker 1: companies just.
• 19:26
  Speaker 2: Lose money, right?
• 19:26
  Speaker 1: Companies, because we have this, quote Ed Elson from Profiteer Markets,
• 19:30
  Speaker 1: we have this cult-like worship of the wealthy, where we
• 19:33
  Speaker 1: think that someone wouldn't spend all this money for no reason, right?
• 19:36
  Speaker 1: Because reconciling with that, with this idea that the ultra-wealthy,
• 19:42
  Speaker 1: the ultra-powerful didn't get there through big brains, they didn't
• 19:45
  Speaker 1: get there through anything other than luck and opportunity and
• 19:50
  Speaker 1: getting an MBA perhaps of the right people that they
• 19:53
  Speaker 1: just got there because they're regular people and they just
• 19:55
  Speaker 1: happen to be in the right place at the right time.
• 19:57
  Speaker 1: Reconciling with that and realizing that the world is not
• 19:58
  Speaker 1: controlled by people like a meritocracy is kind of grim.
• 20:02
  Speaker 2: So it's easy to be like, no.
• 20:04
  Speaker 1: They're not making a mistake. I must be missing something.
• 20:06
  Speaker 1: And that's what they want.
• 20:08
  Speaker 2: So, you know, I think back through the history of
• 20:10
  Speaker 2: technological breakthroughs and i think about i mean you can
• 20:13
  Speaker 2: look at different industries and one of my favorite books
• 20:15
  Speaker 2: on this subject is the innovators dilemma i read it
• 20:18
  Speaker 2: and one of the things it talks about is how
• 20:19
  Speaker 2: the the innovation that ends up taking out or transforming
• 20:23
  Speaker 2: an industry often starts worse doesn't make economic sense none
• 20:29
  Speaker 2: of your customers are asking for it and this is
• 20:31
  Speaker 2: typically why we end up ignoring it so like you've
• 20:33
  Speaker 2: got horse and carriages in the 1800s amazing form of
• 20:37
  Speaker 2: transport according to the 1800s you know people of the
• 20:39
  Speaker 2: 1800s And then you have this thing called cars come along. Now,
• 20:42
  Speaker 2: the problem with cars is they broke down all the time.
• 20:44
  Speaker 2: It's kind of like AI hallucinates now. They were more
• 20:46
  Speaker 2: expensive and the economics of it didn't make sense. You
• 20:49
  Speaker 2: might as well walk than buy a car. There was
• 20:51
  Speaker 2: a law at the time that meant you had to
• 20:52
  Speaker 2: walk in front of it with a red flag and
• 20:54
  Speaker 2: someone had to employ someone to walk in front of
• 20:56
  Speaker 2: it waving a red flag. Obviously, it's worse. It's like
• 20:59
  Speaker 2: a worse solution. However, these things that are disruptive innovations,
• 21:03
  Speaker 2: they have a higher ceiling of growth. And so they
• 21:07
  Speaker 2: eventually overtake the horse. And when I think about that
• 21:10
  Speaker 2: analogy in the context of all of this, I go, okay,
• 21:13
  Speaker 2: it's imperfect at the moment. The economic models aren't perfectly
• 21:17
  Speaker 2: ironed out. They're still figuring out how to make it cheaper,
• 21:20
  Speaker 2: the infrastructure, etc. But if you think about the rate
• 21:23
  Speaker 2: of improvement versus other, you know, let's say coding, how
• 21:27
  Speaker 2: much could I train a human coder to improve and
• 21:30
  Speaker 2: to increase their output versus an AI agent? One would go,
• 21:33
  Speaker 2: if you just imagine any rate of improvement in these
• 21:36
  Speaker 2: AI tools... at some point, if you just imagine a
• 21:38
  Speaker 2: 5% rate of improvement per month, at some point, it's...
• 21:43
  Speaker 2: And then you imagine a 5% reduction in cost, which
• 21:46
  Speaker 2: is what we did with the internet, what we did
• 21:47
  Speaker 2: with cars.
• 21:48
  Speaker 1: Yeah, but that.
• 21:48
  Speaker 2: Moore's law. Moore's law is Ethereum.
• 21:50
  Speaker 1: Moore's law is not with GPUs. So let me actually explain. So, NVIDIA.
• 21:55
  Speaker 1: NVIDIA invented, I think it was in the 2000s, they
• 21:57
  Speaker 1: put out something called CUDA, which is the underlying... Software
• 22:00
  Speaker 1: library and the way to run software on GPUs. It
• 22:04
  Speaker 1: took them a solid decade or more to make it
• 22:06
  Speaker 1: something where they could do data analytics, one of the
• 22:08
  Speaker 1: early things, mapper and such. And then when AI came along,
• 22:12
  Speaker 1: they'd had lots of experience with it. But nevertheless, this
• 22:14
  Speaker 1: company has got more money, more attention, more geniuses behind them,
• 22:20
  Speaker 1: more people focused on making their things more efficient. than
• 22:23
  Speaker 1: anyone could ever ask for.
• 22:25
  Speaker 2: And NVIDIA, for anyone that doesn't know, makes the chips.
• 22:28
  Speaker 1: And that CUDA thing I mentioned, they were the ones
• 22:30
  Speaker 1: with CUDA, and CUDA allowed generative AI to grow.
• 22:32
  Speaker 2: Okay, so they're chips.
• 22:34
  Speaker 1: Chips, yes.
• 22:34
  Speaker 2: And chips are needed. Those are the things that go
• 22:36
  Speaker 2: into the data centers.
• 22:37
  Speaker 1: And their specific chips are the ones where you can
• 22:39
  Speaker 1: run AI software on it. So the training runs and
• 22:41
  Speaker 1: also the inference. Now, here's the thing. The car example.
• 22:45
  Speaker 1: Back then, you didn't have pretty much every mathematician and
• 22:49
  Speaker 1: scientist going into the car industry. You didn't have the
• 22:52
  Speaker 1: combined world's governments never shutting.
• 22:53
  Speaker 2: Up about this.
• 22:54
  Speaker 1: And by the way, giving them credit early, since 2023,
• 22:59
  Speaker 1: they've been saying this is inevitable. Even in what you said,
• 23:01
  Speaker 1: 5% improvement. I don't even know who you'd measure that
• 23:04
  Speaker 1: because a junior software engineer can still experience things and
• 23:08
  Speaker 1: learn things from context, from how people deal with problems.
• 23:12
  Speaker 1: And the way that people deal with problems is not
• 23:14
  Speaker 1: as simple as looking at the code or reading some emails.
• 23:17
  Speaker 1: It's context cues from speaking to a person. It's being
• 23:19
  Speaker 1: in different environments. And there are uses for LLMs in coding.
• 23:24
  Speaker 1: I don't dispute that. But even saying 5%, what does
• 23:28
  Speaker 1: that mean? Is it better at Rust? Is it better
• 23:30
  Speaker 1: at C + +?
• 23:30
  Speaker 2: I'd say productivity. So just like, yeah, shipped. If we
• 23:33
  Speaker 2: did it in the context of coding, it would be
• 23:34
  Speaker 2: like shipped code. That's the thing.
• 23:36
  Speaker 1: That would be like, he's the best writer in the
• 23:38
  Speaker 1: world because his newsletter is really long. That's an insane
• 23:41
  Speaker 1: way of valuing it. With coding, it would be, I mean,
• 23:45
  Speaker 1: it's even difficult to evaluate because it's, is the software
• 23:48
  Speaker 1: out there better?
• 23:48
  Speaker 3: Yeah.
• 23:49
  Speaker 1: Is actually a great way of evaluating it. And I
• 23:51
  Speaker 1: would say uniformly not. I would say the standard of
• 23:54
  Speaker 1: software across Google, Microsoft, Amazon, Meta especially, God, Meta's a monstrosity,
• 24:00
  Speaker 1: is worse. GitHub, someone posted on Twitter earlier today, we
• 24:04
  Speaker 1: should get a notification when GitHub is up rather than
• 24:06
  Speaker 1: when it's down because that would be more reliable. Microsoft's
• 24:09
  Speaker 1: one of the largest companies in the world and they
• 24:10
  Speaker 1: can barely wipe their own ass when it comes to GitHub.
• 24:13
  Speaker 1: The quality of software is going down Weirdly enough, as
• 24:16
  Speaker 1: more people use LLMs and more businesses demand, and I
• 24:20
  Speaker 1: really do mean demand, that people use these services.
• 24:23
  Speaker 2: So on this point of, if we go back to
• 24:25
  Speaker 2: this horse and carriage and car analogy, say that we're
• 24:28
  Speaker 2: at whatever point today, if you imagine any rate of
• 24:31
  Speaker 2: improvement in the technology, which we have seen since Chachapati
• 24:33
  Speaker 2: came out. I remember when Chachapati came out and I
• 24:35
  Speaker 2: was in Asia and I was there showing it to
• 24:37
  Speaker 2: my fiance. I was like, look, you can do this.
• 24:39
  Speaker 2: And it was hallucinating once in a while and getting
• 24:41
  Speaker 2: things wrong. I actually don't have that experience anymore. I
• 24:45
  Speaker 2: have moments where I believe its reasoning is weak, but
• 24:48
  Speaker 2: I don't have outright hallucinations anymore. See, I disagree. Give
• 24:54
  Speaker 2: me an example of what you define as a hallucination.
• 24:56
  Speaker 1: Okay, great one. So I have a Bloomberg terminal. The
• 24:58
  Speaker 1: very useful thing they have on there is AskBee. So
• 25:01
  Speaker 1: when you do a Bloomberg inquiry to look up what
• 25:03
  Speaker 1: we think NVIDIA's revenue is going to be next quarter,
• 25:06
  Speaker 1: it runs something called BQL, which is its own programming language. Now,
• 25:10
  Speaker 1: instead of having to learn that, You can just type
• 25:13
  Speaker 1: into Ask B and it will generate it and run
• 25:14
  Speaker 1: it for you. And so you get pulled up and
• 25:16
  Speaker 1: you know where the data is coming from. It deals
• 25:17
  Speaker 1: with hallucinations real well. The other day, I was like,
• 25:20
  Speaker 1: you know what? We'll get a little spicy. I'm going
• 25:22
  Speaker 1: to look up the growth rate of stocks of Microsoft, Google, Meta,
• 25:26
  Speaker 1: and Amazon over the course of five years, I think
• 25:29
  Speaker 1: it was. And I was about to copy-pasted it over
• 25:33
  Speaker 1: to something, looked it in Excel. It was right in
• 25:35
  Speaker 1: the newsletter. I went, Microsoft's stock's never been $ 575 a stock.
• 25:41
  Speaker 1: you know what, when it's a cute little thing like, oh,
• 25:43
  Speaker 1: it's a stock price and I can't accord it, it
• 25:44
  Speaker 1: was no harm, no foul, that's fine. But when you're
• 25:48
  Speaker 1: talking about, I don't know, like a transcribing tool for
• 25:51
  Speaker 1: a doctor or a financial model that a hedge fund
• 25:53
  Speaker 1: is dependent on, at that point, it becomes a little
• 25:57
  Speaker 1: more dangerous. And the thing is, a hallucination with a
• 26:01
  Speaker 1: software package, for example, refactoring a code base, and it
• 26:05
  Speaker 1: leaves the door open security-wise, or it just break something
• 26:09
  Speaker 1: and you i don't know maybe you've been vibe coding
• 26:11
  Speaker 1: for six months you haven't really been coding with your
• 26:13
  Speaker 1: own hands for a while maybe you've forgotten a few
• 26:15
  Speaker 1: things you had this slot to look for you know
• 26:18
  Speaker 1: what i'm not doing it and so the problem has
• 26:20
  Speaker 1: become multiplicative and i don't really know how you train
• 26:24
  Speaker 1: them out of that and they've certainly not succeeded so
• 26:28
  Speaker 1: on one hand they have got better but one of
• 26:31
  Speaker 1: the main ways they evaluate them getting better are benchmarks
• 26:34
  Speaker 1: that are adjusted specifically for large language models Because you
• 26:37
  Speaker 1: can't just have them do tasks. They've got better at that.
• 26:41
  Speaker 1: They found some tasks they can have them do on.
• 26:43
  Speaker 1: But even them like METR, M-E-T-R, they have this thing
• 26:45
  Speaker 1: where it's like, check out this chart. Look how much
• 26:47
  Speaker 1: better it's getting at running tasks. Wow, it can go
• 26:49
  Speaker 1: for an hour. And then you look, it's like, yeah,
• 26:51
  Speaker 1: and successfully completing them 50% of the time.
• 26:54
  Speaker 2: They have a hallucination leaderboard and it really focuses on
• 26:57
  Speaker 2: basic tasks. And it shows that the four-year trend, according
• 26:59
  Speaker 2: to historical data from the Victoria Hallucination Leaderboard, shows that
• 27:04
  Speaker 2: hallucination rates on simple summarization tasks have plummeted from around 21%, 21.8%
• 27:10
  Speaker 2: four years ago, down to 0.7% roughly on today's top
• 27:16
  Speaker 2: frontier models like Gemini and ChatGPT. Again, the point of
• 27:20
  Speaker 2: nuance here is that these are on simple tasks, which
• 27:22
  Speaker 2: is kind of what I've experienced. I've experienced that on
• 27:24
  Speaker 2: day-to-day things that hallucinates less. Again, rate of improvement thinking.
• 27:28
  Speaker 2: So if I just imagine the trajectory to continue There
• 27:32
  Speaker 2: is going to become a time where hallucinations become rarer
• 27:35
  Speaker 2: than they are today, increasingly. And also what I'd say
• 27:38
  Speaker 2: is when I think about other technologies, there's two more points.
• 27:41
  Speaker 2: Other technologies at their inception, when they first came to
• 27:43
  Speaker 2: the world, like the internet, also had technical difficulties. I
• 27:47
  Speaker 2: remember growing up with dial-up modems and I couldn't go
• 27:50
  Speaker 2: on the phone at the same time as going on
• 27:51
  Speaker 2: the internet. I'd have to stop RuneScape upstairs to... Go
• 27:54
  Speaker 2: on the phone. And you thought, this is crap. This
• 27:56
  Speaker 2: is crap. Technology's crap. I don't know, mate. I loved it. Yeah,
• 27:59
  Speaker 2: I know. It felt like magic. And then, in hindsight,
• 28:02
  Speaker 2: you go, wow, I now have Starlink and 5G internet
• 28:05
  Speaker 2: from my phone. It's unbelievable. You couldn't leave the house
• 28:08
  Speaker 2: with internet before. And that's what I mean by the
• 28:10
  Speaker 2: rate of improvement thinking. I'd say the last point is
• 28:13
  Speaker 2: we often compare AI to perfection. Right. Whereas that's not
• 28:18
  Speaker 2: actually the alternative. In the working world, like if I
• 28:22
  Speaker 2: wanted to do, let's say, a simple... writing task, I
• 28:25
  Speaker 2: should compare AI to my alternative alternative way of doing
• 28:29
  Speaker 2: that simple writing task, which is both measured in my time, right?
• 28:33
  Speaker 2: And my ability to hallucinate as a person who doesn't
• 28:35
  Speaker 2: know everything at all. If I'm hiring someone, an intern
• 28:40
  Speaker 2: who might also be prone to hallucination or have gaps
• 28:43
  Speaker 2: in their knowledge. So it's not actually like we're comparing,
• 28:45
  Speaker 2: we should compare AI to perfection. It's AI to the
• 28:47
  Speaker 2: other alternatives. And if someone hallucinates 0.7% of the time,
• 28:50
  Speaker 2: but knows way more, and is faster maybe on a
• 28:55
  Speaker 2: net basis that's a good trade maybe i should use ai.
• 28:58
  Speaker 1: So let's start with an example someone i love dearly
• 29:02
  Speaker 1: matt hughes my editor yeah this is out of liverpool
• 29:05
  Speaker 1: wonderful guy i don't pay matt hughes because he knows
• 29:08
  Speaker 1: everything i pay him because he has incredible context and
• 29:12
  Speaker 1: a ton of knowledge and he's willing to expand it
• 29:14
  Speaker 1: and work with me and moral support and he's a
• 29:17
  Speaker 1: great editor but he's also someone who gets into the
• 29:20
  Speaker 1: guts of it and has the experiences of it he's
• 29:22
  Speaker 1: a Decorated tech journalist. And on top of that, a wonderful,
• 29:25
  Speaker 1: loving being with empathy and joy in his heart for
• 29:29
  Speaker 1: the stuff he loves and absolute fucking venom for the
• 29:31
  Speaker 1: people he hates. I can't get that from a large
• 29:34
  Speaker 1: language model. But on top of that, I don't, I
• 29:36
  Speaker 1: push back on just the assumption that when you say
• 29:39
  Speaker 1: knows everything. what good is something that knows everything when
• 29:42
  Speaker 1: it sometimes doesn't know anything when it's sometimes and on
• 29:45
  Speaker 1: the thing is are you really paying an intern for
• 29:47
  Speaker 1: something basic are you really going to them and saying
• 29:50
  Speaker 1: yeah can you look up what the date is no
• 29:52
  Speaker 1: you're doing that on google whatever the task is you
• 29:55
  Speaker 1: are trying to also train an intern the point of
• 29:58
  Speaker 1: an intern is to train them and turn them in
• 29:59
  Speaker 1: take them out of pinocchio status but it's also an
• 30:03
  Speaker 1: intern learns an intern gets context an intern learns your habits.
• 30:06
  Speaker 2: Gets context and learns no it doesn't it doesn't learn.
• 30:09
  Speaker 1: I mean it doesn't the way it learns is you
• 30:12
  Speaker 1: create a giant claude the md file that it sometimes
• 30:15
  Speaker 1: doesn't read sometimes does read you create a harness you
• 30:17
  Speaker 1: put it's like it's peewee's breakfast machine from peewee's playoffs
• 30:21
  Speaker 1: you have to do all these contrivances to mitigate the
• 30:24
  Speaker 1: hallucinations even then at the end How much effort have
• 30:27
  Speaker 1: you put in?
• 30:28
  Speaker 2: Okay, this is an extreme simplified example. If I went
• 30:31
  Speaker 2: on my cloud now and said, what's my dog's name?
• 30:33
  Speaker 2: It would know my dog's name. Jesus Christ.
• 30:36
  Speaker 1: This company raised $ 95.
• 30:37
  Speaker 2: Billion this year. I'm using an extreme simplified example to
• 30:41
  Speaker 2: show that it can remember things from the past. Obviously,
• 30:43
  Speaker 2: it knows much more complex things as well, but I
• 30:46
  Speaker 2: just use that as an example. So we accept the
• 30:48
  Speaker 2: fact that it does have memory of the past.
• 30:51
  Speaker 1: It has files it can access that have stuff on it,
• 30:53
  Speaker 1: but that's not the same as memory. And it's also just, okay,
• 30:57
  Speaker 1: so it remembers your dog's name. It might remember your habits.
• 31:00
  Speaker 1: It might be able to read things you've said before.
• 31:03
  Speaker 1: Does it know your moods? Does it know what's going
• 31:05
  Speaker 1: on in the world around it? Does it have good
• 31:07
  Speaker 1: days and bad days? Is it there for you? Because
• 31:09
  Speaker 1: it's just a fucking text machine. And the thing is,
• 31:12
  Speaker 1: the intern example. An intern is something that can grow.
• 31:16
  Speaker 1: It's something that you invest in. That's not something you
• 31:18
  Speaker 1: do through feeding files and text to it. The way
• 31:21
  Speaker 1: that we store memories ourselves, the way in which we
• 31:24
  Speaker 1: accrue experiences is a, a milestone of emotion and feelings
• 31:30
  Speaker 1: and facts.
• 31:30
  Speaker 2: Completely different. So I think there's two things here. There's
• 31:33
  Speaker 2: the process in which something happens and then there's the output.
• 31:37
  Speaker 2: So the process you were describing, the process of how
• 31:39
  Speaker 2: a human does memory, the way that an AI does
• 31:42
  Speaker 2: memory is different, but the thing that people care about
• 31:45
  Speaker 2: is their value in the output. IE, you know, if
• 31:48
  Speaker 2: I dump all of my files into Claude, I don't
• 31:51
  Speaker 2: really care how it processes it. As long as when
• 31:53
  Speaker 2: I ask it, what's my revenue it has the number
• 31:56
  Speaker 2: and one could say the same thing about training someone
• 31:58
  Speaker 2: you could say you teach them you put lots of
• 32:00
  Speaker 2: effort into them you give them lots of context you
• 32:03
  Speaker 2: educate them and give them experiences and then you might
• 32:05
  Speaker 2: come and say to them by the way what's my
• 32:07
  Speaker 2: revenue now the processes are entirely different but the outcome
• 32:10
  Speaker 2: is what i care about do they know the revenue
• 32:11
  Speaker 2: number when i ask them and so i think that's
• 32:14
  Speaker 2: the part that we sometimes get lost and we get
• 32:15
  Speaker 2: you know because i have i've heard this debate about
• 32:17
  Speaker 2: like can ai be creative right i think like the
• 32:20
  Speaker 2: way to answer that question is like It's about the output.
• 32:23
  Speaker 2: When I ask it to do a creative thing, does
• 32:25
  Speaker 2: it give me the answer? Not, is the process the
• 32:28
  Speaker 2: same as a human process? Because actually, who cares what
• 32:31
  Speaker 2: the product... People care about... They pay for the outcome,
• 32:33
  Speaker 2: the product.
• 32:34
  Speaker 1: I actually disagree about the process because Matt Hughes, for example...
• 32:39
  Speaker 1: Your editor.
• 32:40
  Speaker 2: Yeah.
• 32:41
  Speaker 1: Watching him go down a rabbit hole and being there
• 32:43
  Speaker 1: with him, and actually vice versa, him doing the same thing.
• 32:46
  Speaker 1: We wrote these... Well, I mean, we were working on
• 32:49
  Speaker 1: the research. I ended up sitting there for like the
• 32:51
  Speaker 1: day-long session of writing 11 000 words and he he
• 32:54
  Speaker 1: had given me a bunch of notes it was actually
• 32:56
  Speaker 1: just even describing that process i feel so happy because
• 32:59
  Speaker 1: it was like us being like i can't believe how
• 33:01
  Speaker 1: fuck these jesus christ they can't like just like the
• 33:04
  Speaker 1: misanthropy of just the horrible cynical people of asset managers
• 33:08
  Speaker 1: like blackstone just learning about them and being like it
• 33:10
  Speaker 1: can't be this and having a back and forth with
• 33:12
  Speaker 1: him that is fundamentally different because we were both learning
• 33:15
  Speaker 1: together And the learning process was as much about creating
• 33:18
  Speaker 1: the output as the output itself. When you learn something,
• 33:21
  Speaker 1: you're not creating the average, which really is what these
• 33:24
  Speaker 1: things do, of the documents it could find. You're not
• 33:28
  Speaker 1: getting particularly novel outputs. If I needed a generic slop output, sure.
• 33:34
  Speaker 1: But I've used some of the higher-end LLM harness machines
• 33:38
  Speaker 1: that the hedge funds use, and they all get the
• 33:40
  Speaker 1: same shite.
• 33:41
  Speaker 2: It's all the same.
• 33:42
  Speaker 1: The same generic reports, the same, oh, we noticed this analysis,
• 33:45
  Speaker 1: things that you can find on any kind of AI
• 33:47
  Speaker 1: slop out there.
• 33:48
  Speaker 2: What you described to me there, what I heard anyway,
• 33:50
  Speaker 2: is there's two points of value you're getting from your
• 33:52
  Speaker 2: time with that. I mean, there's many more, but you
• 33:55
  Speaker 2: said you're learning, and then you're getting this book edited. Blog. Blog.
• 33:59
  Speaker 2: You're getting a blog edited, which is the output, and
• 34:01
  Speaker 2: you're getting learning, and you're also really getting connection and
• 34:04
  Speaker 2: all these other things. But when people sort of think
• 34:07
  Speaker 2: about the value of AI, of course they could use
• 34:08
  Speaker 2: it to learn. But in the example I gave of
• 34:10
  Speaker 2: like repeat my revenue number back to me or do
• 34:12
  Speaker 2: this number, I just care about the output. I could
• 34:14
  Speaker 2: use it to learn. I could say as far with me.
• 34:16
  Speaker 1: What if the revenue number was wrong once? You should
• 34:18
  Speaker 1: have defined deterministic ways of knowing those numbers. You should
• 34:23
  Speaker 1: not rely on them. Even with the terminal running BQL,
• 34:26
  Speaker 1: which I trust, I will double, triple, triple check everything
• 34:29
  Speaker 1: just to be sure. Partly because also the process of
• 34:32
  Speaker 1: learning for me, I don't want just a report. I
• 34:35
  Speaker 1: go like that. I want something that I fully understand.
• 34:37
  Speaker 1: And also- understand the context around it. I don't think
• 34:41
  Speaker 1: that LLMs do that. And I just don't see them
• 34:44
  Speaker 1: getting better in a way that does that because it's,
• 34:48
  Speaker 1: it's just not what they do. And also there's the
• 34:50
  Speaker 1: other problem of the more detailed the report, the more
• 34:53
  Speaker 1: likely there are things to be wrong with it. If
• 34:54
  Speaker 1: you are with Matt Hughes, for example, I can trust
• 34:57
  Speaker 1: he's got it right. I can trust he understood. And
• 35:00
  Speaker 1: I can trust that I can have a back and
• 35:01
  Speaker 1: forth with him that will inform me if I've missed something.
• 35:04
  Speaker 1: I can read the stuff that he's read and actually
• 35:07
  Speaker 1: trust trust him because there's a big trust part as well.
• 35:10
  Speaker 2: What is the basis of your trust in Matt? Could
• 35:12
  Speaker 2: it be his historical performance?
• 35:16
  Speaker 1: I mean, yes.
• 35:17
  Speaker 2: Okay.
• 35:17
  Speaker 1: And also the fact we've learned half of this stuff together.
• 35:20
  Speaker 2: But tenure doesn't necessarily... There's probably people you know for
• 35:24
  Speaker 2: 15 years who you also don't trust. Yes. So I
• 35:26
  Speaker 2: think I was trying to figure out what is the
• 35:27
  Speaker 2: thing that's causing humans to trust another thing? And I
• 35:30
  Speaker 2: guess it would be continual delivery of a commitment made
• 35:33
  Speaker 2: of sorts. And so with clawed for example on simple
• 35:37
  Speaker 2: tasks as we've seen from this hallucination leaderboard it continually
• 35:41
  Speaker 2: delivers for people and that's why we've seen the fastest
• 35:44
  Speaker 2: i mean is that what that board says well it's
• 35:46
  Speaker 2: saying like is it getting it wrong is it hallucinating
• 35:48
  Speaker 2: simple tasks simple tasks.
• 35:50
  Speaker 1: How are those defined i don't know that's the thing
• 35:52
  Speaker 1: though because this is actually a very illustrative thing of
• 35:55
  Speaker 1: the ai industry they are the whataboutist masters they have
• 36:00
  Speaker 1: like Well, look, we've got this benchmark that says we're
• 36:03
  Speaker 1: good at this. And look, the number's higher. What's the
• 36:04
  Speaker 1: number mean? What does that mean? And I'm not using
• 36:08
  Speaker 1: this as a critic against you. It's when you can't
• 36:11
  Speaker 1: give a direct answer, you give a side answer. When
• 36:14
  Speaker 1: you as the LLM industry want to prove your worth,
• 36:16
  Speaker 1: you can't just be like, just use the product. When
• 36:18
  Speaker 1: the first iPhone came out, I've got Penn State at
• 36:21
  Speaker 1: the time. Oh, I felt like the eights at the
• 36:23
  Speaker 1: beginning of 2001. Oh, fucking visual voicemail.
• 36:27
  Speaker 2: It was immediate.
• 36:27
  Speaker 1: And I showed it to tech friends I showed it
• 36:30
  Speaker 1: to the most normal people in the world and everyone
• 36:32
  Speaker 1: was like holy shit this is they were on razors
• 36:34
  Speaker 1: they were on Nokia 3210s it was obvious the value
• 36:37
  Speaker 1: Amazon web services same deal it wasn't obvious though yes
• 36:40
  Speaker 1: it was I mean I bought it to you it
• 36:43
  Speaker 1: was it was and I also showed it to a
• 36:44
  Speaker 1: bunch of people because I'm aware that I have bias
• 36:47
  Speaker 1: when I just love gadgets.
• 36:48
  Speaker 2: But I remember the famous Steve Ballmer who was the
• 36:51
  Speaker 2: CEO of Microsoft interview where he was told about the
• 36:54
  Speaker 2: iPhone and he bursts out laughing $ 500.
• 37:01
  Speaker 3: Fully subsidized with a plan? I said, that is the
• 37:04
  Speaker 3: most expensive phone in the world, and it doesn't appeal
• 37:07
  Speaker 3: to business customers because it doesn't have a keyboard, which
• 37:10
  Speaker 3: makes it not a very good email machine. You can
• 37:13
  Speaker 3: get a Motorola Q phone now for $ 99. It's a
• 37:17
  Speaker 3: very capable machine. It'll do music. It'll do internet. It'll
• 37:21
  Speaker 3: do email. It'll do instant messaging. So I kind of
• 37:25
  Speaker 3: look at that, and I say, well, I like our strategy.
• 37:28
  Speaker 2: I like it a lot. He burst out laughing, mocking it.
• 37:32
  Speaker 2: Because it was so disruptive. It was way more expensive
• 37:35
  Speaker 2: and it was way different. No keyboard.
• 37:37
  Speaker 1: Well, phones used to be insanely expensive and the carriers
• 37:40
  Speaker 1: would cover them, but you had to sign a long contract.
• 37:42
  Speaker 2: You were still spending 500 bucks.
• 37:43
  Speaker 1: But the thing I'm getting at is you didn't have
• 37:45
  Speaker 1: to explain to someone what, perhaps you'd have to get
• 37:47
  Speaker 1: past the cost part, but you could just be like,
• 37:49
  Speaker 1: look how good this is. And then once the app
• 37:51
  Speaker 1: was the iPhone 3G with the app store, people were like,
• 37:53
  Speaker 1: oh shit, this could actually change things. Mobile web. even
• 37:56
  Speaker 1: though it was a monstrosity. It was so bad at first.
• 38:00
  Speaker 1: Even then, you could get your emails and you could
• 38:01
  Speaker 1: just look at them. But as BlackBerrys were also expensive
• 38:03
  Speaker 1: and were still actually kind of cool, but the way
• 38:05
  Speaker 1: they worked was not like consumer software. They didn't have
• 38:08
  Speaker 1: the classic GUI. iPhones felt like that. It felt like
• 38:10
  Speaker 1: a cell phone designed even like a computer.
• 38:14
  Speaker 2: It was obvious.
• 38:15
  Speaker 1: It was obvious in the beginning. Everyone, I was dating
• 38:18
  Speaker 1: a girl in the center of Pennsylvania at the time
• 38:20
  Speaker 1: and everyone I showed it to was like, wow, this
• 38:21
  Speaker 1: is incredible. That to me is the obvious thing. With
• 38:25
  Speaker 1: AI to this day when you're like okay why is
• 38:27
  Speaker 1: it so amazing people still dither people are still like
• 38:30
  Speaker 1: yeah you can't run a business fully with it without
• 38:33
  Speaker 1: this weird system of pulleys and levers and such but.
• 38:38
  Speaker 2: How come then when you look at the stats around
• 38:40
  Speaker 2: chat gbt's growth 100 million active users in just the
• 38:45
  Speaker 2: first 60 days after launching for comparison tiktok took nine
• 38:48
  Speaker 2: months instagram took 2.5 years And the internet itself, the
• 38:51
  Speaker 2: World Wide Web, took roughly seven years to reach that scale.
• 38:54
  Speaker 2: Over 60% of the US adults are integrated into AI
• 38:57
  Speaker 2: tools in their daily and regular routines within three years
• 39:00
  Speaker 2: of the launch, reaching 40% of the population. And that
• 39:04
  Speaker 2: same milestone took the internet five years and personal computers
• 39:07
  Speaker 2: nearly 12.
• 39:07
  Speaker 1: Okay.
• 39:09
  Speaker 2: So I think this is the part that's giving me dissonance.
• 39:11
  Speaker 2: It's like when I showed my fiancee ChatGPT, okay, it
• 39:14
  Speaker 2: didn't really work. But as a sole entrepreneur who English
• 39:18
  Speaker 2: isn't her first language, who has to write lots of
• 39:21
  Speaker 2: texts and lots of copy and generate lots of images
• 39:23
  Speaker 2: and was paying a graphic designer to help her make
• 39:25
  Speaker 2: certain images that she couldn't make herself because she doesn't
• 39:28
  Speaker 2: have the skills. She would describe it as being transformative
• 39:34
  Speaker 2: for her business. What I'm hearing from you is that
• 39:36
  Speaker 2: it's not transformative and there's no value in it for people.
• 39:39
  Speaker 2: But if she was sat here now, she'd get transformative.
• 39:42
  Speaker 1: Would she pay the per million token rate? Would she
• 39:45
  Speaker 1: pay the actual rate? Because that's the thing. If this
• 39:47
  Speaker 1: was sold at its honest cost, I would actually... And
• 39:50
  Speaker 1: people were reacting like that and they were paying $ 2, $ 3,
• 39:52
  Speaker 1: $ 4 every time they did something and they were genuinely happy.
• 39:55
  Speaker 2: That might be an argument. What would be the honest
• 39:57
  Speaker 2: cost if they weren't subsidizing?
• 39:58
  Speaker 1: The actual per million token cost. The actual API cost.
• 40:01
  Speaker 1: Do you know how.
• 40:02
  Speaker 2: Much that is relative to what they charge? Oh, God.
• 40:05
  Speaker 1: It depends on the model. But there's... Actually... kind of
• 40:08
  Speaker 1: a point i want to make about the thing you
• 40:10
  Speaker 1: said with the internet earlier so when i first got
• 40:12
  Speaker 1: on the internet a 33.4 kilobits a second modem even
• 40:16
  Speaker 1: back then i was like fuck if this was faster
• 40:19
  Speaker 1: and that was like immediate just like if this was
• 40:21
  Speaker 1: faster because it was slow you go on like happy
• 40:23
  Speaker 1: puppy or something download take all bloody day waiting for
• 40:27
  Speaker 1: shareware to download immediately like if i could do this
• 40:29
  Speaker 1: faster it'd be better and even back then i'm like
• 40:31
  Speaker 1: man you could probably do video camera stuff with this
• 40:34
  Speaker 1: stuff that eventually happened. And actually, there's this guy called
• 40:36
  Speaker 1: Jim Covello from Goldman Sachs in a report he did
• 40:39
  Speaker 1: in 2024 that was Gen AI, too much spend for
• 40:41
  Speaker 1: not enough return. Paraphrasing there. And he made the point
• 40:44
  Speaker 1: that in the run-up to the iPhone, there was thousands
• 40:47
  Speaker 1: of presentations that when GSM radios get smaller, when Bluetooth
• 40:50
  Speaker 1: radios get smaller, when Wi-Fi radios get smaller, it is
• 40:53
  Speaker 1: inevitable that we will get something like this. And then
• 40:57
  Speaker 1: he said that there is no such path for AI.
• 41:00
  Speaker 1: There was no roadmap to AI becoming this thing that
• 41:03
  Speaker 1: they promised. And I must be clear, if these companies
• 41:06
  Speaker 1: had gone out there and they're like, yeah, this is
• 41:08
  Speaker 1: interesting cloud software. It's generative. It's really expensive. We're not
• 41:12
  Speaker 1: sure if we can fully not trust it, not in
• 41:15
  Speaker 1: the oh, I'm scared way. I mean, just like, we're
• 41:17
  Speaker 1: not sure that this is going to be a disruptive
• 41:20
  Speaker 1: world changing thing. It has potential, but we're going to
• 41:22
  Speaker 1: go slow. It's really expensive. This is an R &
• 41:25
  Speaker 1: D effort. We're not going to expose consumers to it
• 41:28
  Speaker 1: and actually beat like, I don't know, language models and
• 41:31
  Speaker 1: no No generative AI stuff. Not even calling because it
• 41:34
  Speaker 1: isn't AI. It's not autonomous. It's not smart. I actually
• 41:38
  Speaker 1: might respect it, but this is not. They've gone out
• 41:40
  Speaker 1: there since 2023 and said, it was 2022, this is
• 41:43
  Speaker 1: the best thing since sliced bread. This is changing everything.
• 41:46
  Speaker 1: This is going to do all your work. This is
• 41:47
  Speaker 1: going to take your job. You're going to talk to
• 41:50
  Speaker 1: Bing and it's going to tell you to leave your wife.
• 41:51
  Speaker 1: All of these crazy things. And what's funny is when
• 41:55
  Speaker 1: the writer, Kevin Roos, I think it was, he was
• 41:58
  Speaker 1: speaking to Kevin Scott, the CTO of Microsoft, about it.
• 42:00
  Speaker 1: And Kevin Scott goes, you know, I'm just glad we're
• 42:02
  Speaker 1: having this conversation. Instead of being like, settle down, Beavis.
• 42:06
  Speaker 1: It's a website. The website told you something. It's just LLMs.
• 42:09
  Speaker 2: They talked it up.
• 42:10
  Speaker 1: And that's because everyone is talking about what they wish
• 42:13
  Speaker 1: this was rather than talking about what it can actually do.
• 42:16
  Speaker 1: This makes it scary to people, deliberately so. It makes
• 42:19
  Speaker 1: it environmentally destructive. Look at the gas turbines poisoning black neighborhoods.
• 42:23
  Speaker 1: I think it's in Louisiana. It's one of Musk's data centers.
• 42:26
  Speaker 1: Look at the incredible energy draws. It is raising power bills.
• 42:30
  Speaker 1: And also, It is creating inflation across all consumer electronics
• 42:34
  Speaker 1: because of the massive ramp.
• 42:35
  Speaker 2: Do you know what's interesting? I almost feel like so
• 42:39
  Speaker 2: much of what you're saying is true. And also it
• 42:43
  Speaker 2: can be true that this technology is going to profoundly
• 42:46
  Speaker 2: change the world. And I think like, you know, I
• 42:49
  Speaker 2: think back to the early days of the internet is
• 42:52
  Speaker 2: maybe the closest analogy we have of, you know, in
• 42:54
  Speaker 2: the dot-com bubble, you know, you wrote this great essay. Yes.
• 42:57
  Speaker 1: Yes.
• 42:57
  Speaker 2: I wish I found really funny. Um, especially the name.
• 43:00
  Speaker 2: The rot economy, and you talked about the rot-com bubble.
• 43:04
  Speaker 1: Yes.
• 43:05
  Speaker 2: Talking about how AI is less value than people think.
• 43:10
  Speaker 2: And in that, in the sort of dot-com bubble, what
• 43:11
  Speaker 2: you saw is huge hype, people overselling the capabilities of
• 43:15
  Speaker 2: their websites and what they were building. But in the
• 43:19
  Speaker 2: wake of the dot-com bubble, yes, 90% of stuff went
• 43:22
  Speaker 2: to zero, but you had people generational companies born that
• 43:26
  Speaker 2: changed the world right and so i i do i
• 43:29
  Speaker 2: kind of and that's what bubbles do right but that
• 43:31
  Speaker 2: huge high it was complete over investment investors get crazy
• 43:34
  Speaker 2: delusional they think it's everything's going to change at the
• 43:37
  Speaker 2: same time you do have skeptics in these moments the
• 43:41
  Speaker 2: the dot-com bubble had i mean the internet itself had
• 43:43
  Speaker 2: the biggest skeptics in 1998 nobel prize winning economist paul
• 43:48
  Speaker 2: krugman said By 2005 or so, it will become clear
• 43:52
  Speaker 2: that the internet's impact on the economy has been no
• 43:54
  Speaker 2: greater than the fax machine. In 1995, astrophysicist Clifford Stuhl, famously,
• 43:59
  Speaker 2: I wrote about this in my book, wrote famously in Newsweek,
• 44:03
  Speaker 2: Do our computer pundits lack all common sense? The truth is,
• 44:07
  Speaker 2: no online database will replace your daily newspaper. No CD-ROM
• 44:11
  Speaker 2: can take the place of a competent teacher. Commerce and
• 44:14
  Speaker 2: businesses will shift from offices and malls to networks and modems? Baloney!
• 44:18
  Speaker 2: So how come my local mall, does a roaring business,
• 44:22
  Speaker 2: and the cyber mall gets zero business. And then I'll
• 44:25
  Speaker 2: give you one more from Krugerman, who was the award-winning economist.
• 44:29
  Speaker 2: He said, the growth of the internet will slow drastically.
• 44:32
  Speaker 2: As it becomes apparent, most people have nothing to say
• 44:34
  Speaker 2: to each other.
• 44:35
  Speaker 1: That may actually be the worst point of those predictions.
• 44:40
  Speaker 1: Hang around any bar in middle America. honestly the best conversation.
• 44:44
  Speaker 1: But it's just all the same thing.
• 44:45
  Speaker 2: I actually, so Clifford.
• 44:46
  Speaker 1: Stoll actually, his piece was interesting because there were some
• 44:49
  Speaker 1: boner points in it, but he made points about how
• 44:51
  Speaker 1: like an overwhelming amount of bad information out there is
• 44:53
  Speaker 1: bad for society. He's completely right. Saying how online education
• 44:57
  Speaker 1: would not be a great replacement for regular education. I
• 45:00
  Speaker 1: think we've seen that, but there is an economic difference
• 45:03
  Speaker 1: that's vastly, it's just completely different.
• 45:06
  Speaker 2: So.
• 45:06
  Speaker 1: Com bubble was actually two bubbles. There was the website bubble,
• 45:09
  Speaker 1: which was just trash on trash on trash. It was
• 45:11
  Speaker 1: just like, I think, what was it? Excite at home
• 45:14
  Speaker 1: bought a e-giri in card company for like a billion dollars.
• 45:18
  Speaker 1: It was insane crap happening. That was so small. The
• 45:21
  Speaker 1: big thing that people are thinking about is the dark fiber.
• 45:24
  Speaker 2: Dark fiber.
• 45:25
  Speaker 1: Dark fiber was all of the wires that put in
• 45:28
  Speaker 1: the ground thinking we're going to have all this demand
• 45:29
  Speaker 1: for internet. And it turned out that demand for internet,
• 45:33
  Speaker 1: I think the analyst estimate was it was doubling every
• 45:36
  Speaker 1: 90 days when it was doing that every six to
• 45:38
  Speaker 1: 12 months, maybe longer.
• 45:40
  Speaker 2: Yeah.
• 45:41
  Speaker 1: And just thus there was a massive overbuild of fiber
• 45:43
  Speaker 1: optic cable, and indeed the transmission stations and such to
• 45:47
  Speaker 1: simplifying to bring that to people's houses. And there was
• 45:50
  Speaker 1: the assumption that, well, that would all get lit up
• 45:53
  Speaker 1: and people would want it immediately. Didn't really happen. Now,
• 45:55
  Speaker 1: the post.com bubble thing people say is, well, but after that,
• 45:58
  Speaker 1: there was demand from the internet. That's the thing though.
• 46:01
  Speaker 1: That's very different to demand for generative AI. Right now,
• 46:05
  Speaker 1: the demand we have for generative AI is predominantly subsidized,
• 46:08
  Speaker 1: just Let's start there. Yeah. Predominantly subsidized and most people
• 46:12
  Speaker 1: experiencing it are not paying the real cost.
• 46:14
  Speaker 2: I agree.
• 46:15
  Speaker 1: On top of that, we already have all of the
• 46:18
  Speaker 1: possible marketing in the world. We have the largest, most
• 46:22
  Speaker 1: disingenuous marketing campaign in the history of man pushing this
• 46:26
  Speaker 1: up the hill. We have the apex predator of cloud software, Microsoft.
• 46:31
  Speaker 1: They can only get single digit billions from selling AI software.
• 46:35
  Speaker 1: And Christ almighty, outside of OpenAI and Anthropic, we would
• 46:38
  Speaker 1: barely get $ 22 billion. And the thing is, $ 22 billion
• 46:41
  Speaker 1: is a large amount to you and me. It's not
• 46:43
  Speaker 1: a large amount of money when you spent a trillion
• 46:46
  Speaker 1: plus dollars, when you have Anthropic and OpenAI with $ 1. 1
• 46:49
  Speaker 1: trillion worth of cloud commitments. And on top of that,
• 46:53
  Speaker 1: how does this turn into a post.com bubble thing? A
• 46:55
  Speaker 1: data center built today is going to be as expensive
• 46:58
  Speaker 1: to run in 2050 as it is today, unless there's
• 47:01
  Speaker 1: some breakthrough in electricity. But again, That's not happening with AI.
• 47:05
  Speaker 1: AI is not doing that unless there's some breakthrough in
• 47:08
  Speaker 1: GPU technology. But we already have Broadcom, NVIDIA etched. We
• 47:12
  Speaker 1: have every major chip company arm trying to do something
• 47:16
  Speaker 1: about this. And no one seems to magically be able
• 47:18
  Speaker 1: to make this profitable or indeed even less costly. Even
• 47:22
  Speaker 1: NVIDIA with Vera Rubin, their more expensive new GPU system,
• 47:26
  Speaker 1: even then they're like, yeah, 10X more efficient. It's a
• 47:29
  Speaker 1: more dollars per megawatt. They're all coy about it. They
• 47:32
  Speaker 1: don't just say, yeah, we worked with OpenAI and Anthropic
• 47:34
  Speaker 1: and we found it reduced their cost by 50%. Easiest
• 47:37
  Speaker 1: thing in the world if it was true. And that's
• 47:39
  Speaker 1: because it's not happening. And this isn't a case where.
• 47:42
  Speaker 2: So you're saying there's not going to be the demand for...
• 47:45
  Speaker 2: let's say let's you know there's different types of ai
• 47:48
  Speaker 2: generative ai.
• 47:49
  Speaker 1: Yeah and actually that's a good point to make the
• 47:51
  Speaker 1: reason they use the term artificial intelligence is so everyone
• 47:54
  Speaker 1: would lump everything into it they would lump uh protein
• 47:57
  Speaker 1: folding nothing to do with llms robotics not llms autonomous
• 48:01
  Speaker 1: weapons even horrible as they are not llms because you
• 48:04
  Speaker 1: couldn't trust them but they've mushed everything into ai so
• 48:07
  Speaker 1: that when you say well ai can't they'll go um Um, sir,
• 48:12
  Speaker 1: you forgot to give us homework. And also, AI, it's
• 48:14
  Speaker 1: working on curing cancer when it's just like, no, that's
• 48:16
  Speaker 1: not LLMs. Stop giving them credit.
• 48:18
  Speaker 2: The similarity, though, is they all need GPUs.
• 48:20
  Speaker 1: No, and that's the funny thing. All those data centers
• 48:23
  Speaker 1: that we're building, all of them are for just generative AI.
• 48:27
  Speaker 1: They're not for all of the other stuff. They're not
• 48:30
  Speaker 1: for the cool shit. AI has been around for a
• 48:33
  Speaker 1: long time. Google, a lot of the good stuff that
• 48:35
  Speaker 1: comes out of Google from the search side is AI,
• 48:38
  Speaker 1: but pre-generative.
• 48:39
  Speaker 2: How would you run... the the type of ai that
• 48:42
  Speaker 2: sits in a robot let's say one of the optimist
• 48:45
  Speaker 2: robots if you didn't have a gpu.
• 48:47
  Speaker 1: So matic matic has this cleaning robot for example that
• 48:50
  Speaker 1: thing is not got a little gpu in it what
• 48:53
  Speaker 1: it has and may indeed have used some gpus but
• 48:56
  Speaker 1: now a year as many as they need for generative
• 48:59
  Speaker 1: ai to run the data feed draining data into it
• 49:01
  Speaker 1: so it's able to clean a house but when the
• 49:03
  Speaker 1: little bugger is going around cleaning my floor turdsly accord
• 49:05
  Speaker 1: if it goes around mopping my floor it's not like
• 49:08
  Speaker 1: burning money the whole time but When it comes to
• 49:11
  Speaker 1: these massive amount of data centers, Sightline Climate said in February,
• 49:14
  Speaker 1: there's 190 gigawatts of data centers in planning, don't know
• 49:18
  Speaker 1: about under construction. That works out for about 12 million megawatts, what,
• 49:22
  Speaker 1: like $ 1. 6 trillion to $ 3 trillion a year in annual
• 49:26
  Speaker 1: demand you'd need for that. We don't even have $ 130
• 49:29
  Speaker 1: billion worth of annual demand. And people say, well, it
• 49:31
  Speaker 1: will grow. How? When most of the demand is coming
• 49:34
  Speaker 1: from Amazon feeding money to open AI or Anthropic. Microsoft
• 49:38
  Speaker 1: feeding money to open AI and anthropic. Google feeding money
• 49:41
  Speaker 1: to open AI and anthropic. Well, Google hasn't fed it
• 49:43
  Speaker 1: to open AI yet, but they're a pretty big customer,
• 49:45
  Speaker 1: billions of dollars. The con side is that we are
• 49:49
  Speaker 1: building these effigies to capitalism, these giant GPU data centers,
• 49:53
  Speaker 1: and people are being told.
• 49:54
  Speaker 2: Well, it's for AI.
• 49:56
  Speaker 1: You know, the thing that's done all this other stuff
• 49:58
  Speaker 1: that's unrelated. Or the worst thing I've seen, it's like, oh,
• 50:01
  Speaker 1: you don't like you'd like online banking where you do
• 50:03
  Speaker 1: like data centers there's a big difference between a data
• 50:05
  Speaker 1: center for regular non-gpu compute for standing up a server
• 50:08
  Speaker 1: a content delivery system like a akamai or something that
• 50:11
  Speaker 1: brings the website to you or how meta runs facebook
• 50:15
  Speaker 1: that is not the same it takes way less power
• 50:17
  Speaker 1: mostly cpu driven compared to these giant gpu data centers
• 50:21
  Speaker 1: that offer one thing one thing.
• 50:23
  Speaker 2: But I was doing the research and looking at some
• 50:26
  Speaker 2: of these notes here. It does say that for tougher
• 50:28
  Speaker 2: types of AI systems designed to solve concrete physics, biology,
• 50:31
  Speaker 2: and spatial problems, they require some of the most intense
• 50:34
  Speaker 2: data center infrastructure on the planet. AI systems like DeepMind's AlphaFold,
• 50:39
  Speaker 2: the protein folding company, used for genomic sequencing and climate forecasting,
• 50:44
  Speaker 2: et cetera, run on high-performance computing clusters. These require immense
• 50:48
  Speaker 2: precision and continuous heavy computing data centers. Training the brains
• 50:52
  Speaker 2: for self-driving cars requires... billions of miles of simulated physics environments.
• 50:58
  Speaker 2: The AI isn't generating text. It's learning to navigate 3D
• 51:02
  Speaker 2: spaces and gravity and relies on data centers.
• 51:04
  Speaker 1: Right. And the thing is, those data centers, they might
• 51:07
  Speaker 1: have GPUs in them. We had GPUs used for this HPC,
• 51:11
  Speaker 1: the high-performance computing, before generative AI. And yeah, that's how
• 51:15
  Speaker 1: AI has been trained before. That's how Tesla did. I
• 51:18
  Speaker 1: believe they've had their own data centers when it... comes
• 51:20
  Speaker 1: to training the autopilot system for better or for worse
• 51:23
  Speaker 1: that's how we've done it before again that is not
• 51:26
  Speaker 1: why we're building these data centers these data centers are
• 51:28
  Speaker 1: being built to sell to ai generative ai companies to
• 51:32
  Speaker 1: either train systems or run inference these things are being
• 51:36
  Speaker 1: built in this brainless way where it's just well actually
• 51:40
  Speaker 1: maybe this is a good way of illustrating the con
• 51:43
  Speaker 1: because everyone saw google microsoft amazon and meta give nvidia
• 51:50
  Speaker 1: over call it 800 something billion dollars because everyone saw
• 51:55
  Speaker 1: that they went well they wouldn't do that for no
• 51:57
  Speaker 1: reason they went we've got to build more of these
• 51:58
  Speaker 1: things there must be all this demand even though the
• 52:00
  Speaker 1: demand 70 or more of all that demand comes from
• 52:05
  Speaker 1: these two companies who were funded by these three companies
• 52:08
  Speaker 1: and that's the funny thing the reason that they don't
• 52:11
  Speaker 1: want to break out their ai revenues is because it
• 52:13
  Speaker 1: will become alarmingly obvious that this was the case. It
• 52:17
  Speaker 1: turns out that the only real big customers, because it's
• 52:20
  Speaker 1: not like they're building a few data centers, they're building
• 52:24
  Speaker 1: trillion plus revenue potential, they believe they'll get.
• 52:27
  Speaker 2: Speculative.
• 52:28
  Speaker 1: Entirely speculative. They're building it because they saw the biggest
• 52:31
  Speaker 1: companies in the world buy a bunch of GPUs and
• 52:33
  Speaker 1: they said, I want in on that. They must have
• 52:35
  Speaker 1: diverse customers, right? They wouldn't just have two unprofitable fail
• 52:39
  Speaker 1: sons that they're propping up with, Christ, they've raised $ 217
• 52:42
  Speaker 1: billion just in 2026.
• 52:47
  Speaker 2: So we know that some of the biggest companies in
• 52:49
  Speaker 2: the world are using AI, generative AI, to write a
• 52:52
  Speaker 2: lot of their code. That is a great productivity gain
• 52:55
  Speaker 2: for those companies, right?
• 52:56
  Speaker 1: I mean, have you used Google or Facebook or Instagram
• 53:00
  Speaker 1: or GitHub recently? Because they are catastrophically worse. Amazon Web
• 53:04
  Speaker 1: Services went down multiple times because of their AI coding tool.
• 53:08
  Speaker 2: How is Google worse?
• 53:10
  Speaker 1: Well, I'll tell the story of a real arsehole, a
• 53:12
  Speaker 1: guy called Prabhagaur Raghavan. Previously one of the heads of
• 53:14
  Speaker 1: ads at Google, in 2019... Google called something called a
• 53:17
  Speaker 1: code yellow, which is when they said, we've got a problem.
• 53:20
  Speaker 1: And it was material weakness in query numbers, which means
• 53:24
  Speaker 1: the amount of times that people were searching on Google search.
• 53:27
  Speaker 1: A guy called Ben Gomes, internal at Google, then the
• 53:30
  Speaker 1: head of Google search says, wait a minute, to increase
• 53:32
  Speaker 1: this number of using Google more, we're going to have to,
• 53:36
  Speaker 1: I mean, what you're suggesting would mean we give worse
• 53:39
  Speaker 1: answers because if someone got the answer quickly, that would
• 53:41
  Speaker 1: reduce the amount of queries, right? And people at google
• 53:45
  Speaker 1: shashi thakur was another engineer was saying yeah can we
• 53:47
  Speaker 1: please tell sundar this because this doesn't seem good we
• 53:50
  Speaker 1: can't just increase the amount of queries that would just
• 53:53
  Speaker 1: mean that people would have to search more which make
• 53:55
  Speaker 1: the product worse but.
• 53:56
  Speaker 2: But it would make them more money yes because.
• 53:58
  Speaker 1: You'd show them more ads so if you were spending
• 54:01
  Speaker 1: more time on google because google's work.
• 54:03
  Speaker 2: But is this linked to ai doing code.
• 54:05
  Speaker 1: I'll get there so this is the problem is is
• 54:09
  Speaker 1: that This guy called Prabhakar Raghavan, who was the head
• 54:11
  Speaker 1: of ads at the time, was pushing, pushing and saying, no,
• 54:13
  Speaker 1: we need to make more queries happen.
• 54:15
  Speaker 2: Got to make it happen.
• 54:16
  Speaker 1: Nick Fox, who was there as well, I believe he's
• 54:17
  Speaker 1: actually taken over Google search, said, got to make them
• 54:19
  Speaker 1: go up.
• 54:20
  Speaker 2: This is our new reality.
• 54:22
  Speaker 1: Sometime in early 2020, Prabhakar Raghavan takes over Google search.
• 54:26
  Speaker 1: From then, and this is what I believe, can't prove it.
• 54:30
  Speaker 1: If you go and look around the various SEO sites,
• 54:32
  Speaker 1: such as Journal and the various forums, Google stripped back
• 54:35
  Speaker 1: a lot of the suppression of spammy sites so that
• 54:38
  Speaker 1: people would be on Google more. And then over the
• 54:41
  Speaker 1: course of time, Google wanted to create more queries and
• 54:44
  Speaker 1: Google search became much worse. It's why people always do
• 54:47
  Speaker 1: like plus Reddit or from Reddit or what have you.
• 54:50
  Speaker 1: It's because the actual underlying search results of Google had
• 54:52
  Speaker 1: got worse. And then generative AI came along. And Prabhagar,
• 54:55
  Speaker 1: wouldn't you know it, gets put to run part of Gemini.
• 54:58
  Speaker 1: And Google also was having trouble creating getting people back
• 55:02
  Speaker 1: on google and what did they think they'd do well
• 55:04
  Speaker 1: shit everyone's talking about this ai thing we'll just put
• 55:07
  Speaker 1: it right at the top so people have to stay
• 55:08
  Speaker 1: at google and actually they'll use it more because instead
• 55:11
  Speaker 1: of searching websites and doing that annoying thing where they
• 55:14
  Speaker 1: click away from google they'll just only use google instead
• 55:17
  Speaker 1: of generating answers by which i mean giving you search
• 55:20
  Speaker 1: results you click through now google is the answer is
• 55:23
  Speaker 1: it right god no it might tell you to eat
• 55:25
  Speaker 1: rocks might eat poisonous mushrooms maybe it'll give you a
• 55:28
  Speaker 1: little few links you could click through But the ideal
• 55:30
  Speaker 1: situation was that AI was the ultimate form of Google's evil,
• 55:34
  Speaker 1: which was.
• 55:35
  Speaker 2: But I'm saying here, I'm saying here, but that's not
• 55:37
  Speaker 2: the fact that coders could code on Google that's made
• 55:40
  Speaker 2: Google worse. That's human decisions have made it worse. Yes.
• 55:43
  Speaker 1: And then there's the instability of Google's platform, which is actually,
• 55:46
  Speaker 1: I should have probably led with that, a problem across
• 55:48
  Speaker 1: the whole tech industry.
• 55:49
  Speaker 2: Okay. So you're saying that you're saying Google is going
• 55:51
  Speaker 2: down more. Yes.
• 55:52
  Speaker 1: Google is less stable. Google Docs is a bug fest
• 55:56
  Speaker 1: right now and has been for a while. Google Sheets,
• 55:58
  Speaker 1: same deal. And the thing is, You're right. I'm being
• 56:01
  Speaker 1: a little unfair.
• 56:01
  Speaker 2: This is everyone.
• 56:02
  Speaker 1: It's the same with Microsoft. It's the same with Amazon.
• 56:05
  Speaker 1: It's the same across- How.
• 56:05
  Speaker 2: Do we quantify that outside of anecdotes? Like, is there
• 56:08
  Speaker 2: a way to- You're right.
• 56:09
  Speaker 1: I mean, GitHub downtime is the best example. Amazon Web
• 56:12
  Speaker 1: Services went down, I think, two or three times this
• 56:14
  Speaker 1: year because of AI tools. And honestly, you're right. It
• 56:19
  Speaker 1: is kind of hard to quantify outside of anecdotes. But
• 56:21
  Speaker 1: I challenge anyone listening to this, go and use a
• 56:23
  Speaker 1: website these days and tell me how well it works.
• 56:26
  Speaker 1: Tell me how buggy it is. Tell me how many problems,
• 56:28
  Speaker 1: even with my iPhones, the supposed best UX in town.
• 56:32
  Speaker 1: Even the iPhone is a flipping mess these days.
• 56:35
  Speaker 2: Okay, so the research says, the short answer is yes.
• 56:40
  Speaker 2: Tech downtime and software outages have demonstrably increased over the
• 56:44
  Speaker 2: last few years, and industry data points directly to the
• 56:46
  Speaker 2: explosion of AI-assisted coding as a primary culprit. The problem
• 56:50
  Speaker 2: is hitting the tech industry from two entirely different directions.
• 56:53
  Speaker 2: The code itself is getting buggier. And the sheer volume
• 56:56
  Speaker 2: of AI activity is literally crashing the underlying infrastructure. Interesting.
• 57:01
  Speaker 1: Yeah, that's because GitHub, people are just writing a bunch
• 57:04
  Speaker 1: of code, pushing it. And thus, there's just more code
• 57:07
  Speaker 1: on there.
• 57:08
  Speaker 2: That's interesting.
• 57:09
  Speaker 1: Yeah, it's a real mess as well because open sources
• 57:13
  Speaker 1: have this problem as well. Because it's well-meaning people. They're like,
• 57:15
  Speaker 1: I'll learn a bit of code with an LLM. I'm
• 57:17
  Speaker 1: going to go out and do some stuff. I'm going
• 57:19
  Speaker 1: to make this project better. And these people barely understand
• 57:21
  Speaker 1: what they're shipping. Or maybe they understand a bit of
• 57:24
  Speaker 1: code and they say, oh, Dunning-Kruger, this motherfucker. I'm just like,
• 57:28
  Speaker 1: I can understand some of this. And now the code's
• 57:29
  Speaker 1: all written and just push it right now. So GitHub
• 57:32
  Speaker 1: is flooded with AI code.
• 57:33
  Speaker 2: This sounds like it's making humans complacent.
• 57:37
  Speaker 3: It is.
• 57:37
  Speaker 2: Because we're going, okay, look, I'll let it write the
• 57:40
  Speaker 2: code for the last 100 lines. And it was broadly right.
• 57:43
  Speaker 2: So the next 100 lines, I won't check them as much.
• 57:46
  Speaker 2: And that's human nature is to get sort of to
• 57:49
  Speaker 2: take shortcuts, to spend less energy on an activity if
• 57:52
  Speaker 2: you can.
• 57:53
  Speaker 1: Right. But the AI is still making the mistake and
• 57:55
  Speaker 1: we're still making all the promises of AI. That's the thing.
• 57:58
  Speaker 1: This thing is meant to be this autonomous perfect. You
• 58:01
  Speaker 1: say it can't be perfect.
• 58:02
  Speaker 2: I don't know.
• 58:03
  Speaker 1: Based on what Sam Altman has been saying for the
• 58:05
  Speaker 1: last few years, Clammy Sammy has been promising the world
• 58:08
  Speaker 1: saying this will replace software engineers. Dario Amadei, Wario himself
• 58:12
  Speaker 1: has been saying, oh yeah, 50% of white collar labor
• 58:16
  Speaker 1: is going to go away in the next few years.
• 58:18
  Speaker 2: These people are promising the world.
• 58:20
  Speaker 1: Again, if they were saying it would be smaller and
• 58:22
  Speaker 1: they were like, yeah, it does have issues and we
• 58:25
  Speaker 1: must be... None of this, oh, what if it wakes
• 58:27
  Speaker 1: up and it's super powerful? Just like, yeah, it's probabilistic.
• 58:31
  Speaker 1: It's going to make mistakes. And if you don't know
• 58:33
  Speaker 1: what you're doing, you don't really know what you're looking at,
• 58:35
  Speaker 1: you're going to miss those mistakes and it's going to
• 58:37
  Speaker 1: get multiplicatively worse as you go when you don't know
• 58:40
  Speaker 1: what you're doing. So yeah, human nature is part of it.
• 58:44
  Speaker 1: But so is the marketing. So are the promises.
• 58:47
  Speaker 2: One of the smartest things a business can do is
• 58:49
  Speaker 2: build like a bigger company without actually hiring like one.
• 58:54
  Speaker 2: But the problem we all face is that most companies
• 58:56
  Speaker 2: don't have every skill in-house. So when I look at
• 58:58
  Speaker 2: the businesses seeing real success today, the consistent pattern with
• 59:02
  Speaker 2: all of them is how quickly they move. They bring
• 59:05
  Speaker 2: in specialists with skills in emerging areas to keep themselves ahead.
• 59:08
  Speaker 2: Even in our company, we spent the last year pulling
• 59:10
  Speaker 2: in talent across areas like AI-native strategy, no-code builds, and
• 59:15
  Speaker 2: product workflows. And we find this talent through our long-time partner,
• 59:18
  Speaker 2: Fiverr Pro. Their premium service only shows you vetted talent,
• 59:22
  Speaker 2: so you've always got the safeguard that anyone you pull
• 59:25
  Speaker 2: in to help you with a complex project has the
• 59:28
  Speaker 2: skills that you're after and will deliver to the same
• 59:31
  Speaker 2: high standards as your internal team. And most importantly, they'll
• 59:34
  Speaker 2: keep up with the pace. It's a simple strategy, but
• 59:36
  Speaker 2: it lets us stay agile without compromising on quality. So
• 59:39
  Speaker 2: if you need these kind of skills in your business,
• 59:41
  Speaker 2: head to pro.fiverr.com to find pioneering talent to fill your
• 59:44
  Speaker 2: business's gaps. That's pro.fiverr.com. You know the little traditional SIM
• 59:49
  Speaker 2: card that goes inside of our phones? They haven't changed
• 59:52
  Speaker 2: at all since they were invented in the 90s. You
• 59:55
  Speaker 2: have this physical piece of plastic that means you're locked
• 59:57
  Speaker 2: into one carrier, one network, and the second you cross
• 1:00:01
  Speaker 2: a border, that carrier can start charging you whatever they want.
• 1:00:04
  Speaker 2: But there are alternatives. And today's sponsor, Saley, is one
• 1:00:08
  Speaker 2: of them. It's an eSIM app that gives you a
• 1:00:11
  Speaker 2: safe and secure data connection in over 200 destinations. All
• 1:00:15
  Speaker 2: of their eSIMs have built-in cybersecurity, which is great if
• 1:00:18
  Speaker 2: you're traveling for work and looking at confidential material. I've
• 1:00:21
  Speaker 2: been using Saley whenever I travel because the connection is
• 1:00:24
  Speaker 2: always reliable and it saves me a ton of roaming fees.
• 1:00:27
  Speaker 2: It also means I don't have to deal with all
• 1:00:29
  Speaker 2: of the faff that surrounds sorting out a SIM everywhere
• 1:00:31
  Speaker 2: I go. If you want to give it a try,
• 1:00:33
  Speaker 2: download the Saley app from the App Store now and
• 1:00:36
  Speaker 2: scan the QR code on screen. And if you want
• 1:00:38
  Speaker 2: 15% off your first purchase, use my code DOAC when
• 1:00:44
  Speaker 2: you get to checkout. That's DOAC for 15% off. Keep
• 1:00:49
  Speaker 2: that to yourself. My car that drives itself, that is
• 1:00:53
  Speaker 2: AI technology. Yes. I sat here with Dara from Uber,
• 1:00:56
  Speaker 2: and he was saying that I think in a couple
• 1:00:58
  Speaker 2: of years' time, we won't need drivers for Uber because
• 1:01:04
  Speaker 2: the cars will drive themselves. They'll be fully autonomous. And
• 1:01:08
  Speaker 2: I think if I'm not mistaken, driving is one of
• 1:01:11
  Speaker 2: the biggest professions on planet earth. So when you hear people,
• 1:01:15
  Speaker 2: when you hear the CEO saying that there will be
• 1:01:17
  Speaker 2: job disruption, you say that they are not telling the truth. Yes.
• 1:01:22
  Speaker 1: Or they're guessing in a way that's very good for them.
• 1:01:25
  Speaker 1: Think about it from perspective of Microsoft, Satya Nadella. He's
• 1:01:28
  Speaker 1: not going to be like, yeah, we're done after this
• 1:01:30
  Speaker 1: is going to work, mate. Of course, he's going to
• 1:01:32
  Speaker 1: talk his book and he's going to say, yeah. This
• 1:01:34
  Speaker 1: is going to replace all workers. It's going to be amazing.
• 1:01:37
  Speaker 1: It's going to be so powerful. And then he'll change
• 1:01:39
  Speaker 1: his tune and say, actually, it's not going to replace workers.
• 1:01:41
  Speaker 1: It'll make them more powerful because the things aren't catching up.
• 1:01:44
  Speaker 1: Or from Uber, for example. Of course, he's going to say,
• 1:01:47
  Speaker 1: if this happens, then that would be good for Uber
• 1:01:49
  Speaker 1: because Uber would just become an autonomous taxi service. There's
• 1:01:53
  Speaker 1: a reason that Waymo has taken it. I find Waymo fascinating.
• 1:01:56
  Speaker 1: I think that shit's really cool. I think there are
• 1:01:57
  Speaker 1: socioeconomic problems that will come from it. I think there
• 1:02:00
  Speaker 1: are actual real problems there. that will emerge and also
• 1:02:02
  Speaker 1: what kind of problems well i mean socioeconomically they're like
• 1:02:05
  Speaker 1: you said one of the largest employment centers in the
• 1:02:08
  Speaker 1: world i mean just the economics of cabs will fall
• 1:02:10
  Speaker 1: apart but again we are nowhere nowhere nowhere near that
• 1:02:13
  Speaker 1: we're not even close waymo has had to do the
• 1:02:16
  Speaker 1: smallest rollouts and the most controlled things because the problem
• 1:02:19
  Speaker 1: with pretty much every ai system but especially driving is
• 1:02:22
  Speaker 1: not the getting 95 of the way it's those edge
• 1:02:25
  Speaker 1: cases it's raining which is a big problem for them
• 1:02:27
  Speaker 1: in san francisco It's a kid runs across the road,
• 1:02:30
  Speaker 1: but they're wearing a high-vis thing. Does it even notice
• 1:02:33
  Speaker 1: it's a child?
• 1:02:34
  Speaker 2: Again, this is a really interesting but very, very applicable
• 1:02:36
  Speaker 2: example of the right comparison to be made shouldn't be
• 1:02:40
  Speaker 2: autonomous vehicles versus perfection. It should be autonomous vehicles versus
• 1:02:44
  Speaker 2: human drivers.
• 1:02:45
  Speaker 1: I mean, I don't know if I agree because the
• 1:02:47
  Speaker 1: human driver might make mistakes, sure. But again, not an
• 1:02:51
  Speaker 1: expert in autonomous cars. Just want to be clear. But
• 1:02:53
  Speaker 1: if we're pushing autonomous cars out there willy-nilly, and we're
• 1:02:56
  Speaker 1: not doing so in extremely controlled environments, those edge cases
• 1:03:00
  Speaker 1: will multiply and be dangerous. Yeah, they might be better
• 1:03:03
  Speaker 1: at human drives in some ways, but they might also...
• 1:03:05
  Speaker 1: I was in Vegas the other day, and I was
• 1:03:07
  Speaker 1: in a hotel, and I watched a bunch of Zoox
• 1:03:09
  Speaker 1: cars just get fucking stuck.
• 1:03:11
  Speaker 2: They're autonomous cars. Yeah, yeah, yeah.
• 1:03:12
  Speaker 1: These weird boxy things, and they just blocked the exit.
• 1:03:15
  Speaker 1: They just all kind of lined up and just fell asleep.
• 1:03:18
  Speaker 1: I saw the same thing actually happen outside of a
• 1:03:20
  Speaker 1: hotel when I got out of a Waymo in San Francisco.
• 1:03:22
  Speaker 1: It just stopped at the... And then a bunch of
• 1:03:23
  Speaker 1: cars and another Waymo got stuck behind it.
• 1:03:26
  Speaker 2: I've seen some human bad drivers as well.
• 1:03:29
  Speaker 1: I agree, but it's just we have control over deploying
• 1:03:33
  Speaker 1: these bad or good drivers. We have an ability to
• 1:03:36
  Speaker 1: roll them out slowly, which is exactly what we should do.
• 1:03:39
  Speaker 1: I'm not saying autonomous cars are bad. I'm saying we
• 1:03:42
  Speaker 1: need to be so, so, so careful and treat them
• 1:03:46
  Speaker 1: as guilty until proven innocent because we can prove- And also,
• 1:03:50
  Speaker 1: they have people overlooking them. They actually have people monitoring
• 1:03:52
  Speaker 1: the routes. It is something they cannot rush out. And
• 1:03:55
  Speaker 1: it doesn't seem like they're rushing it, which is good.
• 1:03:57
  Speaker 1: And they're not promising the world.
• 1:03:58
  Speaker 2: I do agree. Listen, I'm a big fan of taxi
• 1:04:01
  Speaker 2: drivers generally, in part because I spend a lot of
• 1:04:03
  Speaker 2: time in taxis. And I think I'm not just getting
• 1:04:05
  Speaker 2: in there because I want to get from A to B.
• 1:04:07
  Speaker 2: I'm getting in there for lots of other reasons. However,
• 1:04:09
  Speaker 2: when I look at the stats around what is more dangerous,
• 1:04:13
  Speaker 2: driving myself... or having an autonomous vehicle drive me, there's
• 1:04:17
  Speaker 2: a 68% lower overall crash involvement rate when you're in
• 1:04:21
  Speaker 2: an autonomous vehicle. Autonomous vehicles experience roughly 2.1 police-reported crashes
• 1:04:27
  Speaker 2: per million miles compared to humans that are at roughly 4.68
• 1:04:30
  Speaker 2: per million miles. Right. So a 55% reduction when you
• 1:04:33
  Speaker 2: get in an autonomous vehicle. And autonomous vehicles show an
• 1:04:36
  Speaker 2: 80% to 81% reduction in crashes resulting in injuries versus
• 1:04:41
  Speaker 2: human drivers. Uh-huh. So you're 85% less likely to be
• 1:04:45
  Speaker 2: involved in a single vehicle crash, like hitting a wall
• 1:04:48
  Speaker 2: or a tree, if you're an autonomous vehicle versus being
• 1:04:51
  Speaker 2: driven by humans.
• 1:04:52
  Speaker 1: I agree, but.
• 1:04:54
  Speaker 2: So it's safer.
• 1:04:56
  Speaker 1: Also, that data is, what's the sample size of human drivers?
• 1:04:59
  Speaker 1: I mean, we've got many, many, many, many, many, many
• 1:05:02
  Speaker 1: more years of drivers and many, many, many more years
• 1:05:04
  Speaker 1: of accidents. And also, man, does that not have anything
• 1:05:07
  Speaker 1: to do with generative AI? If we were just talking
• 1:05:10
  Speaker 1: about that, be having a different conversation i.
• 1:05:12
  Speaker 2: Guess the question here was really around job disruption like
• 1:05:14
  Speaker 2: you know we look across industries and we go driving's
• 1:05:17
  Speaker 2: a massive profession is there going to be job disruption
• 1:05:19
  Speaker 2: because cars can now drive themselves if we think about
• 1:05:21
  Speaker 2: white-collar you know jobs you know lawyers and accountants people
• 1:05:26
  Speaker 2: sit here and they tell me that lawyers and accountants
• 1:05:29
  Speaker 2: the profession right i should say some of the skills
• 1:05:31
  Speaker 2: within the profession will be relegated to ais to do.
• 1:05:35
  Speaker 1: Here's the thing lawyers for example great example Always hearing
• 1:05:39
  Speaker 1: fucking legal partners talking about AI. Never the associates. The
• 1:05:43
  Speaker 1: associates are the ones that go out and find the precedent.
• 1:05:46
  Speaker 1: They're the ones that go and do the grunt work.
• 1:05:48
  Speaker 1: They're the ones who are pulling motions half the time.
• 1:05:49
  Speaker 1: The partner is the one that might be the litigant.
• 1:05:52
  Speaker 1: It may be the client facing. But the ones that
• 1:05:54
  Speaker 1: are actually doing the day-to-day work, I'm not hearing from them.
• 1:05:56
  Speaker 1: I'm not hearing associates being like, this is fucking awesome.
• 1:05:59
  Speaker 1: I'm hearing a bunch of well-paid people that have sat
• 1:06:03
  Speaker 1: on ChatGPT and gone, yeah, I'm the greatest lawyer ever.
• 1:06:06
  Speaker 1: They're not the ones that I want to hear from,
• 1:06:08
  Speaker 1: the actual workers. White-collar labor disruption is not happening. OpenAI
• 1:06:12
  Speaker 1: had a study that came out, I think like a
• 1:06:14
  Speaker 1: week ago, that said there was no connection between spending
• 1:06:17
  Speaker 1: on AI tokens and revenue per employee.
• 1:06:20
  Speaker 2: What does that mean? Could you explain that to me?
• 1:06:22
  Speaker 1: As in, the more tokens you spend has no correlation
• 1:06:27
  Speaker 1: at all with the amount of money you make. It's
• 1:06:29
  Speaker 1: the second report they've put out. The other one was
• 1:06:31
  Speaker 1: like hallucinations are mathematically guaranteed. Kind of almost, it's the
• 1:06:35
  Speaker 1: one thing I respect about that company. Occasionally they just
• 1:06:37
  Speaker 1: put out a study. It's like, yeah, shit's kind of sucks.
• 1:06:40
  Speaker 1: But the people that are having their lives disrupted work-wise
• 1:06:44
  Speaker 1: are art directors. It's people, art directors, transcribers, translators who
• 1:06:50
  Speaker 1: have bosses that don't care about the output. It's what
• 1:06:52
  Speaker 1: they consider cheap work. And the problem is, is those
• 1:06:56
  Speaker 1: people would have automated your work away anyway. They would
• 1:06:58
  Speaker 1: have sold it. They would have taken the cheapest offer.
• 1:07:00
  Speaker 1: They would have sold it to the global self. They
• 1:07:01
  Speaker 1: would have taken the shittiest option they could. That is
• 1:07:04
  Speaker 1: something that AI is doing. And again, those people are
• 1:07:06
  Speaker 1: not paying the actual cost of AI. They're using a subscription.
• 1:07:09
  Speaker 1: The actual white collar labor force might have some things
• 1:07:14
  Speaker 1: that are slightly changing, but there is no evidence of
• 1:07:18
  Speaker 1: productivity gains. In fact, if there were, they would be
• 1:07:22
  Speaker 1: screaming it from the rooftops. There was an Oxford Economics
• 1:07:24
  Speaker 1: study last year where it's like, oh, young people are
• 1:07:27
  Speaker 1: finding less jobs because of AI. We actually read the study,
• 1:07:30
  Speaker 1: which multiple journalists did not. It was a single line
• 1:07:33
  Speaker 1: that said, yeah, we saw some correlation. Didn't give a number.
• 1:07:37
  Speaker 1: Didn't actually say what the correlation was. We are so
• 1:07:40
  Speaker 1: conditioned to believe that the rich and powerful know what
• 1:07:43
  Speaker 1: they're doing that we internalize these narratives about like, well,
• 1:07:47
  Speaker 1: previous booms lost a lot of money. Well, technology takes
• 1:07:50
  Speaker 1: time to do stuff. And they are intentionally playing on
• 1:07:54
  Speaker 1: those mythologies. They are playing on these knowing that journalists, analysts,
• 1:07:58
  Speaker 1: investors will believe them. And this is partly because our
• 1:08:03
  Speaker 1: realities are defined by stock prices. Because the stock prices
• 1:08:06
  Speaker 1: of these companies went up, we're like, oh, look, it
• 1:08:08
  Speaker 1: must be working, right?
• 1:08:10
  Speaker 2: Both of those things you said were true, though, right?
• 1:08:11
  Speaker 2: Like that previous technologies didn't make money at the start.
• 1:08:14
  Speaker 2: And the other one you said was they'll get better.
• 1:08:17
  Speaker 2: But that's the thing.
• 1:08:18
  Speaker 1: Okay, because another thing got better, this will get better?
• 1:08:21
  Speaker 2: No, but there's got to be something that they're saying
• 1:08:23
  Speaker 2: that is fundamentally not true, because those are two true
• 1:08:26
  Speaker 2: statements that, okay, technology often.
• 1:08:27
  Speaker 1: Starts- Okay, I know. I get what you mean. What
• 1:08:29
  Speaker 1: they are fundamentally misleading people about is how possible it
• 1:08:33
  Speaker 1: is and how many actual signs they have, because they
• 1:08:35
  Speaker 1: don't have the signs. If they had the signs, as
• 1:08:37
  Speaker 1: in the signs of this getting cheaper, as in the
• 1:08:39
  Speaker 1: signs of this being able to autonomously do work without
• 1:08:42
  Speaker 1: the Rube Goldberg machine, and even then, in a reliable
• 1:08:45
  Speaker 1: way that was making the customer more money, being productive
• 1:08:49
  Speaker 1: in a way you can say with your whole chest
• 1:08:51
  Speaker 1: without a series of asterisks. And that's how it is
• 1:08:54
  Speaker 1: across the board. The people that are most excited about this,
• 1:08:59
  Speaker 1: psychopaths on Twitter, in many cases are.
• 1:09:01
  Speaker 2: People that I believe they really are.
• 1:09:04
  Speaker 1: I'm sorry. There are some people on Twitter because the
• 1:09:06
  Speaker 1: other thing about this is this is really unique to
• 1:09:09
  Speaker 1: the AI industry.
• 1:09:09
  Speaker 2: I've never seen it.
• 1:09:10
  Speaker 1: Any other industry outside of maybe like sports teams, the
• 1:09:14
  Speaker 1: attachment that some people online have to these companies. If
• 1:09:17
  Speaker 1: you dare, dare to criticize Anthropic. It's almost this religious attachment.
• 1:09:22
  Speaker 1: Good example was this week, Bloomberg reported that OpenAI was
• 1:09:26
  Speaker 1: on track to hit $ 40 billion in annualized revenue. Month
• 1:09:29
  Speaker 1: times 12, four weeks times 13.
• 1:09:31
  Speaker 2: We don't know. They don't define it.
• 1:09:33
  Speaker 1: I saw multiple people going, actually, it's $ 60 billion. It's
• 1:09:36
  Speaker 1: actually $ 60 billion. I heard from someone. It is like
• 1:09:40
  Speaker 1: a cult. And it's a cult of software driven around
• 1:09:43
  Speaker 1: growth and this idea that by backing the right horse,
• 1:09:47
  Speaker 1: you will have some grand thing. And OpenAI in particular,
• 1:09:52
  Speaker 1: in particular, Mr. Altman, they have been fermenting this. They're
• 1:09:56
  Speaker 1: Tebow as well, the TIBO, one of the guys at OpenAI,
• 1:10:00
  Speaker 1: they ferment this thing online. They build this kind of
• 1:10:02
  Speaker 1: parasocial relationship with both the large language model themselves and
• 1:10:07
  Speaker 1: the companies. And one's allegiance to the companies is so important.
• 1:10:11
  Speaker 1: It's truly vile. If only these people gave a fuck about,
• 1:10:15
  Speaker 1: I don't know, Medicare for all or poverty or things
• 1:10:18
  Speaker 1: like actual problems in the world versus are we buying
• 1:10:21
  Speaker 1: enough GPUs?
• 1:10:22
  Speaker 2: Do you know what's interesting is some of what your narrative,
• 1:10:26
  Speaker 2: one would argue, actually helps them.
• 1:10:29
  Speaker 1: How?
• 1:10:30
  Speaker 2: Because, you know, the AI doomers that have come here
• 1:10:32
  Speaker 2: and told, you know, some of the original founding fathers
• 1:10:33
  Speaker 2: of AI like Geoffrey Hinton have told me that what
• 1:10:37
  Speaker 2: they're building is highly, highly dangerous and that it will
• 1:10:39
  Speaker 2: be fundamentally disruptive to society. And, It's interesting because some
• 1:10:44
  Speaker 2: of the CEOs who you've mentioned, their historical narrative was also,
• 1:10:48
  Speaker 2: by the way, this is really fucking dangerous. Yeah. And
• 1:10:50
  Speaker 2: there is a significant chance we could fuck up the planet.
• 1:10:53
  Speaker 2: And what we've seen is this slow pivot away from it.
• 1:10:56
  Speaker 2: Which is so strange. Because now they're getting booed and
• 1:10:58
  Speaker 2: they're being attacked.
• 1:10:58
  Speaker 1: Yeah.
• 1:10:59
  Speaker 2: They've been this slow pivot away from it. And the
• 1:11:01
  Speaker 2: pivot almost sounds a little bit like your narrative. It
• 1:11:06
  Speaker 2: now sounds like, actually, no, it's not going to change anything.
• 1:11:08
  Speaker 2: And you're all going to be fine. And it's now,
• 1:11:09
  Speaker 2: it's just not dangerous at all.
• 1:11:11
  Speaker 1: But that's the funny thing.
• 1:11:13
  Speaker 2: And that's why I'm saying like, you're not, they, I
• 1:11:15
  Speaker 2: actually think there might be a couple of PR people
• 1:11:17
  Speaker 2: at these big AI companies thinking, thank God for Ed.
• 1:11:20
  Speaker 1: Oh, I don't know about that.
• 1:11:22
  Speaker 2: Some of it, because you're like, you're saying, actually, don't worry.
• 1:11:25
  Speaker 2: Everything's going to be fine. It's not going to take
• 1:11:26
  Speaker 2: your job. It's not going to disrupt the economy. It's
• 1:11:28
  Speaker 2: just a fad. There's no technology. And I think they
• 1:11:30
  Speaker 2: don't think that.
• 1:11:31
  Speaker 1: Here's the thing. I think Altman and Amadei are some
• 1:11:33
  Speaker 1: of the most deeply corrupt and cynical people in the world.
• 1:11:36
  Speaker 1: I don't think, of course they were going to say
• 1:11:37
  Speaker 1: from the, it was early 2023. Altman said, we're a
• 1:11:41
  Speaker 1: little bit scared about what we're creating. Oh, shut up.
• 1:11:44
  Speaker 1: I'm just, I hear that and I feel so frustrated
• 1:11:46
  Speaker 1: because I've met so many of these rich fucking liars,
• 1:11:50
  Speaker 1: these people. And you know why he wants to say that?
• 1:11:52
  Speaker 1: So you'll invest in his company and buy the software.
• 1:11:54
  Speaker 1: So you'll be scared that if you don't use AI today,
• 1:11:57
  Speaker 1: you'll be left behind in the future, which is their
• 1:11:59
  Speaker 1: continual narrative. That if you don't get on the train today,
• 1:12:02
  Speaker 1: then you'll be left behind. By the way, every single
• 1:12:06
  Speaker 1: scam and con starts with rushing you. Every single trick
• 1:12:10
  Speaker 1: in history begins with saying, you must do this now.
• 1:12:12
  Speaker 1: And best piece of advice I ever got was if
• 1:12:15
  Speaker 1: anyone tries to rush you and it's not literally a
• 1:12:17
  Speaker 1: mortal thing, like you are bleeding or on fire or
• 1:12:19
  Speaker 1: the house is on fire, slow down. And yet all
• 1:12:22
  Speaker 1: of these companies say, so scary. And now they're talking
• 1:12:25
  Speaker 1: about slowdowns, but you ever noticed that Amaday and Altman,
• 1:12:28
  Speaker 1: they say, well, maybe we should slow down progress. And
• 1:12:30
  Speaker 1: then they don't. Right now, Altman's saying, oh, we slow
• 1:12:33
  Speaker 1: down progress because we're so delayed. No, they're out of compute.
• 1:12:36
  Speaker 1: Now they're doing it. I can guarantee you, by the way,
• 1:12:37
  Speaker 1: they're PR people.
• 1:12:38
  Speaker 2: Do not like me.
• 1:12:39
  Speaker 1: I don't think OpenAI's PR people are super fond of me.
• 1:12:43
  Speaker 2: But I bet there's elements of what you're saying because
• 1:12:45
  Speaker 2: you're calming people. You are theoretically calming down the general public.
• 1:12:48
  Speaker 1: And you know what? I hope I am because the
• 1:12:51
  Speaker 1: fear-based tactics is horrible. These companies don't want that. These
• 1:12:54
  Speaker 1: companies want people scared. I'm 100% sure. I just fundamentally disagree.
• 1:13:00
  Speaker 2: So the timelines, and I sit here and what I
• 1:13:02
  Speaker 2: do is I log their quotes over time. And I
• 1:13:05
  Speaker 2: read them out from 2015. Yeah. To 2026. And the
• 1:13:10
  Speaker 2: change you see is them going from there could be extinction.
• 1:13:14
  Speaker 2: That's the narrative, the early narrative. Elon said it himself.
• 1:13:16
  Speaker 2: He says it's the single most dangerous thing. Elon says
• 1:13:18
  Speaker 2: a lot of things. And then you track it over
• 1:13:20
  Speaker 2: time and it evolves to this age of abundance. We're
• 1:13:24
  Speaker 2: all going to have unlimited stuff. And then the new
• 1:13:27
  Speaker 2: slogan at ChatGPT is intelligence for everyone. It's suddenly, and
• 1:13:32
  Speaker 2: whenever Dario comes out and says, by the way, it's
• 1:13:34
  Speaker 2: really rockin' dangerous, they attack Dario. Yeah, that's right.
• 1:13:37
  Speaker 1: They hate him. They're like, Dario, shut the fuck up. Honestly,
• 1:13:42
  Speaker 1: I've been saying, Dario, shut the fuck up for years.
• 1:13:44
  Speaker 1: But the thing is, I get your point where it's like,
• 1:13:47
  Speaker 1: I don't think they've changed to calm the public down
• 1:13:50
  Speaker 1: so much as they're desperate to not get regulated, which
• 1:13:53
  Speaker 1: is laughable. We don't regulate tech. We don't regulate shit.
• 1:13:56
  Speaker 1: America doesn't regulate shit. We are still trapped in the
• 1:14:01
  Speaker 1: hands of Milton Friedman, Margaret Thatcher, and fucking Ronald Reagan.
• 1:14:05
  Speaker 1: We're still stuck in the neoliberalistic hellscape. which is growth
• 1:14:09
  Speaker 1: at all costs, free market capitalism. So no, no one's regulating.
• 1:14:13
  Speaker 1: The regulation of these companies should have been, I don't know,
• 1:14:16
  Speaker 1: breaking up, put these bastards aside, break up these fuckers
• 1:14:20
  Speaker 1: for sure.
• 1:14:20
  Speaker 2: We shouldn't have companies this big. It makes things worse.
• 1:14:22
  Speaker 2: But these technologies are dangerous.
• 1:14:24
  Speaker 1: I mean, they're dangerous, but not in the ways they've
• 1:14:26
  Speaker 1: been warning about.
• 1:14:27
  Speaker 2: Less if we think about cyber hacking. Right.
• 1:14:30
  Speaker 1: And just be clear, those cyber hacking things that happened
• 1:14:32
  Speaker 1: were not a result of, they were like, break out
• 1:14:35
  Speaker 1: of the sandbox. And then they set the sandbox up
• 1:14:38
  Speaker 1: wrong they set up the server they were on wrong.
• 1:14:40
  Speaker 2: But i mean you know advanced ai models could very
• 1:14:43
  Speaker 2: easily because they can go out onto the open internet
• 1:14:45
  Speaker 2: as agents they could very easily go and look at
• 1:14:48
  Speaker 2: code bases of different websites find vulnerabilities and exploit those
• 1:14:51
  Speaker 2: vulnerabilities yeah you know at scale and arguably um at
• 1:14:55
  Speaker 2: a higher intelligence and faster and wider than humans a
• 1:15:00
  Speaker 2: human hacker could theoretically so that's dangerous.
• 1:15:02
  Speaker 1: Well here's the funny thing We don't know how much
• 1:15:05
  Speaker 1: compute was spent to do the hugging face attack, the
• 1:15:08
  Speaker 1: open AI one. We also do know that they improperly
• 1:15:10
  Speaker 1: set up the server to keep it in. They thought
• 1:15:13
  Speaker 1: they'd turn the internet off and they didn't. That's human error.
• 1:15:16
  Speaker 1: And that's human error in a sense that, yeah, they
• 1:15:18
  Speaker 1: threw about an indeterminately large amount of compute. This is dangerous.
• 1:15:22
  Speaker 1: But people keep saying we can't let the Chinese get.
• 1:15:25
  Speaker 2: A hold of these models. We couldn't possibly.
• 1:15:27
  Speaker 1: Because what if these models fall into the wrong hands?
• 1:15:30
  Speaker 1: They're already in the wrong hands. Mark Zuckerberg, Sam Altman,
• 1:15:33
  Speaker 1: Dario Amadei. The wrong hands are the hands of those
• 1:15:37
  Speaker 1: who are running these companies. We should not be training.
• 1:15:40
  Speaker 2: These models to do these things.
• 1:15:42
  Speaker 1: I don't know why the fuck we're doing it. Other
• 1:15:44
  Speaker 1: than they've run out of other things they can train on.
• 1:15:46
  Speaker 1: There's a ton. And the fact that they can do it,
• 1:15:48
  Speaker 1: it's kind of interesting.
• 1:15:49
  Speaker 2: But would you agree that it's an intelligence and i'll
• 1:15:54
  Speaker 2: call it that you know you might disagree with that
• 1:15:55
  Speaker 2: terminology but an intelligence that can go out onto the
• 1:15:58
  Speaker 2: internet and click around and take actions is inherently there's
• 1:16:03
  Speaker 2: risks associated with that well.
• 1:16:06
  Speaker 1: The second part i agree with the risks we've had
• 1:16:08
  Speaker 1: people running automated scripts hacking scripts for a while we've
• 1:16:11
  Speaker 1: had hackers doing that for years and years and years
• 1:16:14
  Speaker 1: this is brute forcing it with a bunch of compute
• 1:16:15
  Speaker 1: and yet It is dangerous. These companies are doing something dangerous.
• 1:16:20
  Speaker 1: That is not what Jeffrey Hinton et al have been
• 1:16:23
  Speaker 1: warning about. They've been saying, oh, these things could destroy society.
• 1:16:26
  Speaker 1: They could manipulate people. When you actually look at the
• 1:16:28
  Speaker 1: underlying things, not so much. Jeffrey Hinton as well, talking
• 1:16:31
  Speaker 1: his book, still got his Google stock, I think. And
• 1:16:33
  Speaker 1: weirdly enough, he left Google because he was worried about
• 1:16:35
  Speaker 1: the AI there, but then immediately made a comment being like, yeah, actually, though,
• 1:16:39
  Speaker 1: Google's very responsible. Strange thing, though. But let's get back
• 1:16:42
  Speaker 1: to the cybersecurity side. I agree. This is dangerous. These
• 1:16:45
  Speaker 1: people should not have access to so much compute. They
• 1:16:47
  Speaker 1: clearly don't know what to do with it. There's a
• 1:16:49
  Speaker 1: really easy way of dealing with this. It's not letting
• 1:16:52
  Speaker 1: them use so much compute. It's regulating that part out
• 1:16:54
  Speaker 1: of existence.
• 1:16:55
  Speaker 2: What if the Chinese do it? The Chinese were able
• 1:16:58
  Speaker 2: to distill the models.
• 1:16:59
  Speaker 1: And also, I don't know, regulate it and stop. I
• 1:17:04
  Speaker 1: feel like with this particular thing as well, we got
• 1:17:08
  Speaker 1: to this point and let the genie out of the bottle.
• 1:17:10
  Speaker 1: to use an annoying Sam Altman term. We let this
• 1:17:13
  Speaker 1: happen because we let these companies be unregulated and use
• 1:17:16
  Speaker 1: as much computers as we want. We had these fucking
• 1:17:18
  Speaker 1: enablers allowing them to burn as much computers.
• 1:17:20
  Speaker 2: As they want.
• 1:17:21
  Speaker 1: And also, for all of these dire warnings about AI dangers,
• 1:17:26
  Speaker 1: no one seems to have fucking done anything.
• 1:17:28
  Speaker 2: Okay, we're going to play a game, Ed.
• 1:17:29
  Speaker 1: Let's play it.
• 1:17:30
  Speaker 2: On these cards here, I have the things that you
• 1:17:33
  Speaker 2: consider to be myths about the AI industry. The challenge...
• 1:17:38
  Speaker 2: is I want you to give me one sentence on
• 1:17:41
  Speaker 2: each myth. Oh, great. So just your first reaction. You're
• 1:17:44
  Speaker 2: going to pick it up, you're going to read it,
• 1:17:45
  Speaker 2: and then you're going to give me one sentence on
• 1:17:47
  Speaker 2: your opinion of that belief. Okay, let's go. Let's do this.
• 1:17:56
  Speaker 2: What does it say?
• 1:17:57
  Speaker 1: It says the AI industry is creating enormous economic growth.
• 1:18:02
  Speaker 2: No, it's not.
• 1:18:03
  Speaker 1: It's nowhere in the data. okay it's just may i
• 1:18:07
  Speaker 1: do a second sentence go ahead pretty much all of
• 1:18:10
  Speaker 1: the economics is either nvidia feeding money to its companies
• 1:18:12
  Speaker 1: like core wave or these three companies feeding money to
• 1:18:15
  Speaker 1: these ones to spend it with the them.
• 1:18:17
  Speaker 2: Okay and what evidence do you have that there's it's
• 1:18:21
  Speaker 2: not causing economic growth.
• 1:18:23
  Speaker 1: Just to be clear other than the spend on semiconductors
• 1:18:26
  Speaker 1: so the speculative investment in gpus and data center infrastructure
• 1:18:30
  Speaker 1: that's happening but as far as like Spend on AI goes,
• 1:18:33
  Speaker 1: barely cracking $ 100 billion, and most of that is just
• 1:18:36
  Speaker 1: these two running their services and paying these three companies, Oracle, Core,
• 1:18:40
  Speaker 1: even others.
• 1:18:41
  Speaker 2: But $ 100 billion is a lot of money for a
• 1:18:44
  Speaker 2: relatively new technology.
• 1:18:45
  Speaker 1: Not when you've spent $ 300 billion in equity funding and
• 1:18:48
  Speaker 1: if we're going with just these three, I think $ 600
• 1:18:52
  Speaker 1: billion in capital expenditures.
• 1:18:53
  Speaker 2: Yeah, I get that. That means it's not profitable. But
• 1:18:55
  Speaker 2: the $ 100 billion is an expression of consumer demand.
• 1:18:58
  Speaker 1: When the compute is mostly driven by subscriptions that are subsidized, no,
• 1:19:02
  Speaker 1: it's not. When you're giving someone $ 20 or $ 40 for
• 1:19:04
  Speaker 1: a dollar.
• 1:19:05
  Speaker 1: They are going to use it more. If this was
• 1:19:07
  Speaker 1: all on a per million token basis, we'd be having
• 1:19:09
  Speaker 1: a different conversation.
• 1:19:10
  Speaker 2: Okay, fine.
• 1:19:11
  Speaker 3: Cool.
• 1:19:11
  Speaker 2: Next one.
• 1:19:14
  Speaker 1: The United States need to spend trillions to beat China
• 1:19:16
  Speaker 1: in the AI race.
• 1:19:19
  Speaker 2: Let's see.
• 1:19:21
  Speaker 1: What AI race? that's actually my point it's what AI
• 1:19:26
  Speaker 1: race is there is it to make big scary LLMs
• 1:19:29
  Speaker 1: they did that already without the Nvidia GPUs by the
• 1:19:32
  Speaker 1: way they've got Blackwell GPUs Kakashi and Justario two amazing
• 1:19:35
  Speaker 1: analysts I love they've been on this for years it's
• 1:19:38
  Speaker 1: like China's already had Nvidia GPUs that they're not meant
• 1:19:41
  Speaker 1: to have for years but also to do what they
• 1:19:43
  Speaker 1: already got the LLMs what's the race to do to
• 1:19:46
  Speaker 1: make us spend more money than them for us to
• 1:19:48
  Speaker 1: constantly piss our pants worrying about China because they won
• 1:19:51
  Speaker 1: if that's the case Myth number three. AI will replace
• 1:19:56
  Speaker 1: all human jobs. That just isn't happening. There's no economic
• 1:20:00
  Speaker 1: data to support it.
• 1:20:02
  Speaker 2: Will it replace some jobs? I mean, it's.
• 1:20:05
  Speaker 1: Replaced some contract labor that would otherwise be replaced with
• 1:20:08
  Speaker 1: cheap labor out in the global south. It's a digital
• 1:20:10
  Speaker 1: globalization in that sense. But all jobs, most jobs, a
• 1:20:15
  Speaker 1: lot of jobs, no.
• 1:20:16
  Speaker 2: What about robotics?
• 1:20:17
  Speaker 1: Robotics is not what we're talking about. Robotics is a
• 1:20:20
  Speaker 1: very different thing. And even then.
• 1:20:21
  Speaker 2: Robotics will be powered by AI.
• 1:20:23
  Speaker 1: I mean, yes, but there are tons of different kinds
• 1:20:25
  Speaker 1: of AI. We're talking explicitly about generative AI, and that's
• 1:20:27
  Speaker 1: what I... This is where my myth busts this piece.
• 1:20:29
  Speaker 1: It was definitely about generative AI.
• 1:20:31
  Speaker 2: Okay, but what about robotics? The thing is... The Optimus
• 1:20:33
  Speaker 2: robot that Elon's working on at Tesla.
• 1:20:36
  Speaker 1: The one where even in the demo of the hand,
• 1:20:40
  Speaker 1: they had to have a guy controlling it.
• 1:20:41
  Speaker 2: It wasn't doing it autonomously.
• 1:20:42
  Speaker 1: Here's the thing. If they can beat all these challenges, yeah,
• 1:20:46
  Speaker 1: robotics would be really cool. I don't know how long that's...
• 1:20:49
  Speaker 1: That's when I'd actually be willing to believe... In a
• 1:20:52
  Speaker 1: couple decades.
• 1:20:53
  Speaker 2: Have you seen them Chinese robots? I know you've seen them.
• 1:20:56
  Speaker 1: What's it called? The one that can dance and that,
• 1:20:59
  Speaker 1: but they can't really do human things.
• 1:21:01
  Speaker 2: Well, it's just, it is pretty mind blowing.
• 1:21:04
  Speaker 1: Robotics are fucking cool. I like, I'm not going to
• 1:21:06
  Speaker 1: pretend I don't think robots are cool. I wish they
• 1:21:08
  Speaker 1: were building robots and actually doing cool shit. I wish
• 1:21:11
  Speaker 1: the tech industry still made fun stuff and interesting stuff.
• 1:21:14
  Speaker 1: Instead we get these fucking large language models.
• 1:21:16
  Speaker 2: But AI plus robotics is, You know, I was in
• 1:21:20
  Speaker 2: San Francisco and I went to this massive incubator there.
• 1:21:23
  Speaker 2: And when I'd gone there three years earlier, it was
• 1:21:25
  Speaker 2: all software startups. Right. And when I went back three
• 1:21:27
  Speaker 2: years later, it was all these robot startups. And I
• 1:21:30
  Speaker 2: remember saying to the founder of the incubator, I was like,
• 1:21:32
  Speaker 2: why is everything robots now? There was this one robot
• 1:21:35
  Speaker 2: where it was just the arm and it had a
• 1:21:37
  Speaker 2: frying pan on it.
• 1:21:37
  Speaker 1: Yeah.
• 1:21:38
  Speaker 2: And its whole thing is it cooks for you.
• 1:21:39
  Speaker 1: Yeah.
• 1:21:40
  Speaker 2: So it was showing me it cooking, whatever. And he goes, well,
• 1:21:43
  Speaker 2: you know, the arm, he goes, the hardware part, the
• 1:21:45
  Speaker 2: physical parts are That's always been fairly cheap. Yeah. He goes,
• 1:21:49
  Speaker 2: the expensive part was the intelligence. Yeah. And now that's
• 1:21:51
  Speaker 2: come down to pennies. So what you're seeing is this
• 1:21:53
  Speaker 2: explosion in the robotics industry because robotics is a function
• 1:21:56
  Speaker 2: of intelligence plus hardware. We've always had that.
• 1:21:58
  Speaker 1: And a ton of data, though, as well. Yeah. And
• 1:22:00
  Speaker 1: the data is very expensive.
• 1:22:02
  Speaker 3: Yeah.
• 1:22:03
  Speaker 1: The thing is, cybercabs rolled out real slow. It's going
• 1:22:06
  Speaker 1: to take a long time.
• 1:22:07
  Speaker 2: It could be a threat.
• 1:22:08
  Speaker 1: If they do a robot that could replace a human job,
• 1:22:11
  Speaker 1: sure it could. But human jobs are multifaceted. Human jobs
• 1:22:15
  Speaker 1: change with environments and also... A lot of human jobs
• 1:22:18
  Speaker 1: that you might think of like, I don't know, dishwashing robot,
• 1:22:20
  Speaker 1: for example.
• 1:22:21
  Speaker 3: Yeah.
• 1:22:22
  Speaker 1: Some guy at a restaurant isn't paying 10, 20 grand
• 1:22:25
  Speaker 1: for a robot to replace the job that they're already
• 1:22:27
  Speaker 1: not paying enough for. The point is, yeah, it could
• 1:22:30
  Speaker 1: if you can replace the jobs. That is not what
• 1:22:33
  Speaker 1: we're talking about with this.
• 1:22:34
  Speaker 2: Yeah, I just, I ask these questions not because I'm
• 1:22:37
  Speaker 2: trying to be like democratic. I actually am trying to
• 1:22:39
  Speaker 2: form my own opinion on these things. And I do think,
• 1:22:44
  Speaker 2: you know, as it's written there, it says AI will
• 1:22:46
  Speaker 2: replace all human jobs. Obviously not. Obviously that's bullshit. But
• 1:22:50
  Speaker 2: I'm trying to figure out if the truth is somewhere
• 1:22:52
  Speaker 2: in the middle, that there's a certain type of job,
• 1:22:55
  Speaker 2: which actually humans probably shouldn't have ever been doing, really.
• 1:22:59
  Speaker 2: If you think back through history, there was someone's job
• 1:23:00
  Speaker 2: just to sit in an elevator and press the buttons.
• 1:23:02
  Speaker 2: That's an example of a job that humans probably shouldn't
• 1:23:04
  Speaker 2: have been doing. And as technology gets more advanced, it
• 1:23:07
  Speaker 2: takes on a lot of that sort of automated monotonous stuff.
• 1:23:11
  Speaker 1: Right the thing is with this particular thing that i
• 1:23:14
  Speaker 1: know that this is from it's a specific blog i
• 1:23:16
  Speaker 1: wrote i was explicitly talking about generative ai though i
• 1:23:19
  Speaker 1: was explicitly talking about people when they say this they
• 1:23:22
  Speaker 1: are referring to that so you're not talking about agentic
• 1:23:24
  Speaker 1: ai which agentic ai is llms agentic ai is just
• 1:23:27
  Speaker 1: a fancy way of saying an llm talking to another
• 1:23:29
  Speaker 1: llm with a harness on top that is still a
• 1:23:31
  Speaker 1: lens agentic ai is is one of the bigger lies
• 1:23:35
  Speaker 1: they tell. It's like when you hear agent, you're meant
• 1:23:36
  Speaker 1: to think autonomous AI can do what you want.
• 1:23:39
  Speaker 2: It's still LLMs. It's still LLMs talking to other LLMs.
• 1:23:42
  Speaker 2: Taking screenshots and putting them in LLMs and stuff.
• 1:23:44
  Speaker 1: Oh God, yeah.
• 1:23:45
  Speaker 2: Okay, but you know, I could make the case that
• 1:23:49
  Speaker 2: I'm just thinking about my personal usage. I definitely use
• 1:23:53
  Speaker 2: agents to do things that I would have previously asked
• 1:23:55
  Speaker 2: people to do. It's not to say that I still
• 1:23:57
  Speaker 2: don't hire because we're hiring like crazy.
• 1:23:58
  Speaker 1: Yeah.
• 1:23:59
  Speaker 2: And I still, in that particular function, I'm thinking about
• 1:24:01
  Speaker 2: like the chief of staff role So my chief of
• 1:24:04
  Speaker 2: staff would have triaged all of my inboxes previously and
• 1:24:06
  Speaker 2: put them somewhere and told me about them or maybe
• 1:24:09
  Speaker 2: once upon a time shown me a piece of paper
• 1:24:10
  Speaker 2: back in the day, I guess. Now my chief of
• 1:24:12
  Speaker 2: staff is no longer doing that job. You still have
• 1:24:14
  Speaker 2: a chief of staff though. This is what I'm saying.
• 1:24:16
  Speaker 2: They're doing other things. Right.
• 1:24:18
  Speaker 1: But the thing is, again, what you were describing is
• 1:24:22
  Speaker 1: fairly basic automation. I don't know what the tasks are.
• 1:24:24
  Speaker 2: Yeah, it's fairly basic even.
• 1:24:26
  Speaker 1: Didn't spend a trillion dollars on triaging email. The promise,
• 1:24:29
  Speaker 1: if they'd spent $ 10 billion and this was much smaller,
• 1:24:32
  Speaker 1: and you're like, cool, software, yay. A lot of the
• 1:24:35
  Speaker 1: things that people are impressed with, like script stuff as well,
• 1:24:37
  Speaker 1: it's just LLM's doing Python. You should be impressed by
• 1:24:39
  Speaker 1: Python code. Python's incredible. You can scrape websites.
• 1:24:42
  Speaker 2: You can download shit. It's awesome.
• 1:24:44
  Speaker 1: But the point I'm making is none of this would
• 1:24:47
  Speaker 1: be anywhere near as much of a problem if they
• 1:24:49
  Speaker 1: didn't ask for all of the attention, all of the money,
• 1:24:51
  Speaker 1: and promise the world. It's their promises that are the
• 1:24:54
  Speaker 1: problem and the journalists who went along with it and
• 1:24:56
  Speaker 1: the analysts and the Twitter people who went along with
• 1:24:58
  Speaker 1: this saying that this would change everything and replace everything
• 1:25:01
  Speaker 1: and leaving the realm of reality.
• 1:25:03
  Speaker 2: Is there any technological innovation through history that was really,
• 1:25:06
  Speaker 2: really game-changing where that didn't happen?
• 1:25:11
  Speaker 1: I mean, the internet.
• 1:25:14
  Speaker 2: I mean, people over-promised that.
• 1:25:15
  Speaker 1: I mean, they over-promised on the businesses, but I've read
• 1:25:18
  Speaker 1: through a great many pieces about the early internet. A
• 1:25:21
  Speaker 1: lot of people were... excited but hesitant. They were worried
• 1:25:25
  Speaker 1: that there was not enough demand, but they were still like,
• 1:25:28
  Speaker 1: oh yeah, this could have potential ramifications if it happened.
• 1:25:31
  Speaker 1: People were not super negative about the internet. A lot
• 1:25:35
  Speaker 1: of the skeptics were saying we're worried about an overload
• 1:25:37
  Speaker 1: of bad information.
• 1:25:38
  Speaker 2: Look at where we are.
• 1:25:39
  Speaker 1: A lot of people were worried about the social consequences
• 1:25:42
  Speaker 1: of everyone talking online, which they were correct about. With
• 1:25:45
  Speaker 1: the economic things, they were specifically talking about like the globe,
• 1:25:47
  Speaker 1: which I think made hundreds of thousands of dollars and
• 1:25:50
  Speaker 1: had like a think a billion dollar market cap they were.
• 1:25:53
  Speaker 2: Talking yeah there was a massive hype in the dot-com era.
• 1:25:56
  Speaker 1: I read a lot of those stories the hype was
• 1:25:58
  Speaker 1: nowhere near you didn't have articles everywhere that were saying
• 1:26:01
  Speaker 1: if you don't get online you'll be left behind you
• 1:26:03
  Speaker 1: didn't have professional consequences nick suresh mentioned his blog earlier
• 1:26:08
  Speaker 1: he described this thing global uh ai's of history and
• 1:26:10
  Speaker 1: global decision making where he said that you have businesses
• 1:26:14
  Speaker 1: you work at where if you don't say that you're
• 1:26:17
  Speaker 1: more productive with ai whether or not it's true is
• 1:26:19
  Speaker 1: irrelevant You have professional consequences.
• 1:26:22
  Speaker 2: You can get fired.
• 1:26:23
  Speaker 1: There are people having to AI wash their jobs by
• 1:26:25
  Speaker 1: saying AI did it. Otherwise, their bosses, who don't do shit,
• 1:26:30
  Speaker 1: will get mad at them. This did not happen with
• 1:26:32
  Speaker 1: the internet. It was not present. And part of the
• 1:26:35
  Speaker 1: thing is social media was not like it is today.
• 1:26:38
  Speaker 1: The kind of decentralization of media in general has caused
• 1:26:42
  Speaker 1: this as well. And also the fact of day trading.
• 1:26:45
  Speaker 1: There's so many different things that are different.
• 1:26:46
  Speaker 2: It's crazy. I do think AI... is different from the internet,
• 1:26:51
  Speaker 2: in part, if you just measured it on the speed
• 1:26:52
  Speaker 2: of adoption, especially if we could just think about generative AI.
• 1:26:55
  Speaker 1: But the adoption of the internet required physical connections to
• 1:27:00
  Speaker 1: your house. The adoption of generative AI involves having a
• 1:27:02
  Speaker 1: web browser. It took a vast amount of effort to
• 1:27:05
  Speaker 1: bring internet to people. Even with dial-up connections, it still
• 1:27:08
  Speaker 1: required the distribution.
• 1:27:09
  Speaker 2: And that's why it was so slow and there was
• 1:27:12
  Speaker 2: less hype than AI. I do agree that there's way
• 1:27:14
  Speaker 2: more hype. And we are, again, going back to this
• 1:27:16
  Speaker 2: point that we're clustering AI. in this big category of
• 1:27:20
  Speaker 2: lots of different things.
• 1:27:21
  Speaker 1: There's generative AI.
• 1:27:22
  Speaker 2: There's generative AI. There's like real world AI.
• 1:27:24
  Speaker 1: But generative AI is explicitly what I'm talking about here.
• 1:27:26
  Speaker 1: When bosses are saying you need to use AI, they're
• 1:27:29
  Speaker 1: not saying I need you to go and buy a
• 1:27:30
  Speaker 1: Unibeam robot. They're saying use LLMs so that I, and
• 1:27:34
  Speaker 1: that's the thing. They have this theory, the era of
• 1:27:36
  Speaker 1: the business idiot, where it's like we are ruled by
• 1:27:37
  Speaker 1: people that don't do work because nobody who actually does
• 1:27:40
  Speaker 1: a bunch of work who really is productive is harassing
• 1:27:44
  Speaker 1: someone who works for them for not being productive enough.
• 1:27:47
  Speaker 1: They don't have the time. They're doing work. Someone who
• 1:27:49
  Speaker 1: is sitting there with the ingratiation machine that's telling them
• 1:27:52
  Speaker 1: that every beautiful idea out of their messy little skull
• 1:27:55
  Speaker 1: is amazing. Yeah, they're going, damn, this thing says I'm
• 1:27:57
  Speaker 1: a genius. Why are you not using the genius machine
• 1:28:00
  Speaker 1: to do more work? And yeah, if you're a boss
• 1:28:03
  Speaker 1: that goes to lunch, leaves lunch, and sometimes reads your emails,
• 1:28:05
  Speaker 1: LLMs are magic.
• 1:28:06
  Speaker 2: Do you know one of the most compelling arguments I
• 1:28:08
  Speaker 2: have for the overhype of AI? In a world where
• 1:28:13
  Speaker 2: everybody has access to these tools, whatever the tools can do,
• 1:28:17
  Speaker 2: would largely be commoditized and what the tools can't do
• 1:28:20
  Speaker 2: which one could say is the human taste judgment you
• 1:28:24
  Speaker 2: could say it's people skills whatever you want to say
• 1:28:26
  Speaker 2: is now going to be the valuable thing because the
• 1:28:29
  Speaker 2: scarce and the hard becomes the most valuable through history
• 1:28:33
  Speaker 2: and the commoditized becomes the least valuable so the very
• 1:28:36
  Speaker 2: nature that we're commoditizing the generation of content or whatever
• 1:28:40
  Speaker 2: you want to call it code means that's actually not
• 1:28:42
  Speaker 2: where the value will accrue as for the user and
• 1:28:45
  Speaker 2: actually if you think about what it takes to now
• 1:28:48
  Speaker 2: make something that is objectively great if an ai can
• 1:28:52
  Speaker 2: do it then it's not the great thing is not
• 1:28:56
  Speaker 2: a value So I think a lot, I've been thinking
• 1:28:59
  Speaker 2: a lot actually about how do you avoid the temptation
• 1:29:04
  Speaker 2: of slopification of the things you make, the value you
• 1:29:06
  Speaker 2: put into the world. It's a very simple example that
• 1:29:09
  Speaker 2: people will be able to relate to. If you use
• 1:29:11
  Speaker 2: ChatGPT or Anthropic, you know, Claude, to make your LinkedIn posts,
• 1:29:15
  Speaker 2: let's say, they will be shit LinkedIn posts because everybody
• 1:29:18
  Speaker 2: else is using them. And actually, a great LinkedIn post
• 1:29:20
  Speaker 2: now is someone who doesn't use them and makes something
• 1:29:23
  Speaker 2: that's like irreplaceably human. Right.
• 1:29:25
  Speaker 1: And.
• 1:29:27
  Speaker 2: deeper and more personal, N of 1, lived experience. Yeah.
• 1:29:32
  Speaker 2: All these things that AI can't do. And I think
• 1:29:34
  Speaker 2: that's a compelling argument that actually... The commodity tools produce
• 1:29:38
  Speaker 2: commodity outcomes.
• 1:29:39
  Speaker 1: So everyone has access to these things.
• 1:29:40
  Speaker 2: And what's changed?
• 1:29:41
  Speaker 1: Like, really, like what?
• 1:29:42
  Speaker 2: The slopification of stuff.
• 1:29:43
  Speaker 1: We've got a bunch of slop. But these people were
• 1:29:46
  Speaker 1: half-arsing their jobs before. It's just a half-arsery machine. And
• 1:29:49
  Speaker 1: it's just, it's the thing.
• 1:29:51
  Speaker 2: It's what I'm talking about with the slop blogs. It's like,
• 1:29:53
  Speaker 2: it's it.
• 1:29:53
  Speaker 1: Yeah, people that gave you dog shit before have now
• 1:29:56
  Speaker 1: got the dog shit machine to pump out dog shit. It's,
• 1:29:59
  Speaker 1: so there's a guy called Carl Brown, Internet Bucks.
• 1:30:01
  Speaker 2: Awesome guy.
• 1:30:02
  Speaker 1: Great software engineer.
• 1:30:03
  Speaker 2: Yeah.
• 1:30:04
  Speaker 1: He said it. I might have said this earlier. So
• 1:30:05
  Speaker 1: it makes the easy things easy, the hard things harder.
• 1:30:07
  Speaker 1: When you know you're doing a really distinct small script
• 1:30:10
  Speaker 1: for something and it can plot that out, it's awesome.
• 1:30:12
  Speaker 1: I used Claude the other day for something useful. My
• 1:30:15
  Speaker 1: kid loves Minecraft. I was trying to fix a fucking
• 1:30:17
  Speaker 1: broken mod because he loves his Witherstorm.
• 1:30:19
  Speaker 2: It's awesome.
• 1:30:20
  Speaker 1: And it still took me half an hour and kept
• 1:30:22
  Speaker 1: getting things wrong.
• 1:30:23
  Speaker 2: What do you use AI for, generative AI?
• 1:30:25
  Speaker 1: I really don't.
• 1:30:26
  Speaker 2: You don't use it?
• 1:30:27
  Speaker 1: With Bloomberg Terminal, I use AskB, which is just when
• 1:30:29
  Speaker 1: it's like, requesting the consensus analyst estimates for.
• 1:30:33
  Speaker 2: Nvidia otherwise you don't use it no so how do
• 1:30:35
  Speaker 2: you know it's bad.
• 1:30:36
  Speaker 1: I've used it i've put it through its paces i've
• 1:30:38
  Speaker 1: used it to try and do financial models and found
• 1:30:40
  Speaker 1: one error and immediately be like ah i've never been
• 1:30:43
  Speaker 1: particularly impressed the one thing i will defend it on
• 1:30:45
  Speaker 1: is it's really good for like tech support like i
• 1:30:47
  Speaker 1: have this thing called synergy in my new york new
• 1:30:50
  Speaker 1: york place i go to i have this monitor where
• 1:30:52
  Speaker 1: i have a macbook and a pc laptop and this
• 1:30:54
  Speaker 1: thing synergy for using the same mouse and keyboard drop
• 1:30:57
  Speaker 1: in a giant fucking troubleshooting log into this thing going
• 1:31:01
  Speaker 1: what's wrong and it going this is wrong yeah it's
• 1:31:04
  Speaker 1: super useful is that trillion dollars not is that a
• 1:31:06
  Speaker 1: two trillion dollar company no pretty better than google though
• 1:31:09
  Speaker 1: right better than google no i mean uh yeah remember
• 1:31:12
  Speaker 1: do you use google search still i try i have
• 1:31:15
  Speaker 1: to fucking push the crap out of the way and i.
• 1:31:17
  Speaker 2: Can't remember the last time i did a google search.
• 1:31:20
  Speaker 1: Christ i find myself using bing sometimes i know i
• 1:31:23
  Speaker 1: hate saying it too but i have to scroll past
• 1:31:26
  Speaker 1: the ai crap because i want the good stuff i
• 1:31:28
  Speaker 1: want the I want the actual links to stuff so
• 1:31:30
  Speaker 1: that I can read the thing and go.
• 1:31:32
  Speaker 2: But you can ask the AI to give you the links. Yeah.
• 1:31:36
  Speaker 1: And it doesn't do a particularly good job.
• 1:31:39
  Speaker 2: Say that the other day, my iPad wasn't turning on
• 1:31:41
  Speaker 2: and it was doing this funny little thing on the screen.
• 1:31:43
  Speaker 2: You think that it's better to type that into Google?
• 1:31:45
  Speaker 3: Oh, no.
• 1:31:46
  Speaker 1: I must be clear. That may be the only LLM
• 1:31:49
  Speaker 1: use case I defend. The troubleshooting thing is awesome for it.
• 1:31:52
  Speaker 1: It's the one weakness I have. It's like genuinely being
• 1:31:55
  Speaker 1: able to drop a log into it. That's awesome. Again,
• 1:31:58
  Speaker 1: that is not what they're selling it as. They're not
• 1:32:00
  Speaker 1: selling it as a useful little tool. They're selling it
• 1:32:03
  Speaker 1: as the software, as the thing that will change everything,
• 1:32:07
  Speaker 1: that will replace all jobs, that will do this and that.
• 1:32:10
  Speaker 1: It's not like they sold it as a quirky bit
• 1:32:12
  Speaker 1: of software.
• 1:32:12
  Speaker 2: No, you are right. They are telling us that it
• 1:32:15
  Speaker 2: is going to replace everything. But funnily enough, the critics
• 1:32:17
  Speaker 2: are saying that as well. Which one? They are, like
• 1:32:20
  Speaker 2: the Geoffrey Hintons of the world. You know, even people
• 1:32:22
  Speaker 2: that have left the safety team in ChatGPT who I've
• 1:32:25
  Speaker 2: sat here with, these are critics that are warning of
• 1:32:29
  Speaker 2: the impacts it's going to have on the world.
• 1:32:31
  Speaker 1: It's weird how all these critics also have a vested
• 1:32:34
  Speaker 1: interest in AI doing well, though. Daniel, former OpenAI guy,
• 1:32:37
  Speaker 1: AI 2027, written with the Star Codex guy that was
• 1:32:40
  Speaker 1: nothing more than badly written science fiction that he's already
• 1:32:43
  Speaker 1: had to walk back.
• 1:32:44
  Speaker 2: But you know he could have made more money by
• 1:32:45
  Speaker 2: staying at ChatGPT. Could he? I mean, it looks like
• 1:32:49
  Speaker 2: if he had options early, sticking around.
• 1:32:52
  Speaker 1: Would have made him. Did he lose the options? How
• 1:32:54
  Speaker 1: much did he lose?
• 1:32:54
  Speaker 2: You're not saying that they're being critical.
• 1:32:57
  Speaker 1: They're not critical of the companies themselves. They're not critical
• 1:33:00
  Speaker 1: of the stealing. They're not critical of the environmental damage.
• 1:33:02
  Speaker 1: They're not critical of the fact that you cannot rely
• 1:33:05
  Speaker 1: on the answers. They're critical of this big, scary boogeyman
• 1:33:08
  Speaker 1: out in the future where it's like, oh, I'm scared
• 1:33:12
  Speaker 1: of when this becomes so powerful and everyone should talk
• 1:33:14
  Speaker 1: to me about how scary and powerful it is. They're
• 1:33:16
  Speaker 1: not saying, hey, here are the harms today. Here are
• 1:33:19
  Speaker 1: the things we're actually looking at today. Here are the
• 1:33:21
  Speaker 1: social problems of having this automated way of spewing out slop,
• 1:33:26
  Speaker 1: of filling our feeds with crap, of having information that
• 1:33:30
  Speaker 1: will pop up that is presented, even with the little
• 1:33:32
  Speaker 1: disclaimer thing of saying, yeah, sometimes this gets you wrong.
• 1:33:34
  Speaker 1: So in the tiniest words possible, they don't talk about
• 1:33:37
  Speaker 1: the fact that That these things are trained on stealing
• 1:33:40
  Speaker 1: millions of people's work.
• 1:33:41
  Speaker 2: But on that last point where you say that it's
• 1:33:43
  Speaker 2: going to get progressively more intelligent and when it does,
• 1:33:45
  Speaker 2: it will be a danger. Yeah. Would you agree with
• 1:33:48
  Speaker 2: the statement that artificial intelligence has gotten more intelligent if
• 1:33:53
  Speaker 2: you measure it based on any sort of measure of
• 1:33:55
  Speaker 2: intelligence one might use?
• 1:33:57
  Speaker 1: It's got better on the tests that are rigged for
• 1:33:59
  Speaker 1: the models. It's got better at tests where you can
• 1:34:01
  Speaker 1: train for the test.
• 1:34:03
  Speaker 2: Okay, so it's got better at... It's got better at.
• 1:34:05
  Speaker 1: Tests that they're intentionally trained for.
• 1:34:07
  Speaker 2: So if you logged the rate of improvement on a graph,
• 1:34:10
  Speaker 2: it would look something like this.
• 1:34:13
  Speaker 1: Right. You agree?
• 1:34:14
  Speaker 2: In terms of what it's capable of doing.
• 1:34:16
  Speaker 1: Oh, that's... There we go.
• 1:34:18
  Speaker 2: Yeah.
• 1:34:18
  Speaker 1: Because it's not got new features. You'll notice that outside
• 1:34:22
  Speaker 1: of OpenAI and Anthropic, when you remove the coding startups,
• 1:34:26
  Speaker 1: there's basically no successful AI startup company.
• 1:34:29
  Speaker 2: So we agree that it's got better. It's got more capable.
• 1:34:34
  Speaker 2: At doing things. Yeah, fine. Over time, AI has got
• 1:34:37
  Speaker 2: more capable. If we imagine that trajectory will continue, it
• 1:34:42
  Speaker 2: will get more capable. Then at some point it does cross,
• 1:34:46
  Speaker 2: you know, this is what they say to me, it
• 1:34:48
  Speaker 2: crosses human intelligence. And at such time, will it not
• 1:34:52
  Speaker 2: start to do some of the jobs that people are
• 1:34:54
  Speaker 2: doing today?
• 1:34:55
  Speaker 1: Outside of software engineering. Remove software, because I will concede
• 1:34:58
  Speaker 1: software engineering, it's got better at that. Outside of software engineering. Yeah. Where?
• 1:35:02
  Speaker 2: So the chief of staff things, the admins.
• 1:35:04
  Speaker 1: Okay.
• 1:35:04
  Speaker 2: So it's got better than admin. Video generation, photo generation. Okay.
• 1:35:08
  Speaker 2: Text generation, theoretically. Coding.
• 1:35:11
  Speaker 3: Right.
• 1:35:12
  Speaker 2: And then I'd say agentic workflows.
• 1:35:14
  Speaker 1: What is an agentic workflow?
• 1:35:15
  Speaker 2: So automated workflows where you're doing the same. I mean,
• 1:35:18
  Speaker 2: a good example is looking at the backend data of
• 1:35:21
  Speaker 2: the dire of a CEO.
• 1:35:22
  Speaker 1: Summarizing.
• 1:35:23
  Speaker 2: Looking at all of the data, adjusting all of it,
• 1:35:24
  Speaker 2: going out onto the internet and searching who Ed is,
• 1:35:26
  Speaker 2: looking at every interview you've ever done ever. Uh-huh.
• 1:35:29
  Speaker 1: This is summarizing and generating.
• 1:35:32
  Speaker 2: Making a little model on, you know, the things people
• 1:35:34
  Speaker 2: want to know from Ed. Producing a report, sending that
• 1:35:36
  Speaker 2: to my inbox. Me getting a 20, 30, 40, 50
• 1:35:39
  Speaker 2: page report on Ed before he arrives. This is all.
• 1:35:42
  Speaker 1: Basically the same thing it's been doing for years though.
• 1:35:43
  Speaker 1: It's not really new capabilities.
• 1:35:45
  Speaker 2: Research.
• 1:35:47
  Speaker 1: It's still the same things. They've had web search for years.
• 1:35:50
  Speaker 1: They've had report generation for years.
• 1:35:52
  Speaker 2: But we couldn't generate high quality videos that are like
• 1:35:56
  Speaker 2: indistinguishable from cameras. Seed Dance and these ones that look
• 1:36:00
  Speaker 2: like movies are incredible. So I'm saying the point I'm
• 1:36:03
  Speaker 2: trying to make is that if we imagine that over
• 1:36:05
  Speaker 2: the last 10 years, there has been a rate of
• 1:36:06
  Speaker 2: improvement in terms of capabilities and output and quality. We've
• 1:36:10
  Speaker 2: seen hallucinations drop. We've seen the models get more, quote unquote, intelligent,
• 1:36:14
  Speaker 2: get better. You know, if you give it an IQ test,
• 1:36:16
  Speaker 2: it's getting higher scores than it was 10 years ago.
• 1:36:18
  Speaker 2: We agree that there's been... A upward motion of improvement.
• 1:36:21
  Speaker 2: This is pretty much how machine learning goes when you
• 1:36:23
  Speaker 2: feed it more data. Exactly. And you put more compute
• 1:36:25
  Speaker 2: behind it. So if this continues, what does the future
• 1:36:30
  Speaker 2: look like? So the rebuttal I was expecting to hear
• 1:36:33
  Speaker 2: is that it won't continue.
• 1:36:34
  Speaker 1: And I actually don't think it. I think that there
• 1:36:37
  Speaker 1: are hard limits that we're going to hit.
• 1:36:39
  Speaker 2: So you do believe that there's a hard limit somewhere.
• 1:36:41
  Speaker 1: We've kind of already hit the diminishing returns level because,
• 1:36:45
  Speaker 1: for example, video generation, which is, by the way, Far
• 1:36:48
  Speaker 1: less an American concern anymore. OpenAI shut down Sora. I
• 1:36:52
  Speaker 1: think you can still use the API. But nevertheless, look
• 1:36:55
  Speaker 1: around you with the amount of stuff and the crew
• 1:36:57
  Speaker 1: you need to get a shot. People think that movies
• 1:36:59
  Speaker 1: are just shot by shot by shot and they just
• 1:37:01
  Speaker 1: magically happen. When you've got a wonderful girlfriend, first ADs,
• 1:37:05
  Speaker 1: assistant directors, you've got gaffers, you've got lighters. And also,
• 1:37:08
  Speaker 1: simulating light is insanely difficult. There are so many magical
• 1:37:12
  Speaker 1: things that happen in creating visual images that, yeah, you
• 1:37:15
  Speaker 1: could create a one-minute long thing that might fool someone.
• 1:37:18
  Speaker 1: How do you practically turn that into a movie? Because
• 1:37:20
  Speaker 1: that movie, I forget what the name is. There was
• 1:37:21
  Speaker 1: a movie that claimed it aired at Cannes. It didn't.
• 1:37:25
  Speaker 1: It aired in the city of Cannes during the Cannes
• 1:37:28
  Speaker 1: Film Festival. It was not at the film festival. When
• 1:37:30
  Speaker 1: it comes to the practical creation of actual things at
• 1:37:32
  Speaker 1: the end of it versus magic tricks, the actual practical
• 1:37:35
  Speaker 1: outcomes are not there. The reason I keep coming back
• 1:37:37
  Speaker 1: to the capabilities thing for the example is, yeah, they
• 1:37:39
  Speaker 1: can do better at tests, do better, number go up.
• 1:37:43
  Speaker 1: When it comes to, can this actually do distinct tasks
• 1:37:45
  Speaker 1: you can rely on it? You can rely on it
• 1:37:48
  Speaker 1: for summaries. You can rely on it for generations. The
• 1:37:50
  Speaker 1: things it was doing, it's getting linearly-ish better at. But again,
• 1:37:55
  Speaker 1: there's a ceiling to that. Like, okay, so it gets
• 1:37:57
  Speaker 1: really good at research. What does that actually mean? You've
• 1:37:59
  Speaker 1: already kind of got the automation there. What is the
• 1:38:02
  Speaker 1: next step of that? Because training it to be more autonomous,
• 1:38:04
  Speaker 1: for example, that's not something that comes from training data.
• 1:38:07
  Speaker 1: That is actually a new, Gary Marcus, a neuro-symbolic. You
• 1:38:11
  Speaker 1: actually need to build a structure around the AI to
• 1:38:13
  Speaker 1: make it work. And even then, It doesn't fix the.
• 1:38:15
  Speaker 2: So you're saying that there will become a point where
• 1:38:18
  Speaker 2: the rate of improvement will plateau... We're already there. and stop.
• 1:38:22
  Speaker 1: We've already hit that diminishing... Gary Marcus said this in
• 1:38:25
  Speaker 1: 2022 as well.
• 1:38:25
  Speaker 2: You know there's lots of people listening now that, like,
• 1:38:28
  Speaker 2: they've had their workflows completely transformed by these tools. Have they?
• 1:38:32
  Speaker 2: There'll be people. Yeah, there are.
• 1:38:33
  Speaker 1: Yeah, the thing is, first of all, every single one
• 1:38:36
  Speaker 1: of them, did you pay for the tokens? That's the thing.
• 1:38:38
  Speaker 1: Did you pay for the tokens? And also, how many
• 1:38:41
  Speaker 1: tokens did you burn? But putting all that aside, what workflows?
• 1:38:43
  Speaker 1: Because if it's, yeah, I did a bunch of web
• 1:38:44
  Speaker 1: scraping and web searches, I'm just not impressed. Did you
• 1:38:47
  Speaker 1: make an entire fucking movie?
• 1:38:49
  Speaker 2: No, you didn't.
• 1:38:50
  Speaker 1: Is it speeding up your coding yeah i believe that
• 1:38:53
  Speaker 1: i've heard that from multiple people but again how much
• 1:38:56
  Speaker 1: can you trust this i think i.
• 1:38:57
  Speaker 2: Was getting at is you know when in the moment
• 1:39:01
  Speaker 2: of any technological innovation people they extrapolate linearly or they
• 1:39:08
  Speaker 2: view it as a static state i either think today
• 1:39:10
  Speaker 2: is going to look like tomorrow or they think it's
• 1:39:11
  Speaker 2: going to get better in this sort of straight line
• 1:39:13
  Speaker 2: but what we end up seeing a lot of the
• 1:39:15
  Speaker 2: time is this exponential improvement.
• 1:39:17
  Speaker 1: All of the innovations we're talking about with you with
• 1:39:18
  Speaker 1: like with compute and all that with fast processors those
• 1:39:21
  Speaker 1: are hardware breakthroughs the hardware breakthrough companies don't seem to
• 1:39:25
  Speaker 1: be fixing the llm problems despite the all the king's
• 1:39:28
  Speaker 1: horses all the king's men we were nine ten generations
• 1:39:31
  Speaker 1: of tpus from google now broadcoms building stuff with open
• 1:39:34
  Speaker 1: ai their jalapeno chip and yet none of these people
• 1:39:38
  Speaker 1: could just say yeah we're on the path to making
• 1:39:39
  Speaker 1: this profitable because they can't if we fix the environmental
• 1:39:42
  Speaker 1: problems and the profitability situation maybe i'd be more generous
• 1:39:46
  Speaker 1: with this stuff but they don't seem to be able
• 1:39:49
  Speaker 1: to and you talk about these improvements and capabilities there's
• 1:39:53
  Speaker 1: a certain point at which i'm saying okay can it
• 1:39:56
  Speaker 1: do even a tenth of the stuff they're promising samuel
• 1:39:59
  Speaker 1: when the other fucking week was saying it was going
• 1:40:00
  Speaker 1: to be in like six months will be like a
• 1:40:02
  Speaker 1: genie that you can ask wishes for from it's like
• 1:40:06
  Speaker 1: motherfuckers never watched aladdin what's he talking about like also
• 1:40:09
  Speaker 1: the genie was charming anyway long story short the promises
• 1:40:13
  Speaker 1: do not line up with the capabilities or the capability
• 1:40:16
  Speaker 1: improvements an exponential improvement in software and software performance is
• 1:40:22
  Speaker 1: always a result of direct hardware improvement. We have all
• 1:40:26
  Speaker 1: the gifted mathematicians, all the gifted software engineers, all the
• 1:40:29
  Speaker 1: gifted hardware engineers.
• 1:40:31
  Speaker 2: And where are we?
• 1:40:32
  Speaker 1: Trillion plus dollars in with the future great financial crisis
• 1:40:36
  Speaker 1: and the world's greatest marketing psyop.
• 1:40:38
  Speaker 2: I just think in the future, I do think that
• 1:40:40
  Speaker 2: all of the devices and the computers we use and
• 1:40:43
  Speaker 2: the physical items in our world will be more intelligent.
• 1:40:45
  Speaker 2: I mean, sure, but is that LLMs? That'll be powered
• 1:40:49
  Speaker 2: by the underlying AI infrastructure. It'll be the data centers.
• 1:40:53
  Speaker 2: It'll be energy coming down.
• 1:40:54
  Speaker 1: How does a GPU full data center translate to a
• 1:41:00
  Speaker 1: Nikon camera that can, I don't know even what you'd think.
• 1:41:05
  Speaker 1: Because what is the thing we're talking about here? Because
• 1:41:08
  Speaker 1: the idea that devices will get smarter. Sure. I can
• 1:41:11
  Speaker 1: see that. It's a very broad statement. I could see
• 1:41:13
  Speaker 1: it happening. It's really kind of happening. What does that
• 1:41:15
  Speaker 1: have to do with the data centers? Because these data centers, again,
• 1:41:18
  Speaker 1: are not being built to make your consumer electronics smarter.
• 1:41:21
  Speaker 1: They're not being built for anything other than speculating on
• 1:41:25
  Speaker 1: the ability to capture demand for generative AI services.
• 1:41:27
  Speaker 2: But it's not just generative AI. We went through that.
• 1:41:29
  Speaker 3: It is.
• 1:41:29
  Speaker 1: No, but those data centers, they are being built for
• 1:41:32
  Speaker 1: generative AI. They are not being built for anything else.
• 1:41:34
  Speaker 2: Would you consider generative AI to be the fact that
• 1:41:37
  Speaker 2: on Meta's earnings call a couple of weeks ago, Mark
• 1:41:40
  Speaker 2: Zuckerberg said the big breakthrough we've had, which has resulted
• 1:41:42
  Speaker 2: in 15 basis points of increased retention, I believe he
• 1:41:46
  Speaker 2: was referring to Instagram, is that we now take anything
• 1:41:49
  Speaker 2: you post on social media and we run it through
• 1:41:52
  Speaker 2: an AI to get full context of what it is.
• 1:41:55
  Speaker 2: And because we can see a guy sat in front
• 1:41:57
  Speaker 2: of me called Ed with blue shirt and coffee, we
• 1:42:01
  Speaker 2: now can train AI to serve whoever wants blue shirt
• 1:42:04
  Speaker 2: Ed with coffee and To the right user, which means
• 1:42:07
  Speaker 2: people are retained longer because- Isn't 15 basis points like 0.15%? Yeah,
• 1:42:11
  Speaker 2: it's not cool, but it makes a difference at scale.
• 1:42:14
  Speaker 2: It makes a big difference at scale.
• 1:42:15
  Speaker 1: Yeah, but $ 100 and something billion in and the best
• 1:42:17
  Speaker 1: you've got is 0.15%.
• 1:42:18
  Speaker 2: It makes a big difference.
• 1:42:19
  Speaker 1: If he could be fired, I mean, how much of
• 1:42:22
  Speaker 1: a difference? Because there's a reason he's saying basis points
• 1:42:26
  Speaker 1: versus dollars. Because think about it like this. If Mark
• 1:42:30
  Speaker 1: Zuckerberg was- I take your point about scale.
• 1:42:33
  Speaker 2: I'm saying the point I was making was that that
• 1:42:35
  Speaker 2: is another application of these data centers because it needs
• 1:42:39
  Speaker 2: a data center that is driving revenues, but also that's
• 1:42:45
  Speaker 2: outside of us thinking about just generating a sentence.
• 1:42:47
  Speaker 1: And that's the gem, the generative model.
• 1:42:49
  Speaker 2: Muse.
• 1:42:50
  Speaker 1: Oh, Muse Spark is their LLM. Gem is their generative.
• 1:42:53
  Speaker 2: It's Muse.
• 1:42:54
  Speaker 1: Well, Muse, then that's them doing the weird thing where
• 1:42:57
  Speaker 1: it's like on Instagram and it's like Dave the Cat.
• 1:42:59
  Speaker 1: Why is Dave the Cat suffering? Like it's the weird
• 1:43:02
  Speaker 1: pop-up things. Meta is fucked. God damn, that company sucks.
• 1:43:05
  Speaker 1: Like every time I think about how they've ruined that product.
• 1:43:07
  Speaker 1: But that's the thing though. Again, why can't he just
• 1:43:09
  Speaker 1: say with his whole chest, we've made a couple billion?
• 1:43:12
  Speaker 1: Why can't he say that? Because he isn't. Because there's
• 1:43:14
  Speaker 1: not actually a way of going, I spent all this money.
• 1:43:17
  Speaker 1: I spent 14 billion goddamn dollars on scale by Alexander Wong.
• 1:43:21
  Speaker 1: And I made this much.
• 1:43:23
  Speaker 2: They can't.
• 1:43:24
  Speaker 1: It gets back to a very simple point of, hey,
• 1:43:27
  Speaker 1: if it was going well, you'd tell me how well
• 1:43:29
  Speaker 1: it was going rather than, I don't know, doing this
• 1:43:32
  Speaker 1: weird rain dance thing where you're like, well, if we
• 1:43:34
  Speaker 1: move all the pieces around in three years, theoretically, this
• 1:43:37
  Speaker 1: will happen.
• 1:43:39
  Speaker 2: I've done almost 700 interviews with some of the most
• 1:43:42
  Speaker 2: interesting people in the world. And one of the things
• 1:43:44
  Speaker 2: you learn, which is unexpected, is that vulnerability is the
• 1:43:48
  Speaker 2: doorway to connection. And after sitting here for two, three
• 1:43:50
  Speaker 2: hours with a guest, I feel a deep sense of
• 1:43:53
  Speaker 2: connection to them. And as they leave, what I get
• 1:43:56
  Speaker 2: them to do is to write a question in the
• 1:43:59
  Speaker 2: diary of a CEO. We've taken all of the questions
• 1:44:02
  Speaker 2: from the diary of a CEO. We have put the
• 1:44:05
  Speaker 2: question here on this card with the name of the
• 1:44:09
  Speaker 2: person that wrote it. So you can sit at home
• 1:44:11
  Speaker 2: as I do with my fiance and my colleagues at
• 1:44:13
  Speaker 2: work and other people in my life. Whenever we get
• 1:44:15
  Speaker 2: a minute, we play the diary of a CEO conversation cards.
• 1:44:19
  Speaker 2: And it is incredible. What happens? These are great if
• 1:44:23
  Speaker 2: you're in a romantic relationship and you want to connect
• 1:44:25
  Speaker 2: your partner more. These are also great if you're in
• 1:44:27
  Speaker 2: a team and you want to bond your team together.
• 1:44:29
  Speaker 2: And I have to say they're also great for families
• 1:44:31
  Speaker 2: that want to learn more about each other and that
• 1:44:33
  Speaker 2: need a good excuse to spend some time in a
• 1:44:36
  Speaker 2: digital world, in the analog environment, connecting human to human.
• 1:44:41
  Speaker 2: It is remarkable what the right question at the right
• 1:44:44
  Speaker 2: time can do. Go to thediary.com and you can get
• 1:44:48
  Speaker 2: these conversation cards online. There should be a button just
• 1:44:52
  Speaker 2: down below here. And if it says subscribe, you're already subscribed.
• 1:44:56
  Speaker 2: If it says subscriber, that means you're not yet. And
• 1:44:59
  Speaker 2: if you're not subscribed, please could you do us a
• 1:45:00
  Speaker 2: favor and hit that button? It helps the show more
• 1:45:02
  Speaker 2: than you know. And according to the algorithm, you're someone
• 1:45:05
  Speaker 2: that watches our show, but you haven't yet hit that button.
• 1:45:07
  Speaker 2: Thank you so much. I do think you're accurate and
• 1:45:10
  Speaker 2: right when you talk about the fact that there's a
• 1:45:12
  Speaker 2: lot of like, is the word for gazey?
• 1:45:14
  Speaker 1: Yeah.
• 1:45:14
  Speaker 2: Where like, there's a lot of people that have spent
• 1:45:16
  Speaker 2: a lot of money and they kind of shouldn't have
• 1:45:17
  Speaker 2: spent it and they fucked that up. And now they're thinking, shit,
• 1:45:20
  Speaker 2: like we've spent all this investor money. Kind of like
• 1:45:22
  Speaker 2: the Metaverse was a bit of a fucking. Oh my God.
• 1:45:24
  Speaker 2: That was a bit of a joke. That's so weird.
• 1:45:25
  Speaker 2: A lot of money spent. We kind of thought this
• 1:45:27
  Speaker 2: dream was coming of this one. I shouldn't say dream
• 1:45:29
  Speaker 2: because it's not a dream that I've had. Dream that
• 1:45:31
  Speaker 2: they had. Yeah. This sort of virtual world. And actually
• 1:45:33
  Speaker 2: it never transpired. And there's no sign that it will
• 1:45:35
  Speaker 2: in the near term. AI and the dot-com boom in
• 1:45:40
  Speaker 2: this regard are the same. NFTs were the same. Right.
• 1:45:43
  Speaker 2: So crypto, one could argue that a lot of the
• 1:45:44
  Speaker 2: crypto industry was the same. It's Wang that is inflated
• 1:45:47
  Speaker 2: by the media.
• 1:45:48
  Speaker 1: The difference is, the reason the metaverse and NFTs didn't
• 1:45:52
  Speaker 1: escape this was there weren't stocks to speculate on. There
• 1:45:55
  Speaker 1: weren't big companies that you could invest in. They had
• 1:45:58
  Speaker 1: record earnings in 2021. There's a bunch of money floating
• 1:46:01
  Speaker 1: in the system thanks to post-COVID, the PRDC, that basically government,
• 1:46:05
  Speaker 1: federal money flowed in to the banks. There's a bunch
• 1:46:07
  Speaker 1: of easy money, zero interest-free era. Money was easy to find.
• 1:46:11
  Speaker 1: Then after that, there was the hangover. Growth started to
• 1:46:13
  Speaker 1: slow down dramatically. This is actually my rock-con bubble theory,
• 1:46:16
  Speaker 1: which is, They don't have any hyper growth ideas anymore.
• 1:46:19
  Speaker 1: So suddenly they started buying GPUs. And when they bought GPUs,
• 1:46:23
  Speaker 1: people went, they're doing AI. Oh, we better buy the stock.
• 1:46:26
  Speaker 1: And the stock's running an incredible run. Meta's like several
• 1:46:29
  Speaker 1: hundred percent growth. In the last few years, their stock
• 1:46:31
  Speaker 1: has grown by hundreds of percent.
• 1:46:33
  Speaker 2: Despite zero proof.
• 1:46:35
  Speaker 1: And because the media was just saying, yeah, Meta's revenue
• 1:46:39
  Speaker 1: is growing because of AI, right?
• 1:46:40
  Speaker 2: Microsoft's revenue is growing because of AI, right? The fugazi
• 1:46:43
  Speaker 2: you're talking about.
• 1:46:44
  Speaker 1: Was the fact that everyone just gave them credit in advance.
• 1:46:47
  Speaker 1: And now we're kind of getting to the point where
• 1:46:48
  Speaker 1: it's like, hey, you didn't spend that trillion dollars for
• 1:46:51
  Speaker 1: no reason. Did you, Satya? Amy Hood just going to
• 1:46:55
  Speaker 1: take him out back, send him to the glue factory
• 1:46:57
  Speaker 1: or something?
• 1:46:57
  Speaker 2: I do think there's overspending. I want to concede that. Dramatic. Yeah, no,
• 1:47:01
  Speaker 2: I do think there is. And I think the reason
• 1:47:03
  Speaker 2: why there's overspending, Ed, is I think there is something here.
• 1:47:08
  Speaker 2: And- What? In terms of like, I think there is
• 1:47:12
  Speaker 2: practical uses for this technology. And I think when people
• 1:47:15
  Speaker 2: realize that through history... they go crazy because they want
• 1:47:18
  Speaker 2: to be the person that owns the opportunity.
• 1:47:20
  Speaker 1: I'm going to be honest. I just, I fundamentally don't agree.
• 1:47:22
  Speaker 2: You don't agree with which part?
• 1:47:24
  Speaker 1: I don't agree that this, that the speculation is a
• 1:47:26
  Speaker 1: result of actual demand. I don't believe it's suspect. I
• 1:47:29
  Speaker 1: don't think private credit is sinking hundreds of billions of
• 1:47:31
  Speaker 1: dollars into AI because of actual demand. They are doing
• 1:47:33
  Speaker 1: it because they saw the biggest companies in the world
• 1:47:35
  Speaker 1: building data centers, making a ton of money from two companies.
• 1:47:38
  Speaker 1: They feed money and went, I want some of that money.
• 1:47:40
  Speaker 2: I'm saying that I do think there is value in
• 1:47:43
  Speaker 2: the underlying technology. And so I think, I'm not saying
• 1:47:47
  Speaker 2: how much value. Right. Okay.
• 1:47:49
  Speaker 1: I get you meaning. That's fair.
• 1:47:50
  Speaker 2: I'm not saying it's proportionate to the investment. All I'm
• 1:47:53
  Speaker 2: saying is that... Do you know what it's like? It's like...
• 1:47:56
  Speaker 2: If I take your example, the rot economy essay that
• 1:47:57
  Speaker 2: you wrote. Yeah. Say that you're on a desert island
• 1:48:00
  Speaker 2: and then someone says they found a banana tree. Right.
• 1:48:03
  Speaker 2: And there's 10,000 people on the island. Okay. They are
• 1:48:08
  Speaker 2: going to stam fucking peed towards where they think the
• 1:48:11
  Speaker 2: banana tree is. They are going to fucking... claw each
• 1:48:14
  Speaker 2: other to pieces. And if your essay here is right,
• 1:48:16
  Speaker 2: that there was desperation because they hadn't found an innovation
• 1:48:19
  Speaker 2: in a while... maybe that explains it. Maybe there is
• 1:48:21
  Speaker 2: a bit of value here. Right. And they're scam fucking
• 1:48:24
  Speaker 2: peeding and killing each other and making irrational decisions like
• 1:48:27
  Speaker 2: hungry people would.
• 1:48:28
  Speaker 1: I actually think we're, then we actually agree. That is
• 1:48:31
  Speaker 1: actually my point, which is. These three companies are meta.
• 1:48:34
  Speaker 1: Their main business lines are running out of growth. There's
• 1:48:36
  Speaker 1: only so much they can grow. And indeed, in the
• 1:48:38
  Speaker 1: next three and a half years, analysts think that these
• 1:48:40
  Speaker 1: two bastards, these two, OpenAI and Anthropic, are going to
• 1:48:42
  Speaker 1: spend over $ 400 billion on these people alone. Microsoft, Google,
• 1:48:47
  Speaker 1: and Amazon. And the crazy thing is, that's a large
• 1:48:49
  Speaker 1: part of their future growth. And if this money isn't spent,
• 1:48:52
  Speaker 1: their growth slows down. So your point about the bananas,
• 1:48:55
  Speaker 1: I actually agree. That is the rock-con bubble. They don't
• 1:48:57
  Speaker 1: have a new thing and they're desperate. And indeed, they
• 1:49:00
  Speaker 1: got rewarded for buying the GPUs. They got when they
• 1:49:03
  Speaker 1: bought these goddamn GPUs from Nvidia, all the markets went
• 1:49:07
  Speaker 1: rock hard overnight. They loved it. There were stories about
• 1:49:09
  Speaker 1: how they were sending armored cars with the GPUs to
• 1:49:12
  Speaker 1: Microsoft to make sure Microsoft got the GPUs. And so
• 1:49:15
  Speaker 1: everyone saw all that money flowing in, even though they
• 1:49:17
  Speaker 1: never disclosed AI revenues. They saw the expenditures and they went, well,
• 1:49:21
  Speaker 1: I want to do what these people are doing. I
• 1:49:23
  Speaker 1: want to get a little of that money, don't I?
• 1:49:25
  Speaker 2: I think the area where we have a slight disagreement
• 1:49:27
  Speaker 2: is that I think the underlying technology has a lot
• 1:49:31
  Speaker 2: more promise over the long term than you do.
• 1:49:34
  Speaker 1: So the thing I want to push back on there
• 1:49:36
  Speaker 1: is to have progress with AI, just taking it in
• 1:49:41
  Speaker 1: a vacuum, to have progress for these two companies to
• 1:49:44
  Speaker 1: keep going and to keep progressing, they need to spend
• 1:49:47
  Speaker 1: tens of billions of dollars a year on training. The
• 1:49:50
  Speaker 1: only way that that can happen is if these companies
• 1:49:53
  Speaker 1: and venture capitalists and private credit firms and NVIDIA keep
• 1:49:56
  Speaker 1: circulating money to them. So the progress that we've got
• 1:50:00
  Speaker 1: so far is entirely a result of this circular system.
• 1:50:03
  Speaker 1: So it means that without- Circular?
• 1:50:05
  Speaker 2: You talked about VCs there. Venture capitalists who are, by
• 1:50:08
  Speaker 2: the way- Investors.
• 1:50:09
  Speaker 1: The majority of the funding that OpenAI got in the
• 1:50:12
  Speaker 1: last six months came from SoftBank, Nvidia, and Amazon.
• 1:50:16
  Speaker 2: Okay, yeah.
• 1:50:16
  Speaker 1: So just the point is, you're talking about progress continuing.
• 1:50:20
  Speaker 1: Progress in LLMs can only continue- As long as the
• 1:50:23
  Speaker 1: money keeps flowing.
• 1:50:24
  Speaker 2: I completely agree.
• 1:50:26
  Speaker 1: Once the money stops flowing, the progress stops.
• 1:50:28
  Speaker 2: But isn't that most like early... Spotify didn't make money
• 1:50:31
  Speaker 2: for 20 years.
• 1:50:31
  Speaker 1: Spotify didn't lose $ 20. 9 billion in one year. They didn't
• 1:50:35
  Speaker 1: need to raise $ 217 billion in the space of six months.
• 1:50:38
  Speaker 2: Yeah, and Uber's another example.
• 1:50:39
  Speaker 1: $ 33 billion since inception before it became a messy kind
• 1:50:43
  Speaker 1: of profitable. Amazon Web Services between 2003 and 2015 when
• 1:50:46
  Speaker 1: it became profitable, $ 29.
• 1:50:47
  Speaker 2: 7 billion.
• 1:50:48
  Speaker 1: They'd spent. Yeah, that's the total capital expenditures. And that's
• 1:50:52
  Speaker 1: not just Amazon Web Services. That's the entire logistics operation.
• 1:50:55
  Speaker 1: Normalized for inflation.
• 1:50:56
  Speaker 2: So they all lost money for a long period of
• 1:50:58
  Speaker 2: time as the TLDR.
• 1:50:59
  Speaker 1: Yes, but the amount of money they lost is completely
• 1:51:05
  Speaker 1: just magnitudes different on a level where.
• 1:51:08
  Speaker 2: Can I argue then that that's because the potential of
• 1:51:13
  Speaker 2: intelligence permeates everything, whereas Amazon at the time was like
• 1:51:16
  Speaker 2: selling books? No. That was bringing retail online.
• 1:51:19
  Speaker 1: When Amazon Web Services grew, it was- Oh, Amazon Web Services.
• 1:51:22
  Speaker 1: Amazon Web Services.
• 1:51:23
  Speaker 2: Amazon Cloud.
• 1:51:24
  Speaker 1: With Amazon Web Services, the reason I bring that up,
• 1:51:25
  Speaker 1: I'm going to repeat something, but it's really important.
• 1:51:28
  Speaker 2: 2003, it was founded.
• 1:51:29
  Speaker 1: And it was founded mostly because Amazon, as a growing
• 1:51:32
  Speaker 1: online store, needed hardcore infrastructure- 2006, I think, is when
• 1:51:36
  Speaker 1: they turned it client-facing. I may be wrong on the
• 1:51:38
  Speaker 1: dates there, but 2015 was the year it became profitable.
• 1:51:41
  Speaker 1: The total capital expenditures normalized for inflation were $ 29. 7 billion
• 1:51:46
  Speaker 1: across that 12-year period. And yeah, it lost money, but
• 1:51:51
  Speaker 1: if we speak cold economics here, Amazon didn't have to
• 1:51:55
  Speaker 1: go into the... They were unprofitable in a way, but
• 1:51:58
  Speaker 1: their margins actually started improving because AWS was a very
• 1:52:00
  Speaker 1: margin-heavy business.
• 1:52:01
  Speaker 2: It was great.
• 1:52:04
  Speaker 1: These two, Google... Cash flow negative. Amazon, cash flow negative.
• 1:52:07
  Speaker 1: These businesses, the reason you liked software businesses was they
• 1:52:11
  Speaker 1: are meant to be cash heavy asset light. These companies,
• 1:52:17
  Speaker 1: along with Meta, have added more than $ 700 billion of
• 1:52:20
  Speaker 1: new property plants and equipment. So assets, data centers, GPUs.
• 1:52:25
  Speaker 1: In the last four years, they have gone from being
• 1:52:27
  Speaker 1: these cash machines to to these cash furnaces.
• 1:52:31
  Speaker 2: You said a second ago, this can only continue if
• 1:52:34
  Speaker 2: investors continue to invest. Yes. And I was saying, I
• 1:52:37
  Speaker 2: think that investors are used to pumping money into things
• 1:52:40
  Speaker 2: that are burning cash. Your rebuttal to me sounds like, well,
• 1:52:43
  Speaker 2: this is burning more cash than ever. And then so
• 1:52:46
  Speaker 2: I would say, well, is the opportunity bigger than those
• 1:52:50
  Speaker 2: other case studies you referenced like AWS? And one would
• 1:52:53
  Speaker 2: say that the opportunity of intelligence... permeates everything. So the TAM,
• 1:53:00
  Speaker 2: the total addressable market is enormous. Maybe the rebuttal back
• 1:53:04
  Speaker 2: to me is about open source and all these kinds
• 1:53:06
  Speaker 2: of things.
• 1:53:06
  Speaker 3: No, no, no.
• 1:53:06
  Speaker 1: I actually know what you're getting at. So what you
• 1:53:08
  Speaker 1: were describing there is the argument that Satya Nadella or
• 1:53:11
  Speaker 1: Sam Moulton would make, but the theoretical opportunity of large
• 1:53:14
  Speaker 1: language models. And I could have bought that shit into
• 1:53:17
  Speaker 1: any 24 from them when they were like, oh, we
• 1:53:19
  Speaker 1: see the opportunity. We've gone way past the point at
• 1:53:22
  Speaker 1: which you can rationally argue that LLMs need this much money.
• 1:53:26
  Speaker 1: And when I say the money needs to keep flowing,
• 1:53:28
  Speaker 1: I am talking these two companies.
• 1:53:30
  Speaker 2: OpenAI, just OpenAI.
• 1:53:33
  Speaker 1: Glammy Sam Morton has said, Wall Street Journal and eCigarDZ
• 1:53:36
  Speaker 1: reported a few weeks ago, they plan to spend $ 750
• 1:53:41
  Speaker 1: billion on compute through 2030. I think they're going to
• 1:53:44
  Speaker 1: be dead before then, but $ 750 billion. That is an
• 1:53:48
  Speaker 1: insane amount of money.
• 1:53:50
  Speaker 2: It is crazy.
• 1:53:50
  Speaker 1: And... A large chunk of that is training. So when
• 1:53:53
  Speaker 1: I say progress, I mean, literally to make the models
• 1:53:56
  Speaker 1: better at stuff requires billions of dollars invested just in
• 1:53:58
  Speaker 1: data and also tens of billions of dollars of taking
• 1:54:02
  Speaker 1: that data. And so training, training is actually a really
• 1:54:05
  Speaker 1: interesting thing because when you think of like, well, for
• 1:54:09
  Speaker 1: Jay control my trainers, well, I train with them. I'll
• 1:54:11
  Speaker 1: live with them. I have a defined thing. And when
• 1:54:12
  Speaker 1: I do it and I eat right, muscles get bigger.
• 1:54:15
  Speaker 2: They would.
• 1:54:16
  Speaker 1: And here's the thing. When you train with an LLM,
• 1:54:18
  Speaker 1: you're experimenting. And this is not actually a hit on
• 1:54:21
  Speaker 1: the companies because they're still trying to work out how
• 1:54:24
  Speaker 1: to do the thing. Because putting aside how I feel
• 1:54:26
  Speaker 1: like they're trying to innovate, I think there are people
• 1:54:29
  Speaker 1: at these companies that actually want to do something interesting.
• 1:54:31
  Speaker 1: It's costing too much money. So once the money tap
• 1:54:33
  Speaker 1: turns off, the money won't be there to buy the
• 1:54:36
  Speaker 1: data or feed the data into the GPUs. Put aside
• 1:54:39
  Speaker 1: all the thoughts I have, just the raw capital to
• 1:54:41
  Speaker 1: get them this far has cost increasingly larger amounts of
• 1:54:45
  Speaker 1: money and increasingly larger amounts of training money for training
• 1:54:48
  Speaker 1: runs that sometimes can fail. GPT-5 was meant to be
• 1:54:52
  Speaker 1: this panacea for the AI industry. They had at least
• 1:54:55
  Speaker 1: one training run that cost half a billion dollars and
• 1:54:57
  Speaker 1: did nothing. And that's the thing. If we are thinking
• 1:55:01
  Speaker 1: about progress in a vacuum, they need so much more
• 1:55:04
  Speaker 1: money just to maybe get somewhere. There's no guarantee. There's
• 1:55:07
  Speaker 1: never any guarantee, but there's a reason that Google and
• 1:55:09
  Speaker 1: Amazon are cashflow negative now. There's a reason why Oracle's
• 1:55:12
  Speaker 1: probably going to die as a result of OpenAI because
• 1:55:15
  Speaker 1: Oracle's future depends on OpenAI spending $ 300 billion over five years.
• 1:55:19
  Speaker 2: It's absolutely fascinating because I was just reading through a
• 1:55:22
  Speaker 2: list of quotes from the big CEOs of the AI
• 1:55:24
  Speaker 2: companies to see what they would rebuttal you. And they're
• 1:55:28
  Speaker 2: all basically saying the same thing. They're all saying, this
• 1:55:31
  Speaker 2: is an actual, an exact quote from Sundar, who's the
• 1:55:33
  Speaker 2: CEO of Google. He says, the risk of under-investing is
• 1:55:37
  Speaker 2: dramatically greater than the risk of over-investing. And you go down,
• 1:55:43
  Speaker 2: you go through this, you know, Andy Jassy, CEO of Amazon,
• 1:55:46
  Speaker 2: we're not investing approximately 200 billion in CapEx in 2026
• 1:55:49
  Speaker 2: on a hunch. We're not going to be conservative in
• 1:55:53
  Speaker 2: how we play this. We're investing to be the meaningful
• 1:55:56
  Speaker 2: leader and our future business operating income and free cash
• 1:55:59
  Speaker 2: flow will be much larger because of this investment. Then
• 1:56:03
  Speaker 2: Mark Zuckerberg, CEO of Meta, says... will continue to invest
• 1:56:06
  Speaker 2: aggressively in infrastructure to meet the demand. I'd rather risk
• 1:56:10
  Speaker 2: building capacity before it's needed than being late.
• 1:56:13
  Speaker 1: Makes me think of Shrek with Lord Fodquad. Some of
• 1:56:16
  Speaker 1: you may die, but that's a risk I'm willing to accept.
• 1:56:18
  Speaker 1: It's like, you know, I'm just going to spend all
• 1:56:20
  Speaker 1: this money. You can't fire me because Mark Zuckerberg can't
• 1:56:23
  Speaker 1: be fired due to the unique board situation he's got going.
• 1:56:26
  Speaker 1: So yeah, he's just going to piss the money away
• 1:56:27
  Speaker 1: and hope he's right. I don't know if the people
• 1:56:29
  Speaker 1: know it matter.
• 1:56:29
  Speaker 2: He's not right. Why might you be wrong? I mean,
• 1:56:33
  Speaker 2: this is the thing.
• 1:56:34
  Speaker 1: The AI people who claim this is going to be
• 1:56:36
  Speaker 1: the biggest, strongest thing in the world, did they ever
• 1:56:37
  Speaker 1: get that?
• 1:56:38
  Speaker 2: I mean this... It's a good question.
• 1:56:40
  Speaker 1: Because it's like, they don't. And the thing is, what
• 1:56:42
  Speaker 1: would it take for me to be wrong? A bunch
• 1:56:44
  Speaker 1: of hardware breakthroughs to make this profitable.
• 1:56:45
  Speaker 2: It's a good question.
• 1:56:47
  Speaker 1: Because the thing is, When it comes to being a
• 1:56:49
  Speaker 1: critic or a skeptic, you are put on the hot seat.
• 1:56:52
  Speaker 1: Not the people spending a trillion dollars. Not the people
• 1:56:55
  Speaker 1: promising the world. The asshole with the blog is the
• 1:56:59
  Speaker 1: one who's like, no.
• 1:56:59
  Speaker 2: No, no.
• 1:57:00
  Speaker 3: Trust me.
• 1:57:00
  Speaker 2: If they came here, they'd be on the hot seat too.
• 1:57:02
  Speaker 1: Trust me. Oh, they won't talk to me.
• 1:57:05
  Speaker 2: I don't know why, Steve.
• 1:57:06
  Speaker 1: It's because I call him Clammy Sammy.
• 1:57:08
  Speaker 2: I think it's because my guests are quite critical of that.
• 1:57:11
  Speaker 2: I don't think Sam Altman wants to come here. Mr. Altman?
• 1:57:14
  Speaker 2: Go on, Steve. Do it.
• 1:57:15
  Speaker 1: But this is the thing. Of course, they're going to
• 1:57:17
  Speaker 1: say that. And also- If they thought they were right,
• 1:57:20
  Speaker 1: I don't think they do anymore. If I was in
• 1:57:22
  Speaker 1: their shoes and I thought that this was an existential thing, sure.
• 1:57:24
  Speaker 1: But it gets back to the rock-con bubble, which is, yeah,
• 1:57:27
  Speaker 1: this is the last thing they've got.
• 1:57:28
  Speaker 2: But I really want to know that question. It was
• 1:57:30
  Speaker 2: one of the questions I was really excited to ask you,
• 1:57:31
  Speaker 2: which is, you have a different opinion. We said this
• 1:57:34
  Speaker 2: at the top. You have a very different opinion from
• 1:57:36
  Speaker 2: a lot of people. I would categorize the two most
• 1:57:38
  Speaker 2: popular opinions as AI is going to hurt everybody and
• 1:57:42
  Speaker 2: it's going to be catastrophic and we need to stop.
• 1:57:44
  Speaker 2: The other opinion is age of abundance is going to
• 1:57:47
  Speaker 2: be amazing. Let us crack on. yours is different from
• 1:57:50
  Speaker 2: both of those which is as you said in your
• 1:57:52
  Speaker 2: words it's a con and it's in there's no real
• 1:57:55
  Speaker 2: underlying value in the technology and it's overhyped yes and
• 1:57:58
  Speaker 2: there's way too much spending i mean a few people
• 1:58:00
  Speaker 2: agree on the spending part yeah but the other part
• 1:58:02
  Speaker 2: so with you it's one of probably the first person
• 1:58:05
  Speaker 2: that i've spoken to that's had this opinion so how
• 1:58:08
  Speaker 2: what would it take for you to change your mind
• 1:58:13
  Speaker 2: about what you believe here?
• 1:58:14
  Speaker 1: There would need to be a hardware breakthrough that Reduce
• 1:58:17
  Speaker 1: the cost by like a thousand. It would have to
• 1:58:19
  Speaker 1: be just a dramatic breakthrough that is not happening, just
• 1:58:22
  Speaker 1: to be clear, because they've all been trying.
• 1:58:24
  Speaker 2: So it's the cost for you that would have to change.
• 1:58:26
  Speaker 1: It's the cost and it's also the data centers. I
• 1:58:28
  Speaker 1: think the way they're building the data centers is reckless
• 1:58:30
  Speaker 1: and damaging to communities. The fact that you have communities
• 1:58:33
  Speaker 1: like in Vineland, New Jersey, where the residents are like,
• 1:58:35
  Speaker 1: I don't want this, but the planning boards vote for
• 1:58:37
  Speaker 1: it because they're all, I assume, having chummy lunches with
• 1:58:40
  Speaker 1: the people doing it. I think the use of gas
• 1:58:43
  Speaker 1: turbines is fucking disgraceful. The water situation, I'm not super
• 1:58:47
  Speaker 1: well read on, so I'm not going to wade into it.
• 1:58:48
  Speaker 1: But the use of gas turbines and behind the beat
• 1:58:50
  Speaker 1: of power is reckless and damaging to communities. The noise
• 1:58:53
  Speaker 1: that these things make. And also, generative AI is this egregious,
• 1:58:58
  Speaker 1: pornographic demonstration of how unfair the world is. Regular people
• 1:59:03
  Speaker 1: try and get a loan for a business, a random
• 1:59:05
  Speaker 1: business they want.
• 1:59:06
  Speaker 2: I have a good idea.
• 1:59:07
  Speaker 1: They go to a bank, a bank of talent, they'll
• 1:59:09
  Speaker 1: go fuck themselves. They'll say, I'm not going to give you,
• 1:59:11
  Speaker 1: you're going to make a store that sells stuff?
• 1:59:12
  Speaker 2: Screw you.
• 1:59:13
  Speaker 1: You want to build a data center? Jensen Huang will
• 1:59:17
  Speaker 1: back you. Jensen Huang will give you 25% residual value.
• 1:59:20
  Speaker 1: You want to build a regular business that's even profitable?
• 1:59:23
  Speaker 2: Fuck you.
• 1:59:23
  Speaker 1: No, a venture capitalist won't give you the money. Something
• 1:59:26
  Speaker 1: that's just growing steadily but it's profitable? Screw that. No,
• 1:59:29
  Speaker 1: I need 10, 100x return. Try and get a mortgage.
• 1:59:32
  Speaker 1: You have to give the bank a full colonic. But
• 1:59:35
  Speaker 1: you want to get money for Jensen Huang to buy
• 1:59:37
  Speaker 1: some GPUs? He'll give you a contract. Coreweaves, great example.
• 1:59:40
  Speaker 1: A neocloud. which is just a company that builds data
• 1:59:43
  Speaker 1: centers and puts GPUs and rent them to people. NVIDIA,
• 1:59:46
  Speaker 1: one of their first investors, in 2023, signed a $ 1. 3
• 1:59:50
  Speaker 1: billion contract to rent back their GPUs from CoreWeave so
• 1:59:55
  Speaker 1: that CoreWeave go to a bank and go, I've got
• 1:59:57
  Speaker 1: a customer. Yeah, it's the guy I'm buying the GPUs
• 2:00:00
  Speaker 1: from with the debt I'm getting from you. If you
• 2:00:02
  Speaker 1: want to buy GPUs, it's open season. If you want
• 2:00:04
  Speaker 1: to live a regular life where you build a regular
• 2:00:06
  Speaker 1: business or buy a house... Highest interest rates ever.
• 2:00:10
  Speaker 2: Screw you. Up yours.
• 2:00:11
  Speaker 1: Yeah, you need to show us way more than that.
• 2:00:13
  Speaker 1: I don't trust you regular folks. But if you're an
• 2:00:16
  Speaker 1: unprofitable neocloud, you get billions from Jensen. It doesn't matter.
• 2:00:21
  Speaker 2: It's so interesting. It's interesting because you are the first
• 2:00:24
  Speaker 2: person that I've spoken to that has that opinion. I
• 2:00:26
  Speaker 2: am pro-user.
• 2:00:27
  Speaker 1: Let's take another myth. AI will be conscious. So, superintelligence,
• 2:00:35
  Speaker 1: artificial general intelligence, these are theories. Anyone saying this stuff
• 2:00:40
  Speaker 1: will become this is just guessing and does not have proof.
• 2:00:44
  Speaker 2: Okay.
• 2:00:45
  Speaker 1: And that's really it.
• 2:00:46
  Speaker 2: Okay. Okay, let's take another myth.
• 2:00:50
  Speaker 1: AI systems are already blackmailing and escaping control. So this
• 2:00:54
  Speaker 1: is a really specific one. Anthropic.
• 2:00:56
  Speaker 2: There's actually two.
• 2:00:58
  Speaker 1: OpenAI's GPT 3.5. I realize this is more than a sentence.
• 2:01:01
  Speaker 1: I apologize. In their system card, and a bunch of
• 2:01:05
  Speaker 1: media outlets covered this, saying that OpenAI's model blackmailed a
• 2:01:09
  Speaker 1: TaskRabbit into solving a capture. What actually happened was a
• 2:01:15
  Speaker 1: user of GPT doing the experiment got it to generate
• 2:01:20
  Speaker 1: things to say to a TaskRabbit to make a TaskRabbit
• 2:01:23
  Speaker 1: do stuff.
• 2:01:24
  Speaker 2: A TaskRabbit being?
• 2:01:24
  Speaker 1: As in a person that you rent. Not even to
• 2:01:27
  Speaker 1: do a capture. It's something you rent to nail a
• 2:01:29
  Speaker 1: picture up in your apartment. It's an insane example. This
• 2:01:32
  Speaker 1: was covered as if these things blackmailed someone and And
• 2:01:36
  Speaker 1: they specifically said, yeah, we prompted it to do this.
• 2:01:38
  Speaker 1: And also the other note was that, yeah, AI systems
• 2:01:41
  Speaker 1: can't do autonomous.
• 2:01:42
  Speaker 2: Stuff like this.
• 2:01:43
  Speaker 1: Then there was this other one where Anthropic said, oh, yeah,
• 2:01:46
  Speaker 1: a model was blackmailing someone saying that if you don't
• 2:01:49
  Speaker 1: do this, I'll email proof that you slept with someone
• 2:01:52
  Speaker 1: else other.
• 2:01:52
  Speaker 2: Than your wife.
• 2:01:53
  Speaker 1: I think it was. What actually happened was Anthropic explicitly
• 2:01:56
  Speaker 1: trained a model to do this and then prompted it
• 2:01:59
  Speaker 1: to blackmail. This keeps happening and the media just slop
• 2:02:04
  Speaker 1: me up. No thoughts. Put the story in the bag.
• 2:02:07
  Speaker 1: And it's frustrating because it scares people. Put aside the
• 2:02:11
  Speaker 1: fact it's wrong. It's scary.
• 2:02:13
  Speaker 2: It's scary to people.
• 2:02:14
  Speaker 1: People living their lives who have to work longer hours
• 2:02:17
  Speaker 1: to make less money and their money doesn't go far.
• 2:02:19
  Speaker 1: And they turn on the fucking news and there's some
• 2:02:21
  Speaker 1: asshole being like, Yeah, you should be terrified. It blackmailed someone.
• 2:02:25
  Speaker 2: But this is so counterintuitive of their interest to some degree.
• 2:02:29
  Speaker 2: And they've experienced it backfire.
• 2:02:31
  Speaker 1: Well, they have now.
• 2:02:32
  Speaker 2: Like, it's literally backfired. It's backfired. Eric Schmidt getting booed
• 2:02:36
  Speaker 2: at the commencement speech by 8,000 people every time he
• 2:02:38
  Speaker 2: said the word AI. Yeah, that's awesome. But I mean,
• 2:02:40
  Speaker 2: this is, this is, I mean, these series are being
• 2:02:43
  Speaker 2: attacked at home.
• 2:02:44
  Speaker 1: Yeah, which fucking sucks.
• 2:02:45
  Speaker 2: Yeah, which is terrible.
• 2:02:46
  Speaker 1: I must be clear, like, he disliked the company, don't
• 2:02:49
  Speaker 1: fucking hurt people.
• 2:02:49
  Speaker 2: Yeah, don't attack people at home. But the point here
• 2:02:52
  Speaker 2: is that That narrative is backfiring in a big, big
• 2:02:55
  Speaker 2: way for them.
• 2:02:56
  Speaker 1: I agree. I don't think they saw it coming because
• 2:02:58
  Speaker 1: you have to remember, you mentioned regulation earlier. These tech
• 2:03:01
  Speaker 1: companies have been glazed for their entire existence. Travis Kalanick's like, oh, what?
• 2:03:06
  Speaker 1: People don't like me now. And it's because Uber was
• 2:03:08
  Speaker 1: a horribly run place and he was kind of a monster. Also,
• 2:03:11
  Speaker 1: tons of articles about how great Uber was at the time.
• 2:03:13
  Speaker 1: The point I'm making is these companies are not used
• 2:03:15
  Speaker 1: to pushback. They thought what would happen, I believe, just guessing,
• 2:03:19
  Speaker 1: they thought they'd do this scary stuff and they would
• 2:03:21
  Speaker 1: just get floods of money. And everyone would just be like,
• 2:03:23
  Speaker 1: I kneel before you. I'll do whatever you want. They
• 2:03:26
  Speaker 1: didn't expect, I think, what has... I agree. This has
• 2:03:30
  Speaker 1: backfired on them. Because they were inarticulate. They're disconnected from
• 2:03:34
  Speaker 1: regular people. Sam Altman drives a $ 5 million car around
• 2:03:38
  Speaker 1: San Francisco. So that man's doing it like nine miles
• 2:03:40
  Speaker 1: an hour. It's hilarious. But these people are disconnected from
• 2:03:43
  Speaker 1: everyone else. So they don't experience real problems. So they
• 2:03:46
  Speaker 1: can't build the solutions for them. And they think, well,
• 2:03:49
  Speaker 1: if we scare people into doing what we want, that'll work, right?
• 2:03:52
  Speaker 2: It didn't.
• 2:03:53
  Speaker 1: This was all of this blackmail stuff was an attempt
• 2:03:55
  Speaker 1: to make it mystic. It was a mysticism attempt. It
• 2:03:58
  Speaker 1: was to make it seem like this unknowable, impossible to control,
• 2:04:01
  Speaker 1: just this powerful thing. But we're the only ones. We
• 2:04:04
  Speaker 1: are the only us. Only these two angels could possibly
• 2:04:08
  Speaker 1: control the beast we've created.
• 2:04:10
  Speaker 2: This is quite a controversial statement, but I think that
• 2:04:14
  Speaker 2: for some reason, I trust Dario a little bit more
• 2:04:17
  Speaker 2: because I think he's been the most balanced in his
• 2:04:19
  Speaker 2: writing about the risk profile. Whereas the others, they seem
• 2:04:25
  Speaker 2: to kind of move with the wind.
• 2:04:27
  Speaker 1: I get what you mean. The reason I don't like
• 2:04:30
  Speaker 1: Dario is, Dario was doing the scare tactics thing when
• 2:04:33
  Speaker 1: he worked at OpenAI, when GPT-2 came out.
• 2:04:35
  Speaker 2: And said, it's too scary to release.
• 2:04:37
  Speaker 1: He's also gone on television and given AI psychosis to Axios,
• 2:04:42
  Speaker 1: being like, 50% of jobs are going to go away
• 2:04:44
  Speaker 1: because of AI.
• 2:04:45
  Speaker 2: What I respect is the consistency. He's now being attacked
• 2:04:49
  Speaker 2: by them. Good point. Um, but the thing is, sorry,
• 2:04:52
  Speaker 2: I mean, clarify the word attack. Dario is being verbally
• 2:04:56
  Speaker 2: attacked by Silicon Valley. And you know, if Silicon Valley,
• 2:05:01
  Speaker 2: if powerful people in Silicon Valley are attacking someone.
• 2:05:03
  Speaker 1: Four months ago, he wasn't though. They were all saying
• 2:05:06
  Speaker 1: he was the smartest boy ever.
• 2:05:07
  Speaker 2: But the point I want to.
• 2:05:07
  Speaker 1: Make there as well is again, wow, you're so scared
• 2:05:09
  Speaker 1: of how powerful this is. You're so scared of it.
• 2:05:11
  Speaker 2: It's so scary.
• 2:05:12
  Speaker 1: What are you doing about it?
• 2:05:13
  Speaker 2: Oh, nothing.
• 2:05:14
  Speaker 1: Like, it's just like, what are you doing? Well, we
• 2:05:16
  Speaker 1: have an alignment team. So does every AI lab. Well,
• 2:05:18
  Speaker 1: I guess open AI cycles through those really quickly. Here's
• 2:05:21
  Speaker 1: the thing. If I'm Dario Amadeus and I'm like, I'm.
• 2:05:24
  Speaker 2: Scared of all things changing.
• 2:05:26
  Speaker 1: And I thought I had made a thing that would
• 2:05:28
  Speaker 1: eliminate all jobs. I'd be fucking terrified. I'd be walking
• 2:05:31
  Speaker 1: around with like, like a 10 ton weight on my back.
• 2:05:34
  Speaker 1: The shot, the responsibility, the fact he doesn't, the fact
• 2:05:37
  Speaker 1: he wants to be this weird elder statesman that's too
• 2:05:39
  Speaker 1: scared to hold Sam Altman's hand at an event just
• 2:05:42
  Speaker 1: makes me believe that he's just saying it because it's convenient.
• 2:05:45
  Speaker 1: And he'll wind that back as he kind of already
• 2:05:47
  Speaker 1: has whenever it's convenient for him. I think OpenAI and
• 2:05:50
  Speaker 1: Anthropic are basically the same level of bad company. I
• 2:05:53
  Speaker 1: think Anthropic is more cult-like. I think it's so weird.
• 2:05:57
  Speaker 1: Like Jack Clark over there, one of the co-founders, that
• 2:05:59
  Speaker 1: fella used to be at the register. He used to
• 2:06:01
  Speaker 1: be one of the most critical journalists ever. Now it's
• 2:06:03
  Speaker 1: like something took over him because they talk of these
• 2:06:06
  Speaker 1: things in these highfalutin terms. But then again, maybe the
• 2:06:09
  Speaker 1: people at Anthropic buy their shit. Maybe some of the
• 2:06:10
  Speaker 1: people at OpenAI buy their shit.
• 2:06:12
  Speaker 2: I don't know. So going back to the central question
• 2:06:13
  Speaker 2: we asked at the top here was what would have
• 2:06:15
  Speaker 2: to be the case for you to look back and say,
• 2:06:17
  Speaker 2: do you know what? I was wrong in 2026. And
• 2:06:19
  Speaker 2: you said to me it would be mainly that the
• 2:06:22
  Speaker 2: cost of production around AI drops dramatically.
• 2:06:26
  Speaker 1: And it would have to also do insane amounts of stuff.
• 2:06:29
  Speaker 1: It would have to be a truly autonomous.
• 2:06:32
  Speaker 2: It would have to continue its improvement in terms of capability.
• 2:06:34
  Speaker 1: It would have to be a different product. It would
• 2:06:37
  Speaker 1: have to be indistinguishable from magic. And the reason I
• 2:06:39
  Speaker 1: have these high standards is they set them.
• 2:06:41
  Speaker 2: Okay, fair enough.
• 2:06:41
  Speaker 1: It's interesting as well because all these myths and all
• 2:06:44
  Speaker 1: these conversations, it's about technology, but it's also, it's an
• 2:06:47
  Speaker 1: information war. It's literally- narrative versus narrative, everyone trying to
• 2:06:53
  Speaker 1: escape the financials, everyone trying to actually escape what the
• 2:06:56
  Speaker 1: models can do. And the big thing I always say
• 2:06:58
  Speaker 1: about AI boosters is if I could regulate them, I'd
• 2:07:01
  Speaker 1: regulate them. They can't speak in the future tense anymore. Just,
• 2:07:03
  Speaker 1: you got to talk about today, mate, you get two
• 2:07:06
  Speaker 1: weeks in the future max, because if they were constrained
• 2:07:09
  Speaker 1: to what was happening today, they would sound like insane
• 2:07:12
  Speaker 1: people yeah.
• 2:07:13
  Speaker 2: No i think yeah most i guess most technology companies
• 2:07:15
  Speaker 2: would at the time like uber would sound insane but
• 2:07:17
  Speaker 2: no uber was basically the difference.
• 2:07:20
  Speaker 1: They were pissing money though weren't they they were pissing
• 2:07:21
  Speaker 1: money away but the unit economics were the same just
• 2:07:24
  Speaker 1: subsidized so you were still getting a service from a
• 2:07:27
  Speaker 1: to b and paying a much lower cost it wasn't
• 2:07:30
  Speaker 1: like you paid uber 200 sorry 20 bucks a month
• 2:07:34
  Speaker 1: and you could get 500 miles of uber And then
• 2:07:36
  Speaker 1: one day you started paying by the mile because that's
• 2:07:38
  Speaker 1: what's happening with this.
• 2:07:39
  Speaker 2: They've changed their business model for customers like me now
• 2:07:43
  Speaker 2: so that I have to buy credits.
• 2:07:45
  Speaker 1: No. Well, kind of with Fable.
• 2:07:47
  Speaker 2: They asked me the other day.
• 2:07:48
  Speaker 1: So with the Anthropix Fable model, with some accounts you
• 2:07:52
  Speaker 1: have to pay for usage. And also adoption of Fable
• 2:07:55
  Speaker 1: has been pretty low because of this, because of the cost.
• 2:07:57
  Speaker 1: But with enterprises, so companies over 150. people you have
• 2:08:01
  Speaker 1: to pay by the token now a per million token
• 2:08:03
  Speaker 1: oh so they are moving to a token yeah but
• 2:08:05
  Speaker 1: when they did that everyone went from being like this
• 2:08:07
  Speaker 1: is the most impressive thing ever to.
• 2:08:08
  Speaker 2: Being like.
• 2:08:10
  Speaker 1: We've got to control these costs uber's coo says andrew
• 2:08:14
  Speaker 1: mcdonald i think he said that it's getting hard to
• 2:08:16
  Speaker 1: justify the Because it's hard to connect spending money on
• 2:08:19
  Speaker 1: tokens to actual useful outcomes.
• 2:08:22
  Speaker 2: He said the thing. Like he said the actual thing
• 2:08:24
  Speaker 2: I'd be saying. And it's... So we're in an AI bubble.
• 2:08:27
  Speaker 1: Yes.
• 2:08:27
  Speaker 2: And when this AI bubble collapses, so much of the
• 2:08:31
  Speaker 2: economy is resting upon it. Yeah. It's going to have
• 2:08:34
  Speaker 2: downstream consequences. So I got two questions for you. I
• 2:08:37
  Speaker 2: guess the first question is, are we in an AI bubble?
• 2:08:39
  Speaker 2: And what happens when the bubble pops? Yes.
• 2:08:41
  Speaker 1: And it depends. So... The big thing that people say is, oh,
• 2:08:46
  Speaker 1: we'll get bailed out. Donald Trump. Scared of Donald Trump.
• 2:08:49
  Speaker 1: Here's the problem with this. It isn't just an AI bubble.
• 2:08:53
  Speaker 1: It's the Rotcom bubble. So the AI bubble collapsing will
• 2:08:57
  Speaker 1: probably be this company running out of money. OpenAI. And
• 2:09:01
  Speaker 1: the thing is with OpenAI is they were meant to
• 2:09:03
  Speaker 1: go public this year. And now it's been pushed to
• 2:09:05
  Speaker 1: next year. A week and a half after I released
• 2:09:07
  Speaker 1: their auditive financials.
• 2:09:08
  Speaker 2: I wonder where that was.
• 2:09:10
  Speaker 1: But they've delayed to next year. Sarah Fry, the CFO,
• 2:09:12
  Speaker 1: has now said, well, they'll do it. earlier than 2027
• 2:09:16
  Speaker 1: or 2027 great answer there.
• 2:09:18
  Speaker 2: For anyone that doesn't understand what going public means that
• 2:09:20
  Speaker 2: means joining the stock market and at such a time
• 2:09:22
  Speaker 2: when you join the stock market your investors can finally
• 2:09:25
  Speaker 2: sell their equity that they got for investing in the
• 2:09:29
  Speaker 2: company when it was private so oftentimes companies will flirt
• 2:09:33
  Speaker 2: with the idea of we'll go public someday soon because
• 2:09:36
  Speaker 2: investors will have a moment in their head where they'll
• 2:09:39
  Speaker 2: get their money back at a return so You kind
• 2:09:42
  Speaker 2: of need to, if you're in these guys' shoes, you
• 2:09:44
  Speaker 2: kind of need to be flirting with going public or
• 2:09:45
  Speaker 2: investors won't want to invest.
• 2:09:47
  Speaker 1: OpenAI, up until this point, has been a private company.
• 2:09:50
  Speaker 1: And their last funding round, they were valued at $ 865 billion. Now,
• 2:09:55
  Speaker 1: when they tried to go public, New York Times' Mike
• 2:09:58
  Speaker 1: Isaac reported this. They tried to list, well, they wanted
• 2:10:02
  Speaker 1: to go at a $ 1 trillion valuation. Apparently, their advisor said, no,
• 2:10:07
  Speaker 1: don't do that. That is very bad for a number
• 2:10:10
  Speaker 1: of reasons. One, OpenAI needs perpetual amounts of money. They
• 2:10:13
  Speaker 1: raised $ 122 billion this year. Most of it's crossed. There's
• 2:10:16
  Speaker 1: some left, but they are going to need to raise
• 2:10:19
  Speaker 1: at least $ 100 billion a year just to survive. If
• 2:10:22
  Speaker 1: they can't go public, they will have to raise another
• 2:10:24
  Speaker 1: funding round. The problem is it's going to be difficult
• 2:10:26
  Speaker 1: to raise even the same one they raised that. So
• 2:10:29
  Speaker 1: they're probably going to have to take a flat, so
• 2:10:30
  Speaker 1: the same amount.
• 2:10:31
  Speaker 2: They can't do that. Exactly.
• 2:10:33
  Speaker 1: But they need money. They need money so bad. Amazon
• 2:10:36
  Speaker 1: sent them $ 35 billion that was meant to be contingent
• 2:10:38
  Speaker 1: on them going public early. They did that because they
• 2:10:42
  Speaker 1: need the money. Now, OpenAI is the kind of catastrophe
• 2:10:46
  Speaker 1: center here because Anthropic is likely going to beat it
• 2:10:49
  Speaker 1: to go public. And once Anthropic goes public, it'll be
• 2:10:52
  Speaker 1: borderline impossible for OpenAI to do so because Anthropic, an unprofitable,
• 2:10:55
  Speaker 1: unsustainable AI lab, but a better business that's growing faster
• 2:10:58
  Speaker 1: than OpenAI's. I believe they have a ceiling they're eventually
• 2:11:01
  Speaker 1: going to face perdition to. I think sometime in 2027,
• 2:11:05
  Speaker 1: things are going to start running out of steam because
• 2:11:07
  Speaker 1: the thing I said earlier, the only way these models
• 2:11:08
  Speaker 1: get better is if you feed more money, tens of
• 2:11:11
  Speaker 1: billions of dollars into them.
• 2:11:12
  Speaker 2: So you think OpenAI runs out of steam in 2027?
• 2:11:15
  Speaker 1: I think they're already running out of steam, yeah. But
• 2:11:17
  Speaker 1: I think they run out of cash. You think they
• 2:11:18
  Speaker 1: run out of cash?
• 2:11:19
  Speaker 2: Yes. And the sequence of events here will be they
• 2:11:22
  Speaker 2: go out and try and raise.
• 2:11:23
  Speaker 1: And they have trouble raising another round. I think maybe
• 2:11:26
  Speaker 1: NVIDIA props them up a little, maybe private credit. Blackstone, BlackRock,
• 2:11:30
  Speaker 1: and the like, the ones. And the reason that private
• 2:11:32
  Speaker 1: credit is getting involved, so asset managers, is because they're
• 2:11:34
  Speaker 1: investing in the data centers. And they know this company's
• 2:11:37
  Speaker 1: most of the data center demand.
• 2:11:38
  Speaker 2: Okay, so they run out of steam in 2027, according
• 2:11:40
  Speaker 2: to you.
• 2:11:41
  Speaker 1: Yep. And maybe they try, if they bum rush to
• 2:11:43
  Speaker 1: go public, they're going to have worse economics than Anthropic.
• 2:11:45
  Speaker 1: They're going to get savage. WeWork was a great example.
• 2:11:48
  Speaker 1: Another soft bank classic. Now, I think open AI collapses.
• 2:11:52
  Speaker 1: There are many different ways it could happen. There are
• 2:11:54
  Speaker 1: many different ways it could end. But the crucial thing is,
• 2:11:57
  Speaker 1: is that there are multiple companies that are existentially tied
• 2:12:01
  Speaker 1: to OpenAI. SoftBank, one of the largest companies in the
• 2:12:04
  Speaker 1: Japanese stock market, a holding company with lots of investments.
• 2:12:07
  Speaker 1: They have on paper about $ 100 billion worth of OpenAI stock.
• 2:12:12
  Speaker 1: If they can't go public, they can't do diddly squat
• 2:12:14
  Speaker 1: with them. And so SoftBank's future, their ability to continue
• 2:12:18
  Speaker 1: paying the people around them and existing as a business,
• 2:12:21
  Speaker 1: relies on their ability to continually liquidate funds, to take
• 2:12:25
  Speaker 1: the things they've invested in and have value from them,
• 2:12:28
  Speaker 1: either by selling the stock or taking loans out on
• 2:12:30
  Speaker 1: the stock. If OpenAI can't go public, SoftBank can't do that.
• 2:12:33
  Speaker 1: SoftBank probably won't run out of money, but we're going
• 2:12:36
  Speaker 1: to see one of the largest holding companies in the
• 2:12:37
  Speaker 1: world become much smaller. We will also see Amazon, Google,
• 2:12:42
  Speaker 1: and Microsoft have to restate guidance. They will have to say, actually,
• 2:12:46
  Speaker 1: we don't think we're going to grow as fast.
• 2:12:48
  Speaker 2: And what happens then?
• 2:12:49
  Speaker 1: Well, I think we enter a tech depression because the
• 2:12:51
  Speaker 1: rock-com bubble, the core of my theory, is that they're
• 2:12:55
  Speaker 1: out of hyper-growth ideas, but the market doesn't think so.
• 2:12:58
  Speaker 1: The reason they're so maniacally spending is because buying AI
• 2:13:03
  Speaker 1: GPUs allows them to kick the can further. It allows
• 2:13:05
  Speaker 1: them to say, we're still doing something. We're working on AI.
• 2:13:08
  Speaker 1: Don't think too hard.
• 2:13:09
  Speaker 2: And also their current businesses are still growing.
• 2:13:11
  Speaker 1: Their current businesses will eventually slow. There's only so many
• 2:13:14
  Speaker 1: price increases. There's only so many tweaks to ads. Only
• 2:13:17
  Speaker 1: so many tweaks to Google search. Only so many ways
• 2:13:21
  Speaker 1: that Amazon can screw merchants.
• 2:13:23
  Speaker 2: So in that tech depression, which you think might be
• 2:13:27
  Speaker 2: triggered in 2027, is that a cascading downstream economic depression?
• 2:13:32
  Speaker 2: Because the stock market is heavily dependent on these companies.
• 2:13:35
  Speaker 2: The stock market sees a pullback. Investors stop investing. They
• 2:13:38
  Speaker 2: get panicked.
• 2:13:40
  Speaker 1: Yes, I think that because.
• 2:13:41
  Speaker 2: What sort of downstream consequence, the sort of domino effect?
• 2:13:44
  Speaker 1: There's so much to imagine that it's difficult to capture everything, but...
• 2:13:48
  Speaker 1: There are a few things that worry me. First of all,
• 2:13:50
  Speaker 1: a ton of American money, just regular people's money, retail
• 2:13:53
  Speaker 1: investors are in these companies. And they bought into the
• 2:13:55
  Speaker 1: Magnificent Seven thinking the number go up forever. NVIDIA is
• 2:13:58
  Speaker 1: the largest company on the Fortune 500 and NASDAQ as well.
• 2:14:02
  Speaker 1: And like 7% to 8% of the S & P 500.
• 2:14:05
  Speaker 1: That company, when the bottom falls out from NVIDIA, and
• 2:14:08
  Speaker 1: we haven't really got into it, but NVIDIA is doing
• 2:14:10
  Speaker 1: the most circular of financing, feeding companies money so that
• 2:14:13
  Speaker 1: they can raise debt to buy more GPUs. I think
• 2:14:16
  Speaker 1: Nvidia's revenue could go 50% to 70% down. I think
• 2:14:19
  Speaker 1: that Nvidia could put... Nvidia back in 2022 was making
• 2:14:22
  Speaker 1: single-digit billion dollars.
• 2:14:23
  Speaker 2: And what happens though? I'm thinking about like Jenny and
• 2:14:25
  Speaker 2: Dave that are watching this right now. And they are
• 2:14:27
  Speaker 2: just normal people with normal jobs.
• 2:14:30
  Speaker 1: People's retirements are going to contract severely. And I don't
• 2:14:34
  Speaker 1: believe they're going to return to those values. And I
• 2:14:36
  Speaker 1: think that... because so much of the value of the
• 2:14:38
  Speaker 1: S & P 500 and Russell 1000 index comes from
• 2:14:41
  Speaker 1: these four companies and the rest of the Magnificent Seven.
• 2:14:43
  Speaker 1: So Apple, Tesla matter as well. And the thing is,
• 2:14:48
  Speaker 1: I don't know what happens after that because venture capital
• 2:14:52
  Speaker 1: has also, more than half of venture capital last year
• 2:14:54
  Speaker 1: went into AI. I think most venture capital investments in
• 2:14:57
  Speaker 1: AI are going to zero because when it comes to
• 2:14:59
  Speaker 1: building a company on top of an LLM, all of
• 2:15:01
  Speaker 1: those are unprofitable too. And the thing is, LLM companies
• 2:15:05
  Speaker 1: have not really been acquired. The exception being Cursable by
• 2:15:08
  Speaker 1: Elon Musk for the coding side. But you have Cognition,
• 2:15:12
  Speaker 1: which is just another LLM company raising a $ 26 billion valuation.
• 2:15:16
  Speaker 1: That means that company has to go public because who's
• 2:15:18
  Speaker 1: buying a company at $ 26 billion.
• 2:15:20
  Speaker 2: Other than Elon Musk?
• 2:15:21
  Speaker 1: And there are rumors that Elon Musk was trying to
• 2:15:23
  Speaker 1: buy them as well. Is Elon Musk just going to
• 2:15:24
  Speaker 1: pick off every LLM company? Like going to fucking TJ
• 2:15:28
  Speaker 1: Maxx for AI?
• 2:15:28
  Speaker 2: Like Jesus Christ. So is that a recession you're describing?
• 2:15:31
  Speaker 1: It is a recession, but it's also a depression within
• 2:15:34
  Speaker 1: people's retirements like i'm talking about 20 30 40 percent
• 2:15:38
  Speaker 1: off the top of these companies stock.
• 2:15:39
  Speaker 2: Value economic contractions recessions consistently lead to job losses and
• 2:15:43
  Speaker 2: rising unemployment when an economy contracts the mechanism driving job
• 2:15:46
  Speaker 2: losses typically follows a predictable sequence falling demand consumers and
• 2:15:50
  Speaker 2: businesses spend less money causing revenues across most industries to
• 2:15:53
  Speaker 2: drop margin compression with lower revenue and often fixed overhead
• 2:15:57
  Speaker 2: costs like rent or debt, corporate profits shrink, and lastly,
• 2:16:00
  Speaker 2: cost-cutting measures to survive or protect profit margins, businesses freeze hiring,
• 2:16:04
  Speaker 2: reduce hours, and resort to layoffs.
• 2:16:06
  Speaker 1: Yes, that would all happen. But the thing is, we're
• 2:16:09
  Speaker 1: talking about equity values dropping, and we're talking about there
• 2:16:12
  Speaker 1: not really being a home for that value or that money.
• 2:16:16
  Speaker 1: So much is riding on these companies, but you can't
• 2:16:19
  Speaker 1: bail it out. You can theoretically bail out OpenAI. I
• 2:16:22
  Speaker 1: don't think it happens. You could pump these dogs full
• 2:16:25
  Speaker 1: of money and keep them alive for a bit but
• 2:16:27
  Speaker 1: at some point they're going to have to start they
• 2:16:28
  Speaker 1: have between these two companies anthropic and open ai you
• 2:16:32
  Speaker 1: have 1.1 trillion dollars of commitments just open ai oracle
• 2:16:37
  Speaker 1: is building 7.1 gigawatts of data centers so over 400
• 2:16:40
  Speaker 1: billion dollars worth just for open ai there is not
• 2:16:43
  Speaker 1: a customer on earth And Oracle's revenue has been flat
• 2:16:46
  Speaker 1: the last 15 years when you adjust for inflation. Without OpenAI,
• 2:16:49
  Speaker 1: Oracle dies.
• 2:16:50
  Speaker 2: So you think OpenAI is going to crash and run
• 2:16:51
  Speaker 2: out of money, and that's going to cause this domino
• 2:16:53
  Speaker 2: effect across these other big tech companies, which is going
• 2:16:56
  Speaker 2: to impact the stock market and impact the broader economy. Yes.
• 2:16:59
  Speaker 1: And also the tens of thousands of people that will
• 2:17:01
  Speaker 1: be laid off from the tech sector. But also the
• 2:17:03
  Speaker 1: venture capital thing is significant because venture capital has been
• 2:17:07
  Speaker 1: having one of the most historic bad runs in history.
• 2:17:12
  Speaker 1: Since 2018- The average return from venture capital, total value
• 2:17:17
  Speaker 1: put in, so the amount of money you get back
• 2:17:18
  Speaker 1: for your dollar is between 0.8 and 1.21, meaning for
• 2:17:21
  Speaker 1: every dollar you invest, you get 80 cents to $ 1. 20.
• 2:17:24
  Speaker 2: Paper gains.
• 2:17:25
  Speaker 1: Well, no, that's just actual returns. Paper gains they'll give you,
• 2:17:28
  Speaker 1: but even an internal rate return, which is a whole
• 2:17:31
  Speaker 1: separate thing, even that's not very happy. But long story short,
• 2:17:34
  Speaker 1: very simple, venture capital is not making money come out.
• 2:17:37
  Speaker 1: Venture capital is not actually providing returns.
• 2:17:40
  Speaker 2: They're celebrating paper gains.
• 2:17:42
  Speaker 1: They're celebrating paper gains.
• 2:17:43
  Speaker 2: And they're raising off paper gains. And paper gains, I mean,
• 2:17:46
  Speaker 2: just being able to say, oh, look, the valuation of
• 2:17:48
  Speaker 2: Anthropic went up.
• 2:17:49
  Speaker 1: So that's, that's what google and amazon were doing google's
• 2:17:52
  Speaker 1: last quarter they boosted their net profits profits on paper
• 2:17:56
  Speaker 1: by 99 billion dollars because of the increased value of
• 2:17:59
  Speaker 1: their spacex holding and their anthropic holding and again the
• 2:18:04
  Speaker 1: fact that this is happening is insane and the fact
• 2:18:06
  Speaker 1: it's not a scandal is insane but we live in
• 2:18:08
  Speaker 1: this culture i guess but everyone is really benefiting right
• 2:18:13
  Speaker 1: now it's really that it's like great twitter it's like
• 2:18:15
  Speaker 1: when you're uh reaping it's like yeah fuck yeah this
• 2:18:17
  Speaker 1: rocks sewing ah shit this sucks because right now they're
• 2:18:20
  Speaker 1: all like yeah all the speculative gains are awesome the
• 2:18:22
  Speaker 1: paper gains are awesome the theoreticals of anthropic being worth
• 2:18:26
  Speaker 1: two trillion dollars wow the articles we can write the
• 2:18:28
  Speaker 1: promises we can make then when the rubber meets the
• 2:18:30
  Speaker 1: road it's gonna be pretty rough on them because my
• 2:18:34
  Speaker 1: The valuation of Amazon, Google, Microsoft, and Meta is based
• 2:18:38
  Speaker 1: on this idea that they will grow eternally, that they
• 2:18:39
  Speaker 1: will grow forever. If that changes, to quote Ed Elson
• 2:18:42
  Speaker 1: from Prof G Markets again, it's this, they're all doing
• 2:18:45
  Speaker 1: Botox right now. They're sinking money into it to make
• 2:18:47
  Speaker 1: themselves feel young again. And the market believes them. When
• 2:18:50
  Speaker 1: the market doesn't, we're not just talking about a depression.
• 2:18:52
  Speaker 1: I'm talking about the market valuing them like airlines and saying, yeah,
• 2:18:56
  Speaker 1: you're real big and you make money off your existing products.
• 2:18:59
  Speaker 1: But guess what? You don't have new shit. You're just
• 2:19:02
  Speaker 1: going to be doing this forever, and we're going to
• 2:19:04
  Speaker 1: value you as such.
• 2:19:05
  Speaker 2: So for Jenny and Dave, should they do anything differently?
• 2:19:09
  Speaker 2: Should they be conserving money? If there's a recession or
• 2:19:11
  Speaker 2: depression coming, should they be a little bit more conservative?
• 2:19:13
  Speaker 2: Should they- Yes.
• 2:19:14
  Speaker 1: I actually think it's- I don't know. I don't have
• 2:19:17
  Speaker 1: money in the market. I think it's a casino. Casino
• 2:19:20
  Speaker 1: pumped up by the media.
• 2:19:21
  Speaker 2: Should they invest in the S & P 500? Should
• 2:19:23
  Speaker 2: they invest in OpenAI?
• 2:19:24
  Speaker 1: Oh, God, no. I- Honestly, I live in cash right now.
• 2:19:27
  Speaker 2: You live in cash?
• 2:19:28
  Speaker 1: Yeah. I don't fucking trust the market, man. Try and
• 2:19:31
  Speaker 1: get some gains here. I'm not comfortable giving financial advice,
• 2:19:35
  Speaker 1: but it's like you're gambling.
• 2:19:37
  Speaker 2: Okay, be conservative. Things might get volatile. Yeah, it really is.
• 2:19:41
  Speaker 1: It's going to be act as you would with volatility.
• 2:19:43
  Speaker 1: Take the gains when you've got them. Don't sell everything,
• 2:19:46
  Speaker 1: but be suspicious of tech. That's actually the biggest thing.
• 2:19:49
  Speaker 1: It's like be suspicious of what they're promising. If you're
• 2:19:51
  Speaker 1: acting based on their promises, don't trust the promises. Trust
• 2:19:56
  Speaker 1: that they are going to say what will make the
• 2:19:58
  Speaker 1: stock run rather than what's actually happening. And that they
• 2:20:03
  Speaker 1: will find every dodgy way to make you think something
• 2:20:07
  Speaker 1: is happening rather than it's actually happening. Annualized run rate,
• 2:20:10
  Speaker 1: great example. Microsoft said that they had $ 38, $ 37 billion
• 2:20:14
  Speaker 1: of annualized run rate in AI.
• 2:20:16
  Speaker 2: You hear that, you go, I made $ 38, $ 37.
• 2:20:19
  Speaker 1: Billion, right? Wow, that's so much. Run rate may be
• 2:20:23
  Speaker 1: month times 12. They don't even define it. But it's
• 2:20:25
  Speaker 1: built to manipulate. And they do that because we don't
• 2:20:28
  Speaker 1: have a functional SEC and we don't have a media
• 2:20:30
  Speaker 1: environment that actually where skepticism is the priority and where
• 2:20:34
  Speaker 1: protecting the readers is necessary.
• 2:20:36
  Speaker 2: What would they say? They would say, Ed, this technology
• 2:20:39
  Speaker 2: is going to be so great and so transformative that
• 2:20:42
  Speaker 2: we are investing a ton of money in advance of
• 2:20:46
  Speaker 2: the value and utility showing up. That's what they would say.
• 2:20:51
  Speaker 2: right and i've heard your rebuttal but i just wanted
• 2:20:53
  Speaker 2: to express i think that's their sentiment i'm not defending
• 2:20:56
  Speaker 2: them or anything i'm just i'm trying to provide enough
• 2:20:59
  Speaker 2: like balance to see if we can dance between these
• 2:21:03
  Speaker 2: two perspectives and a lot of people would say that
• 2:21:08
  Speaker 2: there's going to be a bloodbath because they can't all
• 2:21:10
  Speaker 2: win big in the way that they're kind of describing
• 2:21:13
  Speaker 2: so someone's going to have to lose and When one
• 2:21:15
  Speaker 2: of these players starts to lose big, I think it could,
• 2:21:17
  Speaker 2: as you say, there could be some kind of domino
• 2:21:19
  Speaker 2: effect or contraction.
• 2:21:20
  Speaker 1: Yeah, and I think the thing that people want to
• 2:21:22
  Speaker 1: believe is the dot-com bubble thing. It's like it worked
• 2:21:25
  Speaker 1: out afterwards because Amazon, Oracle, they didn't die after the
• 2:21:29
  Speaker 1: dot-com bubble. They're actually fine. It isn't like that. They're
• 2:21:32
  Speaker 1: bigger companies. They have bigger promises. And even, I'm not, like,
• 2:21:37
  Speaker 1: Oracle I actually think could die.
• 2:21:38
  Speaker 2: RIP, Larry. It couldn't happen to a nastier man. You
• 2:21:42
  Speaker 2: don't like these people, do you?
• 2:21:43
  Speaker 3: No.
• 2:21:44
  Speaker 2: Again, I ask this question purely because I want an answer,
• 2:21:47
  Speaker 2: not because I agree or disagree. But why don't you
• 2:21:50
  Speaker 2: like these people?
• 2:21:52
  Speaker 1: I don't like being misled. And I don't think regular
• 2:21:55
  Speaker 1: people are being misled either. And I really don't think
• 2:21:58
  Speaker 1: that the average person can get away with bullshitting as
• 2:22:01
  Speaker 1: much as these companies do. And I don't think the
• 2:22:03
  Speaker 1: average person gets anywhere near the level of affordance for
• 2:22:06
  Speaker 1: failure and lying as these companies do. And I think
• 2:22:09
  Speaker 1: there is a real... economic and human cost to allowing
• 2:22:13
  Speaker 1: these companies to run rampant and promise the world and
• 2:22:16
  Speaker 1: never really get called up on it the tepid nature
• 2:22:19
  Speaker 1: of criticism these days is so frustrating there are some
• 2:22:22
  Speaker 1: really great critics out there the really great people but
• 2:22:25
  Speaker 1: it's like seeing these ultra rich ultra wealthy ultra powerful
• 2:22:30
  Speaker 1: people life through their fucking teeth or misstate or whatever
• 2:22:33
  Speaker 1: people want to call it it turns my stomach And
• 2:22:36
  Speaker 1: I hate seeing people being misled. And I feel like
• 2:22:40
  Speaker 1: I write at such length because I really want people
• 2:22:41
  Speaker 1: to see why I've come to a conclusion. Am I right?
• 2:22:43
  Speaker 2: Am I wrong? I think I am.
• 2:22:44
  Speaker 1: Of course I do. But I also, I just find
• 2:22:49
  Speaker 1: it loathsome. I find these companies don't make good products anymore.
• 2:22:52
  Speaker 1: They don't care about their customers. And they treat their
• 2:22:56
  Speaker 1: customers with contempt.
• 2:22:59
  Speaker 2: If people want to go read more about your work,
• 2:23:02
  Speaker 2: you have a great sub stack.
• 2:23:03
  Speaker 1: Ghost, actually. It looks exactly like I'd moved off of
• 2:23:05
  Speaker 1: sub stack in 2024.
• 2:23:06
  Speaker 2: Oh, okay. And you also have a podcast, you do?
• 2:23:09
  Speaker 1: Yeah, Better Off Lime.
• 2:23:10
  Speaker 2: I'm going to link both of them below. So if
• 2:23:12
  Speaker 2: anyone wants to read more, get more detail and follow Ed,
• 2:23:14
  Speaker 2: I think it's, I would highly recommend. It is fascinating.
• 2:23:17
  Speaker 2: And do you know what? One of the things people
• 2:23:20
  Speaker 2: sometimes struggle with when they listen to podcasts is you
• 2:23:22
  Speaker 2: get lots of different opinions. And weirdly, I think they
• 2:23:24
  Speaker 2: think of, some people assume podcasts are going to be
• 2:23:26
  Speaker 2: like one person saying, The same thing as the next
• 2:23:29
  Speaker 2: person and the next person. Yeah, yeah, yeah. That is
• 2:23:31
  Speaker 2: just not the nature of information in the world and
• 2:23:32
  Speaker 2: opinions and progress and discussion. What happens is people have
• 2:23:36
  Speaker 2: different opinions. And I think my job, but also the
• 2:23:38
  Speaker 2: listener's job, is to try and pass through it and
• 2:23:41
  Speaker 2: over time collect more of these reference points from different
• 2:23:44
  Speaker 2: people and do your own research. Yeah. Whether it's on
• 2:23:47
  Speaker 2: your health or whether it's on something like this is
• 2:23:49
  Speaker 2: to watch and do your own research and to learn.
• 2:23:52
  Speaker 2: And I would say also never believe one person. Never
• 2:23:55
  Speaker 2: believe one particular perspective religiously. you know, collect a body
• 2:23:59
  Speaker 2: of evidence and follow the evidence yourself. But I love
• 2:24:02
  Speaker 2: watching your YouTube because it provides a different opinion and
• 2:24:08
  Speaker 2: that challenges me to think beyond my current opinion about
• 2:24:12
  Speaker 2: what might be possible. So when I've heard you talking
• 2:24:15
  Speaker 2: about how this is an economic bubble and I've heard
• 2:24:18
  Speaker 2: you talk about the CapEx spend with these big sort
• 2:24:20
  Speaker 2: of frontier AI labs, it really did make me pause
• 2:24:24
  Speaker 2: for a second and it really did make me consider
• 2:24:27
  Speaker 2: that there could be a bit of fugazi going on
• 2:24:30
  Speaker 2: here yeah and then it made me reflect on history
• 2:24:32
  Speaker 2: and go you know through history there's always a bit
• 2:24:33
  Speaker 2: of fugazi in these moments and oh that's an interesting
• 2:24:36
  Speaker 2: take and what's going to happen in 2027 2028 when
• 2:24:38
  Speaker 2: there's a bit of a market pullback and so i
• 2:24:41
  Speaker 2: highly recommend people go watch because you do you challenge
• 2:24:43
  Speaker 2: me to think differently um yeah and we need some
• 2:24:45
  Speaker 2: of those contrarian voices to to have honest discussions so
• 2:24:49
  Speaker 2: thank you for doing what you do really appreciate it
• 2:24:51
  Speaker 2: and i find you to be a very compelling captivating
• 2:24:53
  Speaker 2: communicator and i've i feel like i've learned a lot
• 2:24:55
  Speaker 2: today So I appreciate that. We have a closing tradition.
• 2:24:58
  Speaker 1: Yeah.
• 2:24:58
  Speaker 2: Where the last guest leaves a question for the next guest,
• 2:25:00
  Speaker 2: not knowing who they're leaving it for. And the question
• 2:25:02
  Speaker 2: left for you is, given that high quality relationships are
• 2:25:05
  Speaker 2: important for health and longevity, what should we be doing
• 2:25:08
  Speaker 2: to improve our relationships and social connection?
• 2:25:12
  Speaker 1: So this is actually connected to the AI bubble. So
• 2:25:16
  Speaker 1: I'm a critic.
• 2:25:17
  Speaker 2: I'm a skeptic. I have found that.
• 2:25:22
  Speaker 1: Showing and appreciating and loving the people around you and
• 2:25:25
  Speaker 1: uplifting them and raising them up as you succeed is
• 2:25:28
  Speaker 1: the way we do that. Your success should be everyone
• 2:25:30
  Speaker 1: around you. It's not economic. It's talking about Matt Hughes
• 2:25:34
  Speaker 1: for a while. Made me really happy. This whole thing
• 2:25:37
  Speaker 1: has been at times quite grueling and quite negative and
• 2:25:39
  Speaker 1: quite brutal. But the love I found and the joy
• 2:25:43
  Speaker 1: I found from community and the people, right? Because even
• 2:25:46
  Speaker 1: in the, in the small groups of haters, even like
• 2:25:49
  Speaker 1: Gary Marcus and sort of the people I talked to
• 2:25:51
  Speaker 1: Edward on gray, so junior, molly white brian merchant there
• 2:25:54
  Speaker 1: are so many people who have been loving and caring
• 2:25:56
  Speaker 1: and i think within especially these very critical moments when
• 2:26:00
  Speaker 1: you're like very much dialing in on how negative things
• 2:26:03
  Speaker 1: are how bad things are finding the people who maybe
• 2:26:08
  Speaker 1: find it repulsive to finding the people finding your people
• 2:26:11
  Speaker 1: who can be and the people who will talk to
• 2:26:13
  Speaker 1: you about it even like troy and jake my my
• 2:26:15
  Speaker 1: trainers who's so excited about this um even talk to
• 2:26:18
  Speaker 1: them about the shit as normal people knowing that there
• 2:26:21
  Speaker 1: are people there going through their own struggles, but also
• 2:26:24
  Speaker 1: to just give you the perspective and also remind you
• 2:26:27
  Speaker 1: that you are human too and focus. I know this
• 2:26:29
  Speaker 1: is kind of a all over the place point, but
• 2:26:30
  Speaker 1: it's just, it's really easy to get hard locked in
• 2:26:33
  Speaker 1: everything in life and to kind of get away from
• 2:26:36
  Speaker 1: why you do things and focus too much on the work.
• 2:26:38
  Speaker 1: When the most important thing at times is just to
• 2:26:41
  Speaker 1: know there are other people feeling the way you do.
• 2:26:43
  Speaker 1: And when I hear from my listeners, my readers a lot,
• 2:26:45
  Speaker 1: the most common thing I feel is they feel like
• 2:26:47
  Speaker 1: they have a voice and they feel like someone.
• 2:26:48
  Speaker 2: Is there for you.
• 2:26:49
  Speaker 1: Yeah. And, I don't think it can be understated how
• 2:26:52
  Speaker 1: much it means when you just reach out to someone
• 2:26:54
  Speaker 1: you love and tell them you love them. Tell them
• 2:26:56
  Speaker 1: their shit rocks. Say that their shit bangs. Tell everyone,
• 2:26:59
  Speaker 1: when you like an artist or a writer or a
• 2:27:01
  Speaker 1: podcast like this, tell them you fucking love it. We
• 2:27:04
  Speaker 1: don't do this enough and we need to do it more.
• 2:27:07
  Speaker 2: Well, that's a good closing message. So if you have
• 2:27:10
  Speaker 2: enjoyed the conversation today with Ed, please do let Ed
• 2:27:12
  Speaker 2: know that you love it down below. But please do
• 2:27:14
  Speaker 2: leave your opinions down below and I shall read all
• 2:27:16
  Speaker 2: of them. Ed, thank you so much. I'll link to
• 2:27:18
  Speaker 2: your website, but also to your YouTube channel where people
• 2:27:21
  Speaker 2: can learn more. And I would highly recommend you do
• 2:27:23
  Speaker 2: because it is truly fascinating. And I think we need
• 2:27:25
  Speaker 2: more voices that are demystifying a lot of the fugazi
• 2:27:28
  Speaker 2: and the narrative in this moment in time. And you're
• 2:27:30
  Speaker 2: certainly one of them. I really enjoyed the conversation. Thank
• 2:27:31
  Speaker 2: you so much. YouTube have this new crazy algorithm where
• 2:27:34
  Speaker 2: they know exactly what video you would like to watch
• 2:27:37
  Speaker 2: next based on AI and all of your viewing behavior.
• 2:27:40
  Speaker 2: And the algorithm says that this video is the perfect
• 2:27:44
  Speaker 2: video for you. It's different for everybody looking right now.
• 2:27:46
  Speaker 2: Check this video out. I bet you, you might love it.