Latest / INSPIRE STORY / #0102_ STORY INSPIRE: AI Whistleblower — Are AI Companies Hiding the Truth From Us? The Dark Reality Behind Artificial Intelligence

Transcript

#0102_ STORY INSPIRE: AI Whistleblower — Are AI Companies Hiding the Truth From Us? The Dark Reality Behind Artificial Intelligence · INSPIRE STORY

• 0:00
  Speaker 1: So much of what's happening today in the AI industry
• 0:02
  Speaker 1: is extremely inhumane.
• 0:04
  Speaker 2: But this is me playing devil's advocate. And logically, it
• 0:07
  Speaker 2: could be the case that the civilization that accelerate their
• 0:09
  Speaker 2: research with AI is going to be the superior civilization.
• 0:13
  Speaker 3: No, it's not. This is a prediction that you're making, right?
• 0:15
  Speaker 2: Elon's making, Zuckerberg's making, Altman's making.
• 0:18
  Speaker 1: And do you know what the common feature of all
• 0:19
  Speaker 1: of them is? They profit enormously off of this myth.
• 0:22
  Speaker 3: You know, I have all these.
• 0:23
  Speaker 1: Internal documents showing that they're purposely trying to create that
• 0:26
  Speaker 1: feeling within the public so that they can extract.
• 0:29
  Speaker 3: And exploit and extract and exploit.
• 0:31
  Speaker 2: So what do we do about it?
• 0:32
  Speaker 3: We need to break up the empires of AI.
• 0:35
  Speaker 1: You know, I've been covering the tech industry for over
• 0:37
  Speaker 1: eight years, interviewed over 250 people, including former or current
• 0:41
  Speaker 1: OpenAI employees and executives. And I can tell you that
• 0:43
  Speaker 1: there are many parallels between the empires of AI and
• 0:46
  Speaker 1: the empires of old.
• 0:47
  Speaker 3: Right, like, they claimed the.
• 0:48
  Speaker 1: Intellectual property of artists, writers, and creators in the pursuit
• 0:51
  Speaker 1: of training these models. Second, they exploit an extraordinary amount
• 0:53
  Speaker 1: of labor, which breaks the career ladder. because someone gets
• 0:57
  Speaker 1: laid off and then they work to train the models
• 0:59
  Speaker 1: on the very job that they were just laid off in,
• 1:02
  Speaker 1: which will then perpetuate more layoffs if that model then
• 1:04
  Speaker 1: develops that skill. And when they talk about that there's
• 1:07
  Speaker 1: gonna be some new jobs created that we can't even imagine,
• 1:09
  Speaker 1: a lot of the jobs that are created are way worse.
• 1:12
  Speaker 3: Than the jobs that were there.
• 1:14
  Speaker 1: And then there's the environmental and public health crisis that
• 1:17
  Speaker 1: these companies have created and how they're able to also
• 1:20
  Speaker 1: spend hundreds of millions to try and kill every possible
• 1:23
  Speaker 1: piece of legislation that gets in their way and will
• 1:25
  Speaker 1: censor researchers that are inconvenient to the empire's agenda. But
• 1:29
  Speaker 1: what I'm saying is not that these technologies don't have utility.
• 1:33
  Speaker 1: It's that the production of these technologies right now is
• 1:36
  Speaker 1: exacting a lot of harm on people.
• 1:38
  Speaker 3: But we have research that shows that the very.
• 1:41
  Speaker 1: Same capabilities could be developed in a different way that
• 1:45
  Speaker 1: doesn't have all of these unintended consequences.
• 1:48
  Speaker 3: So let's talk about all of that.
• 1:53
  Speaker 2: This is super interesting to me. My team gave me
• 1:54
  Speaker 2: this report to show me how many of you that
• 1:56
  Speaker 2: watch this show subscribe. And some of you have told us,
• 1:58
  Speaker 2: according to this, that you are unsubscribed from the channel randomly.
• 2:02
  Speaker 2: So favor to ask all of you, please could you
• 2:04
  Speaker 2: check right now if you've hit the subscribe button, if
• 2:06
  Speaker 2: you are a regular viewer of the show and you
• 2:07
  Speaker 2: like what we do here. We're approaching quite a significant
• 2:09
  Speaker 2: landmark on this show in terms of a subscriber number.
• 2:12
  Speaker 2: So if there was one simple free thing that you
• 2:15
  Speaker 2: could do to help us, my team, everyone here, to
• 2:17
  Speaker 2: keep this show free, to keep it improving year over
• 2:20
  Speaker 2: year and week over week, it is just to hit
• 2:22
  Speaker 2: that subscribe button and to double check if you've hit it.
• 2:24
  Speaker 2: Only thing I'll ever ask of you, do we have
• 2:26
  Speaker 2: a deal? If you do it, I'll tell you what
• 2:27
  Speaker 2: I'll do. I'll make sure every single week, every single month,
• 2:31
  Speaker 2: we fight harder and harder and harder and harder to
• 2:33
  Speaker 2: bring you the guests and conversations that you want to hear.
• 2:35
  Speaker 2: I've stayed true to that promise since the very beginning
• 2:37
  Speaker 2: of the Diary of a CEO, and I will not
• 2:39
  Speaker 2: let you down. Please help us. Really appreciate it. Let's
• 2:42
  Speaker 2: get on with the show. Karen, how? You've written this
• 2:50
  Speaker 2: book in front of me here called Empire of AI,
• 2:52
  Speaker 2: Dreams and Nightmares in Sam Altman's OpenAI. I guess my
• 2:56
  Speaker 2: first question is, what is the research and the journey
• 2:59
  Speaker 2: you went on in order to write this book we're
• 3:02
  Speaker 2: going to talk about and the subjects within it today?
• 3:04
  Speaker 1: I took a strange route into journalism. I studied mechanical
• 3:07
  Speaker 1: engineering at MIT. And so when I graduated, I moved
• 3:11
  Speaker 1: to San Francisco. I joined a tech startup. I became
• 3:13
  Speaker 1: part of Silicon Valley. And I basically received an education
• 3:18
  Speaker 1: in what Silicon Valley is about because a few months
• 3:20
  Speaker 1: into joining a very mission-driven startup that was focused on
• 3:22
  Speaker 1: building technologies that would help facilitate the fight against climate change,
• 3:27
  Speaker 1: The board fired the CEO because the company was not profitable.
• 3:31
  Speaker 1: And this was, in hindsight, a very pivotal moment for
• 3:34
  Speaker 1: me because I thought if this hub is ultimately geared
• 3:39
  Speaker 1: towards building profitable technologies, and many of the problems in
• 3:43
  Speaker 1: the world that I think need solved are not profitable
• 3:46
  Speaker 1: problems like climate change, then what are we actually doing here?
• 3:52
  Speaker 1: How did we get to a point where innovation is
• 3:54
  Speaker 1: not actually necessarily working in the public benefit and sometimes
• 3:58
  Speaker 1: even undermining the public benefit in pursuit of profit. In
• 4:01
  Speaker 1: that moment, I had a bit of a crisis where
• 4:05
  Speaker 1: I thought, well, I just spent four years trying to
• 4:09
  Speaker 1: set myself up for this career that I now don't
• 4:11
  Speaker 1: think I am cut out for. And I thought, well,
• 4:17
  Speaker 1: I might as well just try something totally different. I've
• 4:19
  Speaker 1: always liked writing. And that's how, after two years, I
• 4:24
  Speaker 1: landed at a role at MIT Technology Review covering AI full-time.
• 4:28
  Speaker 1: And that gave me a space to then explore all
• 4:31
  Speaker 1: of these questions of Who gets to decide what technologies
• 4:34
  Speaker 1: we build? How does money and ideology also drive the
• 4:38
  Speaker 1: production of those technologies? And how do we ultimately make
• 4:41
  Speaker 1: sure that we actually reimagine the innovation ecosystem to work
• 4:46
  Speaker 1: for a broad base of people all around the world?
• 4:50
  Speaker 1: And so that is kind of how I then set
• 4:53
  Speaker 1: off on this journey. of ultimately writing a book. I
• 4:56
  Speaker 1: didn't realize that I was working towards writing a book,
• 4:59
  Speaker 1: but starting in 2018 when I took that job was
• 5:04
  Speaker 1: essentially the moment in which I began researching the story
• 5:07
  Speaker 1: that I document in it.
• 5:09
  Speaker 2: Very timely time to start working in artificial intelligence. For
• 5:12
  Speaker 2: anyone that doesn't know, this is pre-OpenAI ChatGPT launch moment
• 5:17
  Speaker 2: that shook the world. But in writing this book, you
• 5:20
  Speaker 2: interviewed a lot of people and went to a lot
• 5:22
  Speaker 2: of places. Can you give me a flavor of how
• 5:23
  Speaker 2: many people you've interviewed where it's taken you around the world,
• 5:26
  Speaker 2: et cetera?
• 5:27
  Speaker 1: I interviewed over 250 people, so over 300 interviews. Over
• 5:31
  Speaker 1: 90 of those people were former or current OpenAI employees
• 5:35
  Speaker 1: and executives. So the book covers the inside story of
• 5:39
  Speaker 1: OpenAI's first decade and how it ultimately got to where
• 5:42
  Speaker 1: it is today. But I didn't want to write a
• 5:45
  Speaker 1: corporate book. I felt very strongly that in order to
• 5:49
  Speaker 1: help people understand the impact of the AI industry, we
• 5:53
  Speaker 1: would also have to travel well beyond Silicon Valley. These
• 5:56
  Speaker 1: companies tell us that AI is going to benefit everyone
• 5:59
  Speaker 1: and that's their mission. But you really start to see
• 6:02
  Speaker 1: that rhetoric break down when you go to the places
• 6:06
  Speaker 1: that look nothing like Silicon Valley, that speak nothing like
• 6:09
  Speaker 1: Silicon Valley, and that have a history and culture that
• 6:12
  Speaker 1: are fundamentally different as well. And that's where you start
• 6:15
  Speaker 1: to really understand the true reality of how this industry
• 6:20
  Speaker 1: works is unfolding around us.
• 6:22
  Speaker 2: Karen, I often try and steer conversations, but in this situation,
• 6:26
  Speaker 2: I feel like it's probably my responsibility to follow. So
• 6:29
  Speaker 2: with that in mind, I'm going to ask you, where
• 6:31
  Speaker 2: does this journey begin and where should we be starting
• 6:34
  Speaker 2: if we're talking about the subjects of empire of AI,
• 6:37
  Speaker 2: AI generally, artificial intelligence? And also I'd say, one thing
• 6:40
  Speaker 2: I'm really keen to do in this conversation, which is
• 6:42
  Speaker 2: I often see in conversations is left out, is let's
• 6:45
  Speaker 2: assume that our viewers know nothing about AI.
• 6:48
  Speaker 3: Yeah.
• 6:49
  Speaker 2: So they don't know what scaling laws are or GPUs
• 6:51
  Speaker 2: or compute or whatever. And let's try and keep this
• 6:53
  Speaker 2: as simple as we possibly can in terms of language
• 6:56
  Speaker 2: or explain all the complicated language so that we can
• 6:59
  Speaker 2: bring as much people with us as we possibly can.
• 7:01
  Speaker 1: Yes.
• 7:02
  Speaker 2: Where should we start?
• 7:03
  Speaker 1: I think we should start with when AI started as
• 7:07
  Speaker 1: a field. So this was back in 1956. And there
• 7:12
  Speaker 1: were a group of scientists that gathered at Dartmouth University
• 7:14
  Speaker 1: to start a new discipline, a scientific discipline, to try
• 7:18
  Speaker 1: and chase an ambition. And specifically, an assistant professor at
• 7:22
  Speaker 1: Dartmouth University, John McCarthy, decided to name this discipline artificial intelligence.
• 7:28
  Speaker 1: This was not the first name that he tried. The
• 7:30
  Speaker 1: previous year, he tried to name it automata studies. And
• 7:34
  Speaker 1: the reason why some of his colleagues were concerned about
• 7:36
  Speaker 1: this name was because it pegged the idea of this
• 7:40
  Speaker 1: discipline to recreating human intelligence. And back then, as is
• 7:45
  Speaker 1: true today, we have no scientific consensus around what human
• 7:50
  Speaker 1: intelligence is. There's no definition from psychology, biology, neurology. And
• 7:55
  Speaker 1: in fact, every attempt in history to quantify and rank
• 8:00
  Speaker 1: human intelligence has been driven by nefarious motives. It's been
• 8:06
  Speaker 1: driven by a desire to prove scientifically that certain groups
• 8:10
  Speaker 1: of people are inferior to other groups of people. There
• 8:14
  Speaker 1: are no goalposts for this field, and there are no
• 8:18
  Speaker 1: goalposts for the industry when they say that they are
• 8:20
  Speaker 1: ultimately trying to recreate AI systems that would be as
• 8:24
  Speaker 1: smart as humans. How do we even define what that means?
• 8:28
  Speaker 1: And when are we going to get there if we
• 8:30
  Speaker 1: don't know how to define the destination. And what that
• 8:35
  Speaker 1: effectively means is that these companies can just use the
• 8:39
  Speaker 1: term artificial general intelligence, which is now the term to
• 8:42
  Speaker 1: refer to this ambitious goal to recreate human intelligence. They
• 8:47
  Speaker 1: can use it however they want to, and they can
• 8:50
  Speaker 1: define and redefine it based on what is convenient for them.
• 8:52
  Speaker 1: So in OpenAI's history, it has defined and redefined it
• 8:55
  Speaker 1: many times. When Sam Altman is talking with Congress, AGI
• 8:59
  Speaker 1: is a system that's going to cure cancer solve climate change,
• 9:03
  Speaker 1: cure poverty. When he's talking with consumers that he's trying
• 9:07
  Speaker 1: to sell his products to, it's the most amazing digital
• 9:10
  Speaker 1: assistant that you're ever going to have. When he was
• 9:14
  Speaker 1: talking with Microsoft, you know, in the deal that OpenAI
• 9:17
  Speaker 1: and Microsoft struck, where Microsoft invested in the company, it
• 9:21
  Speaker 1: was defined as a system that will generate $ 100 billion
• 9:23
  Speaker 1: of revenue. And on OpenAI's own website, they define it
• 9:28
  Speaker 1: as highly autonomous systems that outperform humans in most economically
• 9:33
  Speaker 1: valuable work. This is like not a coherent, vision of
• 9:38
  Speaker 1: one technology. These are very different definitions that are spoken
• 9:42
  Speaker 1: out loud to the audience that needs to be mobilized
• 9:45
  Speaker 1: to ward off regulation or get more consumer buy-in into
• 9:50
  Speaker 1: the industry's quest or to get more capital, more resources
• 9:55
  Speaker 1: for continuing on this journey with ambiguous definitions.
• 10:00
  Speaker 2: I mean, speaking about different definitions through time, in 2015,
• 10:04
  Speaker 2: in a blog post that Sam Altman wrote before OpenArea
• 10:07
  Speaker 2: was officially announced, he explicitly outlined the existential risk by saying,
• 10:12
  Speaker 2: development of superhuman machine intelligence is probably the greatest threat
• 10:16
  Speaker 2: to the continued existence of humanity. There are other threats
• 10:19
  Speaker 2: that I think are more certain to happen. For example,
• 10:22
  Speaker 2: an engineered virus. But AI is probably the most likely
• 10:26
  Speaker 2: way to destroy everything.
• 10:28
  Speaker 1: In general, when Altman is writing for the public or
• 10:32
  Speaker 1: speaking for the public, he does not just have the
• 10:34
  Speaker 1: public as the audience in mind. There are other people
• 10:38
  Speaker 1: that he is trying to motivate or mobilize when he
• 10:42
  Speaker 1: says these things. And in that particular moment, Altman was
• 10:46
  Speaker 1: trying to convince Elon Musk to join him on co-founding OpenAI.
• 10:51
  Speaker 1: And Musk, in particular, was spending all of his time
• 10:55
  Speaker 1: sounding the alarm on what he saw as a huge
• 10:59
  Speaker 1: existential threat that AI could pose. And so in that
• 11:02
  Speaker 1: blog post, if you look at the language that Altman
• 11:05
  Speaker 1: uses side by side with the language that Musk was
• 11:07
  Speaker 1: using at the time, it mirrors all the things that
• 11:11
  Speaker 1: Musk was saying.
• 11:11
  Speaker 2: It's identical. Ten years ago, Musk was going on podcasts, saying, tweeting, whatever,
• 11:16
  Speaker 2: that the greatest existential risk to humanity was AI.
• 11:19
  Speaker 3: Yeah.
• 11:20
  Speaker 1: And so, you know, like his parenthetical, there are other
• 11:23
  Speaker 1: things that might actually be more likely to happen, like
• 11:26
  Speaker 1: engineered viruses. It's because up until then, Altman had been
• 11:30
  Speaker 1: talking just about engineered viruses. And so now that he
• 11:35
  Speaker 1: needs to pivot to speak to an audience of one,
• 11:37
  Speaker 1: to Musk... he needs to kind of resolve the contradiction
• 11:41
  Speaker 1: between what he's now elevating as his new central fear
• 11:46
  Speaker 1: to be the same as Musk's new central fear with
• 11:48
  Speaker 1: what he had previously been saying. So that's why he's like,
• 11:51
  Speaker 1: I think this is now, even though before I said this.
• 11:55
  Speaker 2: And are you saying that Sam Altman manipulated Musk? Because
• 11:59
  Speaker 2: Elon did end up donating a huge amount of money
• 12:02
  Speaker 2: to OpenAI and co-founding it, I believe. with Sam Altman?
• 12:07
  Speaker 3: Elon Musk did end up co-founding it with Altman.
• 12:08
  Speaker 1: And certainly from Musk's perspective, he does feel manipulated because
• 12:13
  Speaker 1: he feels like Altman was engineering his language in a
• 12:20
  Speaker 1: way that would make Musk trust him as a partner
• 12:25
  Speaker 1: in this endeavor.
• 12:26
  Speaker 3: And of course, then Musk leaves.
• 12:29
  Speaker 1: And Through some of the documents that came out during
• 12:32
  Speaker 1: the lawsuit that Musk and Altman are engaged in now,
• 12:36
  Speaker 1: it has become clear that there was a degree to
• 12:39
  Speaker 1: which Musk was actually muscled out a little bit. And
• 12:43
  Speaker 1: so that's why he's left with this very intense personal
• 12:48
  Speaker 1: vendetta against Altman, saying that somehow Altman tricked him into
• 12:52
  Speaker 1: being part of this.
• 12:53
  Speaker 2: So in 2015, Sam Altman is writing these blog posts
• 12:57
  Speaker 2: saying this is one of the greatest existential threats. At
• 12:59
  Speaker 2: the same time, in 2015, Musk is doing some very
• 13:02
  Speaker 2: famous speeches at the time. At MIT, he said that
• 13:06
  Speaker 2: AI was the biggest existential threat and compared developing AI
• 13:09
  Speaker 2: to summoning the demon. And what you're saying here is
• 13:12
  Speaker 2: you're saying that Sam Altman was just mirroring the language
• 13:15
  Speaker 2: that Elon was using to get Elon involved in open AI.
• 13:18
  Speaker 2: And later it appears, and again, there's a legal case
• 13:21
  Speaker 2: taking place now, that Sam might have muscled Elon out
• 13:24
  Speaker 2: in some capacity. Yeah.
• 13:26
  Speaker 1: So we know from the lawsuit and the documents that
• 13:28
  Speaker 1: have come out in the lawsuit that Ilya Sutskever, who
• 13:32
  Speaker 1: was the chief scientist of OpenAI at the time, and
• 13:34
  Speaker 1: Greg Brockman, chief technology officer at the time, when they
• 13:38
  Speaker 1: were deciding whether or not to maintain OpenAI as a nonprofit,
• 13:43
  Speaker 1: because it was originally found as a nonprofit, they decided, okay,
• 13:45
  Speaker 1: we need to create a for-profit entity. But the question was,
• 13:48
  Speaker 1: who should be the CEO of this for-profit entity? Should
• 13:50
  Speaker 1: it be Musk or should it be Altman? Because they
• 13:52
  Speaker 1: were the two co-chairmen of the nonprofit. And in the emails,
• 13:57
  Speaker 1: it became clear that Ilya and Greg first chose Musk
• 14:02
  Speaker 1: to be the CEO. But through my reporting, I discovered
• 14:08
  Speaker 1: that Altman then appealed personally to Greg Brockman, who was
• 14:13
  Speaker 1: a friend of his that they'd known each other for
• 14:16
  Speaker 1: many years through the Silicon Valley scene, and said, don't
• 14:19
  Speaker 1: you think that it would be a little bit dangerous
• 14:22
  Speaker 1: to have Musk be the CEO of this company, this
• 14:26
  Speaker 1: new for-profit entity? Because, you know, he's a famous guy.
• 14:30
  Speaker 1: He has a lot of pressures in the world. He
• 14:32
  Speaker 1: could be threatened. He could act erratically. He could be unpredictable.
• 14:37
  Speaker 1: And do we really want a technology that could be
• 14:40
  Speaker 1: super powerful in the future to end up in the
• 14:43
  Speaker 1: hands of this man? And that convinced Greg. And Greg
• 14:48
  Speaker 1: then convinced Ilya, you know, I think there's a point here.
• 14:51
  Speaker 1: Do we really want to give this much power to Musk?
• 14:55
  Speaker 1: And that is why Musk then leaves because then they
• 14:59
  Speaker 1: start— the two switch their allegiances. They say, actually, we
• 15:02
  Speaker 1: want Altman to be the CEO. And then Musk is like,
• 15:04
  Speaker 1: if I'm not CEO, I'm out.
• 15:06
  Speaker 2: So it sounds like Sam, again, managed to persuade someone
• 15:09
  Speaker 2: to do something. I guess this begs the question, what
• 15:14
  Speaker 2: do you think of Sam Altman?
• 15:17
  Speaker 3: I think he's a very controversial figure.
• 15:19
  Speaker 2: You did an interesting pause. It's a pause where someone
• 15:23
  Speaker 2: tries to select their words.
• 15:26
  Speaker 1: Well, this is what's so interesting about... about those interviews
• 15:31
  Speaker 1: is people are extremely polarized on Altman. No one has
• 15:37
  Speaker 1: in-between feelings about him. Either they think he's the greatest
• 15:41
  Speaker 1: tech leader of this generation akin to the Steve Jobs
• 15:44
  Speaker 1: of the modern era, or they think that he's really
• 15:48
  Speaker 1: manipulative and an abuser and a liar. And what I
• 15:52
  Speaker 1: realized because I interviewed so many people is it really
• 15:56
  Speaker 1: comes down to what that person's vision of the future
• 16:00
  Speaker 1: is and what their goals are. So if you align
• 16:05
  Speaker 1: with Altman's vision of the future, you're going to think
• 16:07
  Speaker 1: he's the greatest asset ever to have on your side
• 16:10
  Speaker 1: because this man is really persuasive. He's incredible at telling stories.
• 16:14
  Speaker 1: He's incredible at mobilizing capital, at recruiting talent, at getting
• 16:18
  Speaker 1: all the inputs that you need. to then make that
• 16:20
  Speaker 1: future happen. But if you don't agree with his vision
• 16:24
  Speaker 1: of the future, then you begin to feel like you're
• 16:28
  Speaker 1: being manipulated by him to support his vision, even if
• 16:33
  Speaker 1: you fundamentally don't agree with it. And this is the
• 16:36
  Speaker 1: story especially of Dario Amede, CEO of Anthropic, who was
• 16:40
  Speaker 1: originally an executive at OpenAI.
• 16:43
  Speaker 2: So for people that don't know, Dario now runs Anthropic,
• 16:45
  Speaker 2: which is the maker of Claude. A lot of people
• 16:47
  Speaker 2: probably are more familiar with Claude. Yeah.
• 16:50
  Speaker 1: And it's one of the biggest competitors to OpenAI. And Amadei,
• 16:55
  Speaker 1: at the time when he was an executive at OpenAI,
• 16:59
  Speaker 1: he thought that Altman was on the same page with him.
• 17:04
  Speaker 1: And then over time began to feel that Altman was
• 17:07
  Speaker 1: actually on exactly the opposite page of him and felt
• 17:12
  Speaker 1: that Altman had used Amadei's intelligence, capabilities, skills, and to
• 17:19
  Speaker 1: build things and bring about a vision of the future
• 17:23
  Speaker 1: that he actually fundamentally didn't agree with. And so that's
• 17:26
  Speaker 1: why people end up with this bad taste in their mouths.
• 17:29
  Speaker 1: And so, you know, I've been covering the tech industry
• 17:32
  Speaker 1: for over eight years and covered many companies. I've covered Meta, Google, Microsoft,
• 17:37
  Speaker 1: in addition to OpenAI. And OpenAI and Altman, it's the
• 17:42
  Speaker 1: only figure that I've seen this degree of polarization with.
• 17:47
  Speaker 1: where people cannot decide whether he's the greatest.
• 17:52
  Speaker 2: Or the worst. You mentioned Dario there. And I found
• 17:56
  Speaker 2: it really, what I found really interesting is to look
• 17:58
  Speaker 2: at how people's quotes evolve over time with their incentives.
• 18:00
  Speaker 2: So I was looking at all of the, all of
• 18:02
  Speaker 2: the things they've said on the record on podcasts, in
• 18:05
  Speaker 2: their blog posts to see how it's evolved over time.
• 18:07
  Speaker 2: And Dario, who was the former VP of research, OpenAI,
• 18:11
  Speaker 2: and has now moved on to Anthropic, who are taking
• 18:13
  Speaker 2: a slightly different approach to developing AI. said back in
• 18:17
  Speaker 2: 2017 while he was still at OpenAI that, this is
• 18:20
  Speaker 2: a quote, I think at the extreme end is the
• 18:22
  Speaker 2: Nick Bostrom style of fear that an AGI could destroy humanity.
• 18:26
  Speaker 2: I can't see any reason in principle why that couldn't happen.
• 18:30
  Speaker 2: My chance that something goes really quite catastrophically wrong on
• 18:34
  Speaker 2: the scale of human civilization might be somewhere between 10%
• 18:38
  Speaker 2: and 25%. And also you mentioned Ilya, who was a
• 18:43
  Speaker 2: co-founder of OpenAI and then left OpenAI I guess the
• 18:46
  Speaker 2: first question I'd ask is why did Ilya leave?
• 18:48
  Speaker 3: That's a great question.
• 18:52
  Speaker 1: So he was instrumental in trying to get Sam Altman fired.
• 18:56
  Speaker 1: And he's another one of the people who over time
• 19:00
  Speaker 1: began to feel like he was being manipulated by Altman
• 19:03
  Speaker 1: towards contributing something that he didn't believe in.
• 19:06
  Speaker 3: How did you know? Because I interviewed a lot of people.
• 19:09
  Speaker 1: Ilya in particular had two pillars that he cared about deeply.
• 19:16
  Speaker 1: One is making sure we get to so-called AGI. And
• 19:21
  Speaker 1: the other is making sure that we get to it safely.
• 19:24
  Speaker 1: And he felt that Altman was actively undermining both things.
• 19:29
  Speaker 1: He felt that Altman was creating a very chaotic environment
• 19:32
  Speaker 1: within the company where he was pitting teams against each other,
• 19:36
  Speaker 1: where he was telling different things to different people.
• 19:39
  Speaker 2: Have you ever spoken to him?
• 19:40
  Speaker 3: I have.
• 19:41
  Speaker 1: So I interviewed him in 2019 for a profile that
• 19:45
  Speaker 1: I did of OpenAI for MIT Technology Review.
• 19:49
  Speaker 2: And back in 2019? 2019, he has a quote where
• 19:51
  Speaker 2: he says, the future is going to be good for
• 19:52
  Speaker 2: AIs regardless. It would be nice if it was also
• 19:54
  Speaker 2: good for humans as well. It's not that it's going
• 19:57
  Speaker 2: to actively hate humans or want to harm them, but
• 19:59
  Speaker 2: it's just going to be so powerful. And I think
• 20:01
  Speaker 2: a good analogy would be the way that humans treat animals.
• 20:04
  Speaker 2: It's not that we hate animals. I think humans love
• 20:06
  Speaker 2: animals and I have a lot of affection for them.
• 20:08
  Speaker 2: But when the time comes to build a highway between
• 20:11
  Speaker 2: two cities, we are not asking the animals for permission.
• 20:14
  Speaker 2: We just do it because it's important to us. And
• 20:16
  Speaker 2: I think by default, that's the kind of relationship that's
• 20:19
  Speaker 2: going to be between us and AI, which are truly
• 20:23
  Speaker 2: autonomous and operating on their own behalf. And that was
• 20:26
  Speaker 2: in 2019, the year that you interviewed him.
• 20:29
  Speaker 1: One of the things that I feel like we should
• 20:31
  Speaker 1: take a step back to examine is going back to
• 20:33
  Speaker 1: this idea of what even is artificial intelligence and what
• 20:36
  Speaker 1: do we mean by intelligence? And a huge part of
• 20:40
  Speaker 1: the views of the different people in the quotes that
• 20:42
  Speaker 1: you're reading derives from a specific belief that they each
• 20:46
  Speaker 1: have in this question of what is intelligence, what constitutes intelligence.
• 20:52
  Speaker 1: For Ilya, he has throughout his research career felt that
• 20:58
  Speaker 1: ultimately our brains are giant statistical models. This is not
• 21:03
  Speaker 1: something that, you know, we actually know, but this is
• 21:06
  Speaker 1: his own hypothesis. Also the hypothesis of his mentor, Jeffrey Hinton,
• 21:11
  Speaker 1: who also was on this podcast. This is why they
• 21:14
  Speaker 1: have such a strong conviction in the idea of building
• 21:17
  Speaker 1: AI systems that are statistical models and that this particular
• 21:21
  Speaker 1: approach is going to lead to intelligent systems as we
• 21:25
  Speaker 1: are intelligent. It's a hypothesis that they have. It's not
• 21:29
  Speaker 1: one that has been proven by science. And some people
• 21:33
  Speaker 1: vehemently disagree with them on this particular thing. But if
• 21:36
  Speaker 1: you step into their shoes and take on that hypothesis
• 21:40
  Speaker 1: and assume that it's true, that our brains are in
• 21:43
  Speaker 1: fact statistical engines, And that these systems that they're building
• 21:49
  Speaker 1: are also statistical engines that they're making bigger and bigger
• 21:52
  Speaker 1: and bigger until they become the size of the human brain.
• 21:55
  Speaker 1: That's why they say that making this comparison where the
• 22:00
  Speaker 1: system will become equal to human intelligence and then maybe
• 22:03
  Speaker 1: exceed human intelligence is relevant in their framework. And Ilya
• 22:09
  Speaker 1: gave a talk at one point at this conference. really
• 22:12
  Speaker 1: prominent AI research conference that happens every year called Neural
• 22:15
  Speaker 1: Information Processing Systems.
• 22:18
  Speaker 3: It's a mouthful. But he gave this keynote where.
• 22:21
  Speaker 1: He shows this chart of the size of brains and
• 22:26
  Speaker 1: the intelligence of a species. And it's roughly linear. The
• 22:31
  Speaker 1: bigger the size of the brain, the more intelligent the species.
• 22:34
  Speaker 1: And so for him, he thinks he's building a digital
• 22:38
  Speaker 1: brain because he thinks brains.
• 22:41
  Speaker 3: Are just statistical engines.
• 22:43
  Speaker 1: So from that logic, it's like, okay, if we then
• 22:45
  Speaker 1: build a bigger statistical engine than the human brain, then
• 22:50
  Speaker 1: based on this chart, it will be more intelligent and
• 22:54
  Speaker 1: then we will be subjected to the same treatment that
• 22:56
  Speaker 1: we've subjected animals to. But it's really important to understand
• 23:00
  Speaker 1: that these are scientific hypotheses of specific individuals within the
• 23:04
  Speaker 1: AI research community. And there's a lot, a lot of
• 23:08
  Speaker 1: debate about whether this is in fact the case. And
• 23:12
  Speaker 1: some of the biggest critics say it's very reductive to
• 23:15
  Speaker 1: think of our brains as simply just statistical engines.
• 23:18
  Speaker 2: Why does it matter to know the mechanism? Is it
• 23:24
  Speaker 2: not just important to know the outcome, which is that
• 23:27
  Speaker 2: it's going to be able to do make a video
• 23:29
  Speaker 2: for me or agents are going to be able to
• 23:31
  Speaker 2: do the work that I do. Does it really, really
• 23:33
  Speaker 2: matter for us to know the mechanism behind it?
• 23:36
  Speaker 3: Yes and no.
• 23:37
  Speaker 1: So it matters because these companies, they are driving their
• 23:43
  Speaker 1: future actions based on this hypothesis. So they have decided,
• 23:50
  Speaker 1: we think that this hypothesis is true. Like, we should
• 23:53
  Speaker 1: just continue building larger and larger statistical models in the
• 23:56
  Speaker 1: pursuit of artificial general intelligence. And that's then having global consequences. Like,
• 24:03
  Speaker 1: in order to continue doing that, they're hoovering up more
• 24:06
  Speaker 1: and more data. They're building more and more data centers.
• 24:09
  Speaker 1: They are exploiting more and more labor in order to
• 24:13
  Speaker 1: continue on this path. Here's a question that I think
• 24:17
  Speaker 1: is important to ask is, why are we trying to
• 24:20
  Speaker 1: build AI systems that are duplicative of humans? We're kind
• 24:23
  Speaker 1: of having this conversation right now where we've just taken
• 24:26
  Speaker 1: the premise of this industry as a good thing. Like,
• 24:31
  Speaker 1: they said that we should be building AGI, so we
• 24:33
  Speaker 1: say that we should be building AGI. But I would
• 24:35
  Speaker 1: like to ask, like, why are we doing that? Why
• 24:39
  Speaker 1: is it that we are building a technology that is
• 24:42
  Speaker 1: ultimately designed to replace and automate people away? That is
• 24:46
  Speaker 1: not the enterprise of technology. Like, we should be building technology.
• 24:52
  Speaker 1: And the purpose of technology throughout history has been to
• 24:55
  Speaker 1: improve human flourishing, not to replace people.
• 25:00
  Speaker 3: And so.
• 25:02
  Speaker 1: This is like a critical part of my critique of
• 25:05
  Speaker 1: these companies and these scientists that have just adopted this
• 25:09
  Speaker 1: goal and have relentlessly pursued it and have had enormous
• 25:12
  Speaker 1: capital and enormous resources to pursue it. Is this the
• 25:15
  Speaker 1: right goal? Like, why are we doing this? Why can't
• 25:19
  Speaker 1: we just build AI systems? that do things like accelerate
• 25:23
  Speaker 1: drug discovery and improve people's healthcare outcomes, which are systems
• 25:28
  Speaker 1: that have nothing to do with the statistical engines that
• 25:31
  Speaker 1: they're trying to build to duplicate the human brain.
• 25:33
  Speaker 2: So why are they doing it? I mean, you've interviewed
• 25:35
  Speaker 2: all these people. I think it's, what, 300 people in total?
• 25:38
  Speaker 2: 80 or 90 of them from OpenAI, the maker of ChatGBC.
• 25:42
  Speaker 2: Why do you think they're doing it?
• 25:45
  Speaker 1: I think it's because they're driven by an imperial agenda,
• 25:47
  Speaker 1: and that is why I call these companies empires of AI.
• 25:50
  Speaker 2: What do you mean by an imperial agenda? What does
• 25:52
  Speaker 2: that term mean?
• 25:53
  Speaker 1: Empire is the only metaphor that I've ever found to
• 25:58
  Speaker 1: fully encapsulate all of the dimensions of what these companies
• 26:01
  Speaker 1: do and the scale that they operate and what motivates
• 26:06
  Speaker 1: them to do what they do. And there are many
• 26:09
  Speaker 1: parallels that you see between what I call the empires
• 26:11
  Speaker 1: of AI and the empires of old. They lay claim
• 26:15
  Speaker 1: to resources that are not their own in the pursuit
• 26:17
  Speaker 1: of training these models. That's the data of individuals, the
• 26:20
  Speaker 1: intellectual property of artists, writers, and creators. They're land grabbing
• 26:24
  Speaker 1: in order to build these supercomputer facilities for training the
• 26:27
  Speaker 1: next generation models. Second, they exploit an extraordinary amount of labor.
• 26:31
  Speaker 1: They contract hundreds of thousands of workers all around the world,
• 26:35
  Speaker 1: including in the US, to ultimately make these technologies.
• 26:40
  Speaker 3: We can talk about that more.
• 26:42
  Speaker 1: And they also designed their tools to be labor automating
• 26:46
  Speaker 1: so that when the technologies are deployed, it also affects
• 26:50
  Speaker 1: labor rights because it erodes away labor rights. And this
• 26:53
  Speaker 1: is a political choice that they have. Third, they monopolize
• 26:58
  Speaker 1: knowledge production. So they project this idea that they're the
• 27:00
  Speaker 1: only ones that really understand how the technology works. And
• 27:03
  Speaker 1: so if the public doesn't like it, it's because they
• 27:05
  Speaker 1: don't actually know enough about this technology.
• 27:08
  Speaker 3: They do this to the public. They do this to policymakers.
• 27:11
  Speaker 1: And they've also captured the majority of the scientists that
• 27:14
  Speaker 1: are working on understanding the limitations and capabilities of AI.
• 27:17
  Speaker 2: You think they're gaslighting the public?
• 27:20
  Speaker 3: In a way. They are, yeah.
• 27:21
  Speaker 1: So if most of the climate scientists in the world
• 27:25
  Speaker 1: were bankrolled by fossil fuel companies, do you think we
• 27:28
  Speaker 1: would get an accurate picture of the climate crisis? No.
• 27:32
  Speaker 1: And in the same way, they employ and bankroll, the
• 27:36
  Speaker 1: AI industry, employs and bankrolls most of the AI researchers
• 27:39
  Speaker 1: in the world. So they set the agenda on AI
• 27:42
  Speaker 1: research in soft ways, simply by funneling money to their priorities.
• 27:47
  Speaker 1: so that only certain types of AI research are produced,
• 27:50
  Speaker 1: but they also will censor researchers when they do not
• 27:55
  Speaker 1: like what the researcher has found. And so I talk
• 27:58
  Speaker 1: about the case of Dr. Timnit Gebru in my book,
• 28:01
  Speaker 1: who was the ethical AI team co-lead at Google when
• 28:05
  Speaker 1: she was literally hired to critique the types of AI
• 28:09
  Speaker 1: systems that Google was building. She then co-wrote a critical
• 28:14
  Speaker 1: research paper that was showing how large language models specifically
• 28:18
  Speaker 1: were leading to certain types of harmful outcomes. And in
• 28:22
  Speaker 1: an attempt to try and stop this research from being published,
• 28:26
  Speaker 1: Google ended up firing Gebru and then fired her other co-lead,
• 28:31
  Speaker 1: Margaret Mitchell. And so they control and quash the research
• 28:39
  Speaker 1: that is inconvenient to the empire's agenda.
• 28:42
  Speaker 2: Did you have an example where this is happening to
• 28:44
  Speaker 2: journalists as well, that are asking questions of their team members?
• 28:49
  Speaker 2: I think I was watching a video of yours where
• 28:51
  Speaker 2: there was a young man that was saying he had
• 28:53
  Speaker 2: someone show up at his door, knocked on his door
• 28:55
  Speaker 2: and asked for information, emails, text messages, and this person
• 28:59
  Speaker 2: was from one of the big AI companies.
• 29:01
  Speaker 1: This was opening, I started subpoenaing some of its critics, yeah,
• 29:05
  Speaker 1: as part of a what appears to be a campaign
• 29:10
  Speaker 1: of intimidation, but also what appeared to be a campaign
• 29:13
  Speaker 1: of fishing for more information to figure out, to map
• 29:17
  Speaker 1: out the network of critics further. But this was a
• 29:20
  Speaker 1: man who runs a small watchdog nonprofit, and they had
• 29:26
  Speaker 1: been doing a lot of work during that time to
• 29:28
  Speaker 1: try and ask questions about OpenAI's attempt to convert from
• 29:33
  Speaker 1: a nonprofit to a for-profit. Ultimately, OpenAI was successful in
• 29:36
  Speaker 1: that conversion, but during the period where it was sort
• 29:40
  Speaker 1: of existential for OpenAI to complete this conversion, there were
• 29:44
  Speaker 1: a lot of civil society groups and watchdog groups like
• 29:47
  Speaker 1: Midas who were trying to prevent the process from happening
• 29:53
  Speaker 1: in the dead of night. They were trying to get
• 29:55
  Speaker 1: more transparency. They were trying to have more public debate
• 29:59
  Speaker 1: about this because it's unprecedented. And it was then that
• 30:04
  Speaker 1: there was a knock on his door and he was
• 30:06
  Speaker 1: served papers.
• 30:08
  Speaker 2: What did the papers say?
• 30:10
  Speaker 1: The papers asked him to reproduce every single piece of
• 30:13
  Speaker 1: communication that he had had that might have involved Musk.
• 30:17
  Speaker 1: So this was like the strange paranoia that opening I
• 30:19
  Speaker 1: had that Musk was somehow funding these people to block
• 30:23
  Speaker 1: the conversion. None of them were actually funded by Musk.
• 30:26
  Speaker 1: So in this particular case, the request he simply was
• 30:29
  Speaker 1: just answered, you know, I don't have any documents because
• 30:32
  Speaker 1: this doesn't exist.
• 30:33
  Speaker 2: So going back to this point of empires, you're saying
• 30:35
  Speaker 2: that one of the factors of an empire is a
• 30:37
  Speaker 2: land grab. And then the next one was.
• 30:40
  Speaker 3: Was labor exploitation.
• 30:42
  Speaker 2: Labor exploitation.
• 30:42
  Speaker 1: The third one, controlling knowledge production. And one of the
• 30:49
  Speaker 1: other ones that's really important to understand about the AI
• 30:52
  Speaker 1: empires in particular is empires always have this narrative that
• 30:59
  Speaker 1: they say to the public, like, we're the good empire.
• 31:02
  Speaker 1: And we need to be an empire in the first
• 31:04
  Speaker 1: place because there are also bad empires in the world.
• 31:07
  Speaker 1: And if you allow us to take all the resources
• 31:11
  Speaker 1: and use all of the labor, then we promise we
• 31:15
  Speaker 1: will bring you progress and modernity for everyone. We will
• 31:19
  Speaker 1: bring you to this utopic state akin to an AI heaven.
• 31:23
  Speaker 1: But if the evil empire does it first, we will
• 31:26
  Speaker 1: descend into a hell.
• 31:28
  Speaker 2: And the evil empire being in this case?
• 31:30
  Speaker 1: In this case, most often it's China. But actually in
• 31:34
  Speaker 1: the early days, OpenAI evoked Google.
• 31:36
  Speaker 3: As the evil empire.
• 31:38
  Speaker 1: So all of their decisions were about, we need to
• 31:40
  Speaker 1: do it first because otherwise Google, this evil corporation that's
• 31:44
  Speaker 1: driven by profit, us as a benevolent nonprofit, like this
• 31:49
  Speaker 1: is a critical contest of who wins.
• 31:54
  Speaker 2: Do you think the people building these AI companies believe
• 31:58
  Speaker 2: that the outcome is going to be all good now?
• 32:02
  Speaker 2: Do you think they think that it's going to be,
• 32:04
  Speaker 2: it's going to serve everyone. It's going to be the
• 32:05
  Speaker 2: age of abundance. Everything's going to go well. What do
• 32:08
  Speaker 2: you think they believe? So this is.
• 32:11
  Speaker 1: So funny is such a core part of the mythology
• 32:15
  Speaker 1: that they create around the industry includes the belief that
• 32:20
  Speaker 1: it could go very badly.
• 32:22
  Speaker 3: It goes hand in hand.
• 32:24
  Speaker 1: Like they need that part of the myth to in
• 32:27
  Speaker 1: order to then say, and that's why we need to
• 32:30
  Speaker 1: be in control of the technology, because that's the only
• 32:32
  Speaker 1: way that it's going to go really, really well. And
• 32:34
  Speaker 1: Altman has said publicly, you know, the worst case lights
• 32:38
  Speaker 1: out for everyone, but best case, we cure cancer, we
• 32:42
  Speaker 1: solve climate change, and there's abundance. And Dario Amadei, same
• 32:46
  Speaker 1: kind of rhetoric. He's like, worst case, catastrophic or existential
• 32:51
  Speaker 1: harm for humanity. Best case, mass human flourishing. So this
• 32:56
  Speaker 1: is like two sides of the same coin. Like they
• 32:58
  Speaker 1: have to use both of these narratives in order to
• 33:03
  Speaker 1: continue justifying an extremely anti-democratic approach to AI development where
• 33:09
  Speaker 1: there should not be broad participation in developing this technology.
• 33:12
  Speaker 1: They must be the ones controlling it at every step
• 33:15
  Speaker 1: of the way.
• 33:16
  Speaker 2: Sam Altman did a tweet saying, there are some books
• 33:19
  Speaker 2: coming out about OpenAI and me. We only participated in
• 33:22
  Speaker 2: two of them. One by... Kesh Hagee?
• 33:25
  Speaker 3: Keech Hagee.
• 33:26
  Speaker 2: Keech Hagee focused on me and one by Ashley Vance
• 33:29
  Speaker 2: on OpenAI. He went on to say, no book will
• 33:33
  Speaker 2: get everything right, especially when some people are so intent
• 33:36
  Speaker 2: on twisting things. But these two authors are trying to.
• 33:41
  Speaker 2: You quote retweeted that tweet from Sam Altman and you said,
• 33:45
  Speaker 2: the unnamed book Empire of AI is mine. Do you
• 33:51
  Speaker 2: believe that tweet from Sam Altman was in reference to
• 33:54
  Speaker 2: your book?
• 33:54
  Speaker 3: 100%. Because there's only three books coming out about him.
• 33:58
  Speaker 2: And he'd caught wind that your book was coming out?
• 34:00
  Speaker 1: He knew my book was coming out because I had
• 34:02
  Speaker 1: contacted OpenAI from the very beginning of my process and said,
• 34:05
  Speaker 1: I'm working on a book now. Will you participate in it?
• 34:08
  Speaker 1: And actually, initially, they said yes. So my history with OpenAI,
• 34:13
  Speaker 1: I profiled the company for MIT Technology Review. I embedded
• 34:16
  Speaker 1: within the office for three days in 2019. My profile
• 34:20
  Speaker 1: comes out in 2020.
• 34:20
  Speaker 3: The leadership are very unhappy.
• 34:25
  Speaker 1: And in my book, I actually quote an email that
• 34:27
  Speaker 1: I received that Sam Altman sent to the company about
• 34:31
  Speaker 1: my profile saying, yeah, this is not great. And from
• 34:37
  Speaker 1: then on, the company's stance to me was, we are
• 34:44
  Speaker 1: not going to participate in anything that you do. We
• 34:47
  Speaker 1: are not going to respond to anything, any of the
• 34:49
  Speaker 1: questions that you receive. And this was, you know, This
• 34:53
  Speaker 1: was things that they explicitly articulated.
• 34:55
  Speaker 3: It wasn't like me inferring.
• 34:57
  Speaker 1: So I had a colleague at MIT Technology Review that
• 35:01
  Speaker 1: also covered AI. And at one point, OpenAI sent him
• 35:04
  Speaker 1: this press release being like, we would love for you
• 35:06
  Speaker 1: to cover this story. And he was like, I'm really busy.
• 35:08
  Speaker 1: Will you send it to Karen? And they were like, oh, no.
• 35:12
  Speaker 1: We have a history.
• 35:13
  Speaker 3: You understand.
• 35:15
  Speaker 1: And so for three years, they refused to talk to me.
• 35:20
  Speaker 1: But then I ended up at the Wall Street Journal
• 35:23
  Speaker 1: where if they felt a bit compelled because it was
• 35:26
  Speaker 1: the journal to reopen the lines of communication. And so
• 35:31
  Speaker 1: I started having, you know, more dialogue with them. Every
• 35:35
  Speaker 1: time I wrote a piece, I would always send them,
• 35:37
  Speaker 1: here's my request for comment. I would always ask them, like,
• 35:39
  Speaker 1: will you sit for interviews? And we did get to
• 35:43
  Speaker 1: a more productive relationship. And then I embarked on the book.
• 35:46
  Speaker 1: So I left the journal to focus on the book
• 35:48
  Speaker 1: full time. And I told them right away, I'm working
• 35:52
  Speaker 1: on this book. I want to continue this productive conversation
• 35:56
  Speaker 1: where I make sure I reflect OpenAI's perspective in the book.
• 36:02
  Speaker 1: And so they were like, we can arrange interviews for you.
• 36:05
  Speaker 1: You can come back to the office. We'll set up
• 36:08
  Speaker 1: some conversations. And then as we were going back and
• 36:13
  Speaker 1: forth on this, the board fires Sam Altman. And that's
• 36:17
  Speaker 1: when things started going kind of south because the company
• 36:21
  Speaker 1: started becoming very sensitive to scrutiny. And so then they
• 36:25
  Speaker 1: started pushing, kicking the can down the road, down the road,
• 36:27
  Speaker 1: down the road.
• 36:28
  Speaker 3: And I kept saying, hey, when are we rescheduling this?
• 36:30
  Speaker 3: What's going on?
• 36:31
  Speaker 1: And then I get an email saying, we are not
• 36:34
  Speaker 1: going to participate at all. You are not coming to
• 36:36
  Speaker 1: the office. You're not doing interviews. And I had actually
• 36:39
  Speaker 1: already booked my tickets. So I was already going to
• 36:42
  Speaker 1: fly to San Francisco to have the interviews.
• 36:46
  Speaker 3: And so then.
• 36:48
  Speaker 1: I told them, I was like, that's fine. I will
• 36:51
  Speaker 1: still engage in the process where I'll give you extensive
• 36:54
  Speaker 1: requests for comment. As through my reporting, I'll keep you
• 36:58
  Speaker 1: updated on all the things that I'm finding so that
• 37:00
  Speaker 1: you can choose to still comment. I gave them 40
• 37:03
  Speaker 1: pages of requests for comment. And I gave them over
• 37:07
  Speaker 1: a month to respond to all of that. So this
• 37:10
  Speaker 1: was when the tweet came out, was we were doing
• 37:13
  Speaker 1: all this back and forth, trying to, and that's when
• 37:16
  Speaker 1: Altman came.
• 37:17
  Speaker 3: Tweeted this.
• 37:22
  Speaker 1: And they never responded to a single one of the,
• 37:24
  Speaker 1: one of the 40 pages.
• 37:25
  Speaker 2: Sam Watman does a lot of interviews.
• 37:27
  Speaker 1: Yeah.
• 37:28
  Speaker 2: You know, he's doing a lot of interviews all the time.
• 37:29
  Speaker 2: He's done every podcast. I've seen him on everything from
• 37:33
  Speaker 2: Tucker Carlson to, I think he's done Theo Von, Joe Rogan, um,
• 37:37
  Speaker 2: podcasts all over the world. I wonder why he won't
• 37:40
  Speaker 2: do mine.
• 37:45
  Speaker 3: Well, maybe. I don't know why.
• 37:47
  Speaker 2: I don't know. I think I'm fair with everyone. I
• 37:49
  Speaker 2: just ask questions I genuinely care about. I don't come
• 37:51
  Speaker 2: in with huge preconceptions or at least meet people for
• 37:54
  Speaker 2: the first time. But I've heard through the grapevine that
• 37:59
  Speaker 2: he doesn't want to do mine.
• 38:00
  Speaker 1: I mean, going back to what you were saying earlier
• 38:02
  Speaker 1: that with this, the way that OpenAI and these companies
• 38:07
  Speaker 1: control research, you asked, do they also do this with journalists?
• 38:12
  Speaker 3: I mean, yes, the answer is yes.
• 38:14
  Speaker 1: And apparently they also do it with anyone who has,
• 38:17
  Speaker 1: you know, a broad mass communications platform. It's not just
• 38:21
  Speaker 1: about the conversation that you're going to have with them.
• 38:24
  Speaker 1: It's about who you also choose to platform. And there's
• 38:29
  Speaker 1: this huge problem in technology journalism where companies know that
• 38:35
  Speaker 1: a really big carrot that they can give to technology
• 38:37
  Speaker 1: journalists is access.
• 38:39
  Speaker 2: Yeah, yeah, yeah.
• 38:40
  Speaker 3: And they will.
• 38:42
  Speaker 1: withholds that access at the drop of a hat if
• 38:45
  Speaker 1: they catch wind that you're speaking to someone that they
• 38:48
  Speaker 1: didn't want you to speak to.
• 38:49
  Speaker 2: This is so true, and I don't think the average
• 38:51
  Speaker 2: person really truly understands this.
• 38:53
  Speaker 3: Yeah.
• 38:54
  Speaker 2: So this kind of sounds like theory as you say it,
• 38:56
  Speaker 2: but I'm not going to name names here because I
• 38:59
  Speaker 2: don't think it's important. But there is a particular person
• 39:02
  Speaker 2: in AI whose team have basically dangled the carrot of
• 39:08
  Speaker 2: them coming here For like 18 months. And I'm like,
• 39:11
  Speaker 2: you don't have to dangle the carrot. I'm going to
• 39:13
  Speaker 2: speak to whoever I want to, regardless of the carrot
• 39:14
  Speaker 2: or not. And when this person comes, if they want
• 39:16
  Speaker 2: to come, I'll give them a fair shot. I'll ask
• 39:19
  Speaker 2: them all genuinely curious questions about what they're doing, their incentives.
• 39:23
  Speaker 2: I won't gotcha them. I don't have a history of
• 39:24
  Speaker 2: ever gotchering anybody. Even if I just like even if
• 39:27
  Speaker 2: i have a different of opinion i'll ask the question
• 39:29
  Speaker 2: yeah but they dangle carrots and they say well if
• 39:31
  Speaker 2: you know he's thinking about it let's think about a
• 39:34
  Speaker 2: date and what the what the strategy is and i
• 39:35
  Speaker 2: don't think they they think most people don't understand is
• 39:38
  Speaker 2: if we just dangle it for long enough then they
• 39:40
  Speaker 2: will um perform in the way that we want them
• 39:44
  Speaker 2: to do and they'll be they'll be pleasant about us,
• 39:48
  Speaker 2: they won't be critical, they won't give a voice.
• 39:51
  Speaker 3: They won't platform our critics.
• 39:52
  Speaker 2: Our critics. And I think a lot of their game
• 39:55
  Speaker 2: is just dangle the carrot forever. Yes, yeah. That's like
• 39:58
  Speaker 2: the optimal outcome is if we just dangle it, if
• 40:00
  Speaker 2: we just tell them, yeah, no, we're just trying to
• 40:01
  Speaker 2: look at the schedule. It just doesn't work. I think
• 40:04
  Speaker 2: in the modern world, you just have to go there
• 40:05
  Speaker 2: and give your opinion and allow the clash of ideas
• 40:08
  Speaker 2: in the public forum. Let the viewers decide for themselves
• 40:11
  Speaker 2: what they think.
• 40:12
  Speaker 1: Yeah, but this is such a huge part of their machinery,
• 40:16
  Speaker 1: is the way that they use these tactics to massage
• 40:21
  Speaker 1: the public image of these companies and make sure that
• 40:23
  Speaker 1: information that they don't want out, and even opinions that
• 40:26
  Speaker 1: they don't want out there, go out there. And so
• 40:29
  Speaker 1: this is, you know, I feel very lucky now that
• 40:35
  Speaker 1: opening eyes shut the door early on me.
• 40:37
  Speaker 3: Mm-hmm.
• 40:38
  Speaker 1: At the time, I didn't feel lucky. I felt like
• 40:40
  Speaker 1: I had screwed myself over. I was like, should I
• 40:42
  Speaker 1: have been nicer to them in the profile so that
• 40:45
  Speaker 1: I could maintain access? Which is a horrible question to
• 40:47
  Speaker 1: ask as a journalist, right? Like you're supposed to report
• 40:50
  Speaker 1: the truth and you're always supposed to report in the
• 40:53
  Speaker 1: interest of the public. Like that is the point of journalism.
• 40:57
  Speaker 1: And in that moment, I was like relatively junior in
• 41:00
  Speaker 1: my career. I was like, did I misunderstand what journalism
• 41:04
  Speaker 1: is about? Like, Should I have actually been playing.
• 41:08
  Speaker 3: The access game? But it was too late.
• 41:11
  Speaker 1: I had the door shut to me. And so I
• 41:13
  Speaker 1: had to build my career understanding that the front door
• 41:17
  Speaker 1: was never going to be open. And that actually really
• 41:20
  Speaker 1: strengthened my own ability to just tell.
• 41:26
  Speaker 3: It like it is.
• 41:27
  Speaker 1: Yeah. And just report what I see are the facts
• 41:30
  Speaker 1: being presented to me, irrespective of whether the company likes
• 41:33
  Speaker 1: it or not. And most often the company really does
• 41:36
  Speaker 1: not like it, but I can continue to do the work.
• 41:40
  Speaker 1: They don't need to open the front door for me.
• 41:41
  Speaker 1: I was still able to do more than 300 interviews.
• 41:45
  Speaker 2: So Sam Altman gets kicked off the OpenAI executive team.
• 41:55
  Speaker 2: Did you find out why that happened?
• 41:57
  Speaker 1: Yeah.
• 41:58
  Speaker 3: There's a scene-by-scene recounting.
• 42:02
  Speaker 2: From who?
• 42:03
  Speaker 1: I can't remember the exact number of sources, so I
• 42:05
  Speaker 1: don't want to misquote myself, but it was around six
• 42:08
  Speaker 1: or seven people that were directly involved or had spoken
• 42:11
  Speaker 1: to people directly involved in the decision-making process. So Ilya
• 42:18
  Speaker 1: Satskever is seeing these serious concerns about the way that
• 42:25
  Speaker 1: Altman's behavior is leading to bad research outcomes and poor
• 42:32
  Speaker 1: decision-making at the company. He then approaches a board member,
• 42:37
  Speaker 1: Helen Toner.
• 42:38
  Speaker 2: Ilya, for anyone that doesn't know, is the co-founder we
• 42:42
  Speaker 2: mentioned earlier, the co-founder of OpenAI we mentioned earlier. Yes.
• 42:46
  Speaker 1: And he kind of does a bit of a sounding
• 42:50
  Speaker 1: board thing to Helen just because Ilya's freaking out. He's like,
• 42:54
  Speaker 1: he's been like sitting on this, these, these concerns for
• 42:57
  Speaker 1: a while. And he's like, if I tell this to someone.
• 43:01
  Speaker 1: this could also be really bad for me if Altman
• 43:04
  Speaker 1: finds out. And so he asks for a meeting with Toner.
• 43:12
  Speaker 1: And in that first meeting, he's like, like he barely
• 43:16
  Speaker 1: says a thing. He's just like dancing around trying to
• 43:20
  Speaker 1: figure out, hey, is this someone that I can maybe
• 43:23
  Speaker 1: trust to divulge more information?
• 43:25
  Speaker 2: And Toner's role and responsibilities at OpenAI were She was.
• 43:29
  Speaker 3: A board member at the time.
• 43:30
  Speaker 2: Just a board member.
• 43:30
  Speaker 3: Yeah. And specifically an independent board member.
• 43:32
  Speaker 1: So Opening Eye, when it was a nonprofit, the board
• 43:36
  Speaker 1: was split between people who had a stake, financial stake
• 43:39
  Speaker 1: in the company, and then people who were fully independent.
• 43:41
  Speaker 1: And this was meant to be a structure that would
• 43:43
  Speaker 1: balance the decision-making to be in the benefit of the
• 43:47
  Speaker 1: public interest rather than to be in the benefit of
• 43:49
  Speaker 1: the for-profit entity that Opening Eye then created. And Ilya,
• 43:55
  Speaker 1: as a, non-independent board member was approaching Toner as an
• 44:00
  Speaker 1: independent board member to try and see whether or not
• 44:06
  Speaker 1: she was potentially seeing or hearing the same things that
• 44:09
  Speaker 1: he was about the effect that Altman was having on
• 44:12
  Speaker 1: the company.
• 44:12
  Speaker 3: This then sets off a series of conversations, first.
• 44:16
  Speaker 1: Between Ilya and Helen, and then between Amir Muradi and
• 44:22
  Speaker 1: some of the board members. So Amir Muradi was At
• 44:25
  Speaker 1: that point, the chief technology officer of OpenAI, where these
• 44:28
  Speaker 1: two senior leaders essentially through these conversations and through documentation
• 44:32
  Speaker 1: that they're pulling together, like email, Slack messages and so forth,
• 44:36
  Speaker 1: they convey to the independent board members, three independent board members,
• 44:40
  Speaker 1: we are very concerned about Altman's leadership. Like he is
• 44:46
  Speaker 1: creating too much instability at the company. And it is like,
• 44:52
  Speaker 1: he is the root of the problem. It's not, they
• 44:56
  Speaker 1: were trying to say to these independent board members, like
• 45:00
  Speaker 1: the problem will not be fixed unless Altman is removed
• 45:03
  Speaker 1: because of the way that he's pitting teams against each
• 45:07
  Speaker 1: other and creating this environment where people are unable to
• 45:10
  Speaker 1: trust each other anymore. And they're competing rather than collaborating
• 45:14
  Speaker 1: on what's supposed to be this really, really important technology.
• 45:17
  Speaker 2: When you say instability, That's quite a vague term. That
• 45:21
  Speaker 2: could mean lots of things. Instability could mean pushing people
• 45:24
  Speaker 2: hard to work harder. What do you mean by instability
• 45:28
  Speaker 2: in as specific terms as you can possibly say them?
• 45:31
  Speaker 1: When ChatGPT came out in the world, OpenAI was wholly unprepared.
• 45:36
  Speaker 1: They didn't think that they were launching a gangbusters product.
• 45:41
  Speaker 1: They thought they were releasing a research preview that would
• 45:45
  Speaker 1: help them get the data flywheel going, collect a bunch
• 45:48
  Speaker 1: of data from users. That would then inform what they
• 45:52
  Speaker 1: thought would be the gangbusters product, which was a chat
• 45:55
  Speaker 1: bot using GPT-4 and chat GPT was using GPT-3.5. And
• 45:58
  Speaker 1: because of that, there were servers crashing all the time.
• 46:08
  Speaker 1: because they had to scale their infrastructure faster.
• 46:11
  Speaker 3: Than any company in history. And there were all of
• 46:16
  Speaker 3: these outages.
• 46:17
  Speaker 1: They were trying to also hire faster than any company
• 46:19
  Speaker 1: in history to try and have more personnel there. And
• 46:22
  Speaker 1: they were then sometimes hiring people that they were like, actually,
• 46:25
  Speaker 1: we made a mistake. We shouldn't have hired you. So
• 46:27
  Speaker 1: they were firing people left and right. And people were
• 46:30
  Speaker 1: just disappearing off of Slack. And that's how their colleagues
• 46:33
  Speaker 1: would learn that they were no longer at the company anymore.
• 46:35
  Speaker 1: And so it was, yes, like many fast growing companies,
• 46:40
  Speaker 1: a very chaotic environment and a particularly chaotic environment because
• 46:44
  Speaker 1: it was extra fast.
• 46:47
  Speaker 3: Like they had to accelerate more than any other startup.
• 46:52
  Speaker 1: And on top of that, Miramarati and Ilyasetsk ever felt
• 46:57
  Speaker 1: that Altman was making it worse. Like he was not
• 47:00
  Speaker 1: actually effectively ameliorating the circumstances that of the chaos, he
• 47:04
  Speaker 1: was actually sowing more chaos, getting these teams to be
• 47:08
  Speaker 1: more divided. And this is where it's important to understand
• 47:13
  Speaker 1: that the executives and the independent board members, they're all
• 47:19
  Speaker 1: operating under this idea that they're building AGI and that
• 47:22
  Speaker 1: AGI could either be devastating or utopic to humanity. And
• 47:29
  Speaker 1: so it's not, yes, it's like any other company. And no,
• 47:33
  Speaker 1: it's not like any other company. You cannot have, like
• 47:36
  Speaker 1: in their view, you cannot have this degree of chaos
• 47:40
  Speaker 1: as the pressure cooker for creating a technology that they,
• 47:44
  Speaker 1: in their conception, could make or break the world. And
• 47:48
  Speaker 1: so that is basically what the independent board members also
• 47:52
  Speaker 1: begin to reflect on. They have these conversations amongst themselves
• 47:56
  Speaker 1: where they're like, well, based on what we're hearing about
• 48:00
  Speaker 1: Altman's behavior, like if this was an Instacart, would that
• 48:03
  Speaker 1: warrant firing him? And they concluded, maybe not, but this
• 48:08
  Speaker 1: is not Instacart. And that's why they were like, well, crap,
• 48:13
  Speaker 1: maybe this is actually, this does rise to the bar
• 48:17
  Speaker 1: where we should consider replacing him because we are ultimately
• 48:21
  Speaker 1: building a technology that we think could have transformative impacts,
• 48:27
  Speaker 1: either in the positive or negative direction.
• 48:29
  Speaker 3: And so that is what happens.
• 48:31
  Speaker 1: It's like these two executives and then the independent board
• 48:34
  Speaker 1: members also, they were hearing other feedback as well from
• 48:37
  Speaker 1: their connections within the company, with other people in the industry.
• 48:40
  Speaker 1: At one point, Adam D'Angelo, who is one of the
• 48:42
  Speaker 1: independent board members and the CEO of Quora, which is
• 48:46
  Speaker 1: a tech startup in the Valley, he is at a
• 48:50
  Speaker 1: party in San Francisco and he starts to hear some
• 48:53
  Speaker 1: of these rumors that There's something weird about the way
• 48:58
  Speaker 1: that OpenAI has structured its OpenAI startup fund, which was
• 49:03
  Speaker 1: this fund that the company had created to start investing
• 49:06
  Speaker 1: in other startups.
• 49:10
  Speaker 3: And he realizes they'd never really.
• 49:13
  Speaker 1: Seen documentation about how the startup fund had been set
• 49:15
  Speaker 1: up from Altman. And finally, they get the documents and
• 49:18
  Speaker 1: it turns out that OpenAI Startup Fund is not OpenAI's
• 49:21
  Speaker 1: startup fund. It's Altman's startup fund. And this was something
• 49:26
  Speaker 1: like one of several experiences that the independent board members
• 49:30
  Speaker 1: were also having where they're like, there's something not right
• 49:33
  Speaker 1: about the fact that there continuously are inconsistencies between the
• 49:40
  Speaker 1: way that Altman is portraying what is being done versus
• 49:44
  Speaker 1: what is actually being done. And so when these two
• 49:47
  Speaker 1: executives approach the board or the independent board members, then
• 49:51
  Speaker 1: they're like, okay, this lines up with also the experiences
• 49:55
  Speaker 1: that we've been having.
• 49:58
  Speaker 3: And at that point, they then.
• 50:00
  Speaker 1: Have this series of very intense discussions where they're meeting
• 50:04
  Speaker 1: almost every day talking about, should we actually really consider
• 50:09
  Speaker 1: removing Altman?
• 50:12
  Speaker 3: And in the end, they conclude, yes, we should.
• 50:16
  Speaker 1: And if we're going to do it, we need to
• 50:18
  Speaker 1: do it quickly because they were very concerned that the
• 50:21
  Speaker 1: moment that Altman found out his persuasive abilities would make
• 50:25
  Speaker 1: it impossible to do. And so they end up firing
• 50:30
  Speaker 1: Altman without telling anyone, you know, they don't talk to
• 50:34
  Speaker 1: any stakeholders to get them on the same page. Microsoft
• 50:37
  Speaker 1: gets a call right before they execute the action saying
• 50:41
  Speaker 1: we're going to fire Altman.
• 50:42
  Speaker 2: And Microsoft, for anyone that doesn't know, are a lead
• 50:44
  Speaker 2: investor in OpenAI at the time.
• 50:46
  Speaker 1: Yes. One of the only investors in OpenAI at the time.
• 50:52
  Speaker 1: And that is what then devolves the whole thing because
• 50:57
  Speaker 1: Every single person that is affected by this decision is
• 50:59
  Speaker 1: now extremely angry that they were not involved. And that
• 51:04
  Speaker 1: is what then creates this campaign to bring Altman back.
• 51:09
  Speaker 3: And then Altman is reinstalled as CEO days later.
• 51:13
  Speaker 2: This company that I've just invested in is growing like crazy.
• 51:16
  Speaker 2: I wanna be the one to tell you about it
• 51:17
  Speaker 2: because I think it's gonna create such a huge productivity
• 51:19
  Speaker 2: advantage for you. Whisperflow is an app that you can
• 51:21
  Speaker 2: get on your computer and on your phone, on all
• 51:23
  Speaker 2: your devices, and it allows you to speak to your technology.
• 51:26
  Speaker 2: So instead of me writing out an email, I click
• 51:28
  Speaker 2: one button on my phone, and I can just speak
• 51:30
  Speaker 2: the email into existence. And it uses AI to clean
• 51:33
  Speaker 2: up what I was saying. And then when I'm done,
• 51:35
  Speaker 2: I just hit this one button here and the whole
• 51:37
  Speaker 2: email is written for me. And it's saving me so
• 51:40
  Speaker 2: much time in a day because Whisper learns how I write.
• 51:43
  Speaker 2: So on WhatsApp, it knows how I am, a little
• 51:45
  Speaker 2: bit more casual. an email a little bit more professional.
• 51:47
  Speaker 2: And also there's this really interesting thing they've just done.
• 51:49
  Speaker 2: I can create little phrases to automatically do the work
• 51:52
  Speaker 2: for me. I can just say Jack's LinkedIn and it
• 51:54
  Speaker 2: copies Jack's LinkedIn profile for me because it knows who
• 51:57
  Speaker 2: Jack is in my life. This is saving me a
• 51:59
  Speaker 2: huge amount of time. This company is growing like absolute crazy.
• 52:02
  Speaker 2: And this is why I invested in the business and
• 52:03
  Speaker 2: why they're now a sponsor of this show. And Whisperflow
• 52:05
  Speaker 2: is frankly becoming the worst kept secret in business productivity
• 52:09
  Speaker 2: and entrepreneurship. Check it out now at Whisperflow, spelt W-I-S-P-R-F-L-O-W-
• 52:16
  Speaker 2: Dot AI slash Steven. It will be a game changer
• 52:18
  Speaker 2: for you. There's a phase a lot of companies hit
• 52:21
  Speaker 2: where they're no longer doing the most important thing, which
• 52:24
  Speaker 2: is selling, and they get really bogged down with admin.
• 52:26
  Speaker 2: And it's often something that creeps up slowly and you
• 52:29
  Speaker 2: don't really notice until it's happened. Slowly, momentum starts to
• 52:32
  Speaker 2: leak out. This happened to us and our sponsor, Pipedrive,
• 52:35
  Speaker 2: was a fix I came across 10 years ago. And
• 52:37
  Speaker 2: ever since, my teams across my different companies have continued
• 52:39
  Speaker 2: to use it. Pipedrive is a simple but powerful sales
• 52:42
  Speaker 2: CRM that gives you the visibility on any deals in
• 52:45
  Speaker 2: your pipeline. It also automates a lot of the tedious, repetitive,
• 52:48
  Speaker 2: and time-consuming parts of the sales process, which in turn
• 52:51
  Speaker 2: saves you so many hours every single month, which means
• 52:53
  Speaker 2: you can get back to selling. Making that early decision
• 52:56
  Speaker 2: to switch to Pipedrive was a real game changer, and
• 52:58
  Speaker 2: it's kept the right things front of mind. My favorite
• 53:01
  Speaker 2: feature is Pipedrive's ability to sync your CRM with multiple
• 53:04
  Speaker 2: email inboxes so your entire team can work together from
• 53:07
  Speaker 2: one platform. And we aren't the only ones benefiting. Over 100,000
• 53:11
  Speaker 2: companies use Pipedrive to grow their business. So if something
• 53:14
  Speaker 2: I've said resonates, head over to pipedrive.com slash CEO, where
• 53:18
  Speaker 2: you can get a 30-day free trial, no credit card
• 53:22
  Speaker 2: or payment required. How does a CEO of a major
• 53:27
  Speaker 2: company get fired by the board? Because board members, there's
• 53:30
  Speaker 2: a quote in your book on page 357 where you
• 53:32
  Speaker 2: say about Ilya saying, I don't think Sam is the
• 53:34
  Speaker 2: guy who should have the finger on the button for AGI. Now,
• 53:38
  Speaker 2: I asked myself this question. You know, I work with
• 53:40
  Speaker 2: lots of people here. We have 150 people that work
• 53:42
  Speaker 2: in this business. And those people know me best. Yeah.
• 53:48
  Speaker 2: They see me on camera. They see me off camera.
• 53:50
  Speaker 2: So if they said that we don't think Stephen is
• 53:53
  Speaker 2: the right person to host the Diary of a CEO,
• 53:55
  Speaker 2: for example.
• 53:55
  Speaker 1: Yeah.
• 53:56
  Speaker 2: It would take a lot for them to say that.
• 53:58
  Speaker 3: Yeah.
• 53:58
  Speaker 2: They must have seen some shit off camera for them
• 54:01
  Speaker 2: to go, we don't think he's the right person to
• 54:03
  Speaker 2: be on camera.
• 54:04
  Speaker 1: Yeah.
• 54:04
  Speaker 2: Or for whatever reason. And in the case of AI,
• 54:06
  Speaker 2: which is much more consequential than a podcast that is,
• 54:08
  Speaker 2: you know, filmed in my old kitchen, it almost sends
• 54:12
  Speaker 2: a chill down one's body to think that the co-founder
• 54:14
  Speaker 2: of a business has gone to the board and said,
• 54:16
  Speaker 2: this isn't the guy to lead this consequential technology.
• 54:19
  Speaker 3: And it wasn't just Ilya. Miramarati then also said, I
• 54:21
  Speaker 3: don't think Altman is the right guy.
• 54:23
  Speaker 2: And then they both left later.
• 54:26
  Speaker 1: So then Altman comes back and lo and behold, Ilya
• 54:29
  Speaker 1: never comes back. So his concerns about the fact that
• 54:33
  Speaker 1: Altman finding out would be bad for him manifested. He
• 54:36
  Speaker 1: ended up not coming back and Miramarati then left shortly thereafter.
• 54:40
  Speaker 2: Quite a lot of these people leave, don't they? OpenAI.
• 54:44
  Speaker 3: They do.
• 54:45
  Speaker 1: So if you consider one of the origin stories of
• 54:52
  Speaker 1: OpenAI is this dinner that happened at the Rosewood Hotel restaurant.
• 54:57
  Speaker 1: which is a very swanky hotel right in the heart
• 55:00
  Speaker 1: of Silicon Valley that was one of Elon Musk's favorites
• 55:04
  Speaker 1: whenever he was coming up from LA to the Bay Area.
• 55:07
  Speaker 1: And there was this dinner that was there where Altman
• 55:10
  Speaker 1: was intending to recruit the OG team that would start OpenAI.
• 55:15
  Speaker 1: So he's kind of telling everyone, you might have a
• 55:19
  Speaker 1: chance to meet Musk because Musk is going to come
• 55:21
  Speaker 1: to this dinner. And he cold emails Ilya and says,
• 55:25
  Speaker 1: gets Ilya to then come, and Ilya specifically wants to
• 55:28
  Speaker 1: come because he wants to meet Musk. And he also
• 55:31
  Speaker 1: emails all these other people, including Greg Brockman, Dario Amadei.
• 55:35
  Speaker 2: These are all people that ended up working at OpenAI.
• 55:37
  Speaker 1: And they all, almost all of them, not every one
• 55:40
  Speaker 1: of them, but almost all of them end up working
• 55:44
  Speaker 1: at OpenAI.
• 55:45
  Speaker 2: And leaving.
• 55:46
  Speaker 1: Almost all of them end up leaving, specifically after they
• 55:50
  Speaker 1: clash with Altman.
• 55:52
  Speaker 2: And Ilya... he left and launched a company called Safe Superintelligence.
• 55:58
  Speaker 3: Yeah.
• 56:00
  Speaker 2: Which is, I mean, that's an indirect if I've ever
• 56:03
  Speaker 2: heard one.
• 56:03
  Speaker 3: Do you know what I mean? Do you know what
• 56:06
  Speaker 3: I mean?
• 56:06
  Speaker 2: If someone like co-founded this podcast with me and then
• 56:10
  Speaker 2: they left and started a podcast called Safe Podcasting, I'd
• 56:17
  Speaker 2: take that as a slight. I'd have people knocking on
• 56:20
  Speaker 2: their door and asking for their texts.
• 56:24
  Speaker 3: One of the things that is happening here is it.
• 56:30
  Speaker 1: Is not a coincidence that every single tech billionaire has
• 56:33
  Speaker 1: their own AI company. They want to create AI in
• 56:39
  Speaker 1: their own image. And that's why they keep not getting along.
• 56:45
  Speaker 1: And in fact, it's not just don't get along. They
• 56:48
  Speaker 1: end up hating each other after working together. And then
• 56:52
  Speaker 1: splinter off into... their own organizations. So after Musk leaves,
• 56:57
  Speaker 1: he starts XAI. After Darger leaves, he starts Anthropic. After
• 57:00
  Speaker 1: Ilia leaves, he starts Safe Superintelligence. After Mira leaves, she
• 57:05
  Speaker 1: starts Thinking Machines Lab. They want to have control over
• 57:13
  Speaker 1: their own vision of this technology.
• 57:16
  Speaker 3: And the best way that they have.
• 57:20
  Speaker 1: Derived from their experiences of trying to put their vision
• 57:26
  Speaker 1: into the arena is by creating a competitor and then
• 57:29
  Speaker 1: competing with OpenAI and with all the other companies out there.
• 57:33
  Speaker 2: Do you think some of these AICs realize that they
• 57:35
  Speaker 2: are quite literally summoning the demon, as Elon said, 10
• 57:37
  Speaker 2: years ago, but they don't really care because being the
• 57:42
  Speaker 2: person that summoned the demon makes you consequential and powerful
• 57:46
  Speaker 2: and historical, even if the outcome is potentially horrific, even
• 57:51
  Speaker 2: if there's like a 20% outcome of it being horrific.
• 57:53
  Speaker 2: I remember, I think it was Dario, he's the one
• 57:56
  Speaker 2: that said there's somewhere between a 10% and 25% chance
• 58:01
  Speaker 2: of things going catastrophically wrong on the scale of human civilization.
• 58:05
  Speaker 2: 25% is a one in four chance. If you put
• 58:12
  Speaker 2: bullets in a four chamber revolver and said, Stephen, The
• 58:16
  Speaker 2: upside is you could become a multi-gazillionaire and be remembered forever.
• 58:20
  Speaker 2: The downside is that there'd be a bullet in your head.
• 58:22
  Speaker 2: There is no chance that I would take that bet
• 58:26
  Speaker 2: with a 25% potential chance of things going catastrophically wrong.
• 58:31
  Speaker 1: So I have a very long answer to this because
• 58:36
  Speaker 1: do they know if they're summoning the demon? It really
• 58:38
  Speaker 1: depends on what we define as summoning the demon. And
• 58:41
  Speaker 1: in this particular case, to go back to what we
• 58:44
  Speaker 1: were saying before, There's a mythology that the AI industry
• 58:48
  Speaker 1: uses where summoning the demon is an integral part of
• 58:55
  Speaker 1: convincing everyone that therefore they can be the only ones
• 58:59
  Speaker 1: that are developing this technology.
• 59:01
  Speaker 2: I got it. So on one end, you've got to say,
• 59:03
  Speaker 2: if we don't, China will. And that's terrible. Yeah. But
• 59:07
  Speaker 2: if we let anyone else do it other than me,
• 59:08
  Speaker 2: then we're fucked as well.
• 59:10
  Speaker 3: Exactly.
• 59:11
  Speaker 2: So that means that I have to do it and
• 59:12
  Speaker 2: you have to give me money and support. Exactly.
• 59:14
  Speaker 1: So when they're saying these things, we should understand it
• 59:20
  Speaker 1: as not as like a genuine prediction based on what
• 59:22
  Speaker 1: they're seeing. Because first of all, we don't predict the future.
• 59:25
  Speaker 3: We make it.
• 59:27
  Speaker 1: We should understand this as an act of speech to
• 59:29
  Speaker 1: persuade other people into believing that they should cede more power,
• 59:35
  Speaker 1: more resources to these individuals. And so do they know
• 59:39
  Speaker 1: that they're summoning the demon? I mean, they are purposely
• 59:42
  Speaker 1: trying to create this feeling within the public that they are,
• 59:49
  Speaker 1: because it is a crucial part of their power.
• 59:53
  Speaker 3: But do they... If we were to define....
• 59:57
  Speaker 1: Just do they realize that the things that they're doing
• 1:00:00
  Speaker 1: are having already really harmful impacts all around the world
• 1:00:04
  Speaker 1: on vulnerable people, vulnerable communities, vulnerable countries? That's where I'm like,
• 1:00:10
  Speaker 1: maybe yes, maybe no. And they don't really care because
• 1:00:15
  Speaker 1: in the frame of mind, like I sometimes use the
• 1:00:19
  Speaker 1: analogy that the AI.
• 1:00:20
  Speaker 3: World is like Dune.
• 1:00:22
  Speaker 2: Dune, for anyone that doesn't know Dune.
• 1:00:24
  Speaker 3: Science fiction epic written by Frank Herbert. And it's set
• 1:00:27
  Speaker 3: in this intergalactic era where there are all these houses
• 1:00:31
  Speaker 3: and they're fighting each other for spice.
• 1:00:33
  Speaker 1: So it's a callback to colonialism and empire. And they
• 1:00:37
  Speaker 1: all are trying to control the spice. But one of
• 1:00:38
  Speaker 1: the features of this story is that there are these
• 1:00:42
  Speaker 1: myths that are seeded on the different planets about a
• 1:00:47
  Speaker 1: religious myth, basically, about the coming of the Messiah that
• 1:00:49
  Speaker 1: are used as ways to control the people. And Paul
• 1:00:53
  Speaker 1: Atreides says, When he arrives at the planet Arrakis with
• 1:00:59
  Speaker 1: the intention of trying to then fight against the Empire
• 1:01:03
  Speaker 1: and avenge his father's death, he steps into a myth
• 1:01:09
  Speaker 1: that has been seeded on this planet that says that
• 1:01:12
  Speaker 1: one day there will be a messiah that comes and
• 1:01:14
  Speaker 1: saves the planet. So he steps into the role of
• 1:01:17
  Speaker 1: the messiah and leans into this idea in order to
• 1:01:21
  Speaker 1: better control the people and rally them behind him as
• 1:01:25
  Speaker 1: a leader to help with this quest. He knows that
• 1:01:30
  Speaker 1: it's a myth in the beginning, but because he lives
• 1:01:33
  Speaker 1: and breathes and embodies it, it kind of starts to
• 1:01:37
  Speaker 1: blur in his mind whether this is really a myth
• 1:01:39
  Speaker 1: or whether he's really the Messiah. And this is what
• 1:01:43
  Speaker 1: I think happens in the AI world. On one hand,
• 1:01:48
  Speaker 1: there are all these executives that, actively engage in myth-making because,
• 1:01:54
  Speaker 1: you know, I have all these internal documents that I
• 1:01:56
  Speaker 1: write about in the book where they are very keenly
• 1:01:59
  Speaker 1: aware of how to bring the public along with them
• 1:02:03
  Speaker 1: by showing them dazzling demonstrations of the technology, by using,
• 1:02:09
  Speaker 1: crafting a mission that will sound really good and make
• 1:02:13
  Speaker 1: people give more leniency to their companies. So they know
• 1:02:19
  Speaker 1: they're doing the myth making. And also, I think many
• 1:02:22
  Speaker 1: of them lose themselves in the myth because they have
• 1:02:26
  Speaker 1: to live and breathe and embody it day in and
• 1:02:28
  Speaker 1: day out. And so when, you know, Dario says he
• 1:02:32
  Speaker 1: thinks that 10 to 25 percent of the future could
• 1:02:35
  Speaker 1: be catastrophic or whatever, the probability is 10 to 25 percent.
• 1:02:40
  Speaker 1: He is actively engaging in the myth-making, but also he's
• 1:02:43
  Speaker 1: losing himself in the myth. Like, I think if you
• 1:02:46
  Speaker 1: were to ask him, do you genuinely believe that? He
• 1:02:48
  Speaker 1: would be like, yes, I genuinely believe that. Because there's
• 1:02:51
  Speaker 1: been a blurring of when he's saying something just to
• 1:02:55
  Speaker 1: say something versus when he actually believes what he's required
• 1:03:02
  Speaker 1: to believe in order to then continue doing the things
• 1:03:07
  Speaker 1: that he's doing.
• 1:03:09
  Speaker 2: And this is the whole... psychology of cognitive dissonance right
• 1:03:12
  Speaker 2: where you the brain struggles to hold two conflicting world
• 1:03:15
  Speaker 2: views at the same time so it's an it's incentivized
• 1:03:18
  Speaker 2: or it endeavors to dismiss one so if you you
• 1:03:20
  Speaker 2: know if you wanted to be a healthy person but
• 1:03:22
  Speaker 2: also a smoker um and i pointed out smoking's bad
• 1:03:25
  Speaker 2: for you the first words out of your mouth are
• 1:03:26
  Speaker 2: going to be yes but yeah smoking helps me with
• 1:03:30
  Speaker 2: stress yeah yes but i only do it when and
• 1:03:33
  Speaker 2: i think i don't know i i kind of see
• 1:03:34
  Speaker 2: that at the moment because these companies have to raise
• 1:03:37
  Speaker 2: extortionate like huge amounts of money to fund their AI
• 1:03:41
  Speaker 2: research and they're building out all of these data centers.
• 1:03:44
  Speaker 2: So when they're out in the public, they're always fundraising.
• 1:03:47
  Speaker 2: All of these major companies are fundraising all the time
• 1:03:49
  Speaker 2: at the moment. So you can't be fundraising and saying,
• 1:03:51
  Speaker 2: I'm going to destroy your children's future. Potentially there's 25%
• 1:03:53
  Speaker 2: chance that your children aren't going to have a great life,
• 1:03:58
  Speaker 2: which might be the truth.
• 1:03:59
  Speaker 1: I mean, that is actually what they say. This is
• 1:04:01
  Speaker 1: what famously Dario Amede does.
• 1:04:03
  Speaker 2: He does that, but the others, Sam's not doing that
• 1:04:05
  Speaker 2: as much anymore.
• 1:04:06
  Speaker 4: Yes.
• 1:04:07
  Speaker 1: And it's because, you know, it goes back to like
• 1:04:11
  Speaker 1: each of them kind of distinguish themselves a little bit
• 1:04:13
  Speaker 1: as the brand that they need to project.
• 1:04:17
  Speaker 2: Do you think any of them are more, have a
• 1:04:20
  Speaker 2: stronger moral compass than others? Because I think Dario often
• 1:04:23
  Speaker 2: gets the credit for having more of a, you know,
• 1:04:25
  Speaker 2: more of a backbone and being more conscious of implications.
• 1:04:31
  Speaker 3: He does get a lot of credit for that.
• 1:04:33
  Speaker 2: He's from Claude and Anthropic for anyone that doesn't know.
• 1:04:37
  Speaker 1: I don't think, It truly matters, that question, the answer
• 1:04:42
  Speaker 1: to that question, because to me, even if you were
• 1:04:45
  Speaker 1: to swap all the CEOs for someone that people would
• 1:04:48
  Speaker 1: say is better at running these companies, it doesn't fix
• 1:04:53
  Speaker 1: the problem that I identify in the book, which is
• 1:04:56
  Speaker 1: that there is a system of power that has been
• 1:04:58
  Speaker 1: constructed where these companies and the people running these companies
• 1:05:02
  Speaker 1: get to make decisions that affect billions of people's lives
• 1:05:04
  Speaker 1: around the world. And those billions of people do not
• 1:05:07
  Speaker 1: get any say in how it goes.
• 1:05:10
  Speaker 2: Those people, they can go to the polls, right? So
• 1:05:14
  Speaker 2: if the public are sufficiently educated, they can go to
• 1:05:16
  Speaker 2: the polls and pick a leader that says they're going
• 1:05:19
  Speaker 2: to legislate or pass laws or try and pass laws. Yes.
• 1:05:23
  Speaker 1: But at the speed and pace at which these companies
• 1:05:26
  Speaker 1: operate and at the sheer scale and size, they're able
• 1:05:30
  Speaker 1: to also spend extraordinary amounts of money, hundreds of millions
• 1:05:33
  Speaker 1: in this upcoming midterms, to try and kill every possible
• 1:05:37
  Speaker 1: piece of legislation that gets in their way and craft
• 1:05:39
  Speaker 1: legislation that would codify their advantage. And so to me,
• 1:05:45
  Speaker 1: I think sometimes as a society, we obsess a little
• 1:05:47
  Speaker 1: bit with, are these leaders good or bad people? And
• 1:05:53
  Speaker 1: to me, the bigger question is, is the governance structure
• 1:05:57
  Speaker 1: that we've created a sound one that allows broad participation
• 1:06:02
  Speaker 1: or an anti-democratic one that has consolidated this decision-making power
• 1:06:06
  Speaker 1: in the hands of the few. Because no person is perfect.
• 1:06:10
  Speaker 1: I don't care who is on the top of these companies.
• 1:06:13
  Speaker 1: They're not going to have the ability to make decisions
• 1:06:17
  Speaker 1: on behalf of so many people around the world who
• 1:06:20
  Speaker 1: live and talk and have a culture and history that
• 1:06:25
  Speaker 1: are fundamentally different from them without things going wrong. And
• 1:06:29
  Speaker 1: so that is why throughout history, we've moved from empires
• 1:06:34
  Speaker 1: to democracy. It's because empire as a structure is inherently unsound.
• 1:06:41
  Speaker 1: It does not actually maximize the chances of most people
• 1:06:45
  Speaker 1: in the world being able to live dignified lives.
• 1:06:49
  Speaker 2: I'm going to try and take on that point of view.
• 1:06:51
  Speaker 2: So this is me playing devil's advocate. Okay. But Karen, if...
• 1:06:57
  Speaker 2: the US don't continue to accelerate their research with AI.
• 1:07:01
  Speaker 2: At some point, China's model is going to become so
• 1:07:05
  Speaker 2: smart and intelligent that we're basically going to have to
• 1:07:07
  Speaker 2: rent it off them. And we're going to be, you know,
• 1:07:09
  Speaker 2: they'll get the scientific discoveries. They'll discover the new era
• 1:07:12
  Speaker 2: of autonomous weapons. And we will be their backyard. And like, logically,
• 1:07:19
  Speaker 2: that argument does appear to be pretty true.
• 1:07:22
  Speaker 3: No, it's not.
• 1:07:23
  Speaker 2: If we scale up, if we just imagine any rate
• 1:07:25
  Speaker 2: of change with this intelligence, At some point, we're going
• 1:07:28
  Speaker 2: to come to a weapon that could theoretically disable all
• 1:07:33
  Speaker 2: of the United States' electricity, their weapons systems. It would
• 1:07:36
  Speaker 2: know exactly how to disable the United States from a
• 1:07:39
  Speaker 2: cyber perspective because it would be that smart. All you've
• 1:07:42
  Speaker 2: got to imagine is any rate of improvement over any
• 1:07:45
  Speaker 2: sort of a long period of time. So this is
• 1:07:49
  Speaker 2: a theory that might be true. And if it's true.
• 1:07:52
  Speaker 3: I mean, yeah, any theory might be true.
• 1:07:56
  Speaker 2: But, you know, again, going to this point of like,
• 1:07:58
  Speaker 2: even if it's a small percentage, it's worth paying attention
• 1:08:00
  Speaker 2: to on the other side of the foot. This is
• 1:08:02
  Speaker 2: a theory that people talk about. It could be the
• 1:08:05
  Speaker 2: case that the most intelligent civilization is going to be
• 1:08:09
  Speaker 2: the superior civilization. Logically, that's a pretty sound thing to.
• 1:08:13
  Speaker 1: Say, no?
• 1:08:14
  Speaker 3: So there's a lot of.
• 1:08:17
  Speaker 1: Fundamentals in this argument that would need to be true
• 1:08:20
  Speaker 1: in order for this to be a viable argument and
• 1:08:22
  Speaker 1: let's knock them down one by one. So the first
• 1:08:25
  Speaker 1: one is that these systems are intelligent and that just
• 1:08:31
  Speaker 1: scaling them is going to bring us more intelligence.
• 1:08:34
  Speaker 2: So far, so true.
• 1:08:35
  Speaker 1: No, it's actually not because First of all, again, we
• 1:08:40
  Speaker 1: don't actually know if these systems are, like, intelligence is not,
• 1:08:45
  Speaker 1: it's not like the right analogy, almost. It's sort of like,
• 1:08:50
  Speaker 1: it's like, is a calculator, a calculator can do math
• 1:08:52
  Speaker 1: problems faster than a human.
• 1:08:54
  Speaker 3: Does that make it intelligent?
• 1:08:56
  Speaker 2: It has a narrow intelligence because it's solving a narrow problem,
• 1:08:58
  Speaker 2: which is like one plus one equals two.
• 1:09:01
  Speaker 1: And these systems, they actually also are quite narrowly intelligent
• 1:09:05
  Speaker 1: in the sense that even though these companies say that
• 1:09:08
  Speaker 1: they're everything machines that can do anything for anyone... they
• 1:09:11
  Speaker 1: actually can only do some things for some people. This
• 1:09:13
  Speaker 1: is like the jagged frontier of these AI models. Some
• 1:09:17
  Speaker 1: of the capabilities are quite good. Other capabilities are not
• 1:09:19
  Speaker 1: that good. You know why that happens? It's because the
• 1:09:22
  Speaker 1: company can only focus on advancing certain types of capabilities.
• 1:09:25
  Speaker 1: It can't literally focus on advancing all types of capabilities.
• 1:09:29
  Speaker 1: They have to actually set their mind to advancing a
• 1:09:31
  Speaker 1: certain by gathering the data that is needed for that capability,
• 1:09:34
  Speaker 1: by getting a bunch of human contractors to annotate and
• 1:09:40
  Speaker 1: train the model to do that exact thing. And so
• 1:09:45
  Speaker 1: scaling these models is actually a perpendicular question to, are
• 1:09:51
  Speaker 1: we actually getting more cyber capabilities specifically and more military
• 1:09:57
  Speaker 1: capabilities specifically?
• 1:09:58
  Speaker 2: I would argue that most of the top people in
• 1:10:01
  Speaker 2: AI believe that the intelligence is gonna continue to scale
• 1:10:04
  Speaker 2: for some time. A lot of them do, like Geoffrey
• 1:10:06
  Speaker 2: Hinton does.
• 1:10:07
  Speaker 1: And again, it's back to his hypothesis about how human
• 1:10:11
  Speaker 1: intelligence works and what the appropriate model of the brain is.
• 1:10:15
  Speaker 1: His hypothesis throughout his career has been the brain is
• 1:10:19
  Speaker 1: a statistical engine. But that's his hypothesis, and that is
• 1:10:22
  Speaker 1: not universally agreed upon, especially among people that are not
• 1:10:26
  Speaker 1: in the AI world. When you talk with neuroscientists and psychologists,
• 1:10:29
  Speaker 1: people who actually study human intelligence in the human brain—
• 1:10:32
  Speaker 1: That is where you start to get a lot of
• 1:10:34
  Speaker 1: debate and disagreement about this particular view that Hinton has.
• 1:10:40
  Speaker 1: And so this is kind of like one.
• 1:10:43
  Speaker 3: Of the things.
• 1:10:44
  Speaker 1: It's like AI is already being used in the military
• 1:10:48
  Speaker 1: and has been used in the military for a long time.
• 1:10:51
  Speaker 1: But specifically accelerating large language models isn't just the only...
• 1:11:00
  Speaker 1: path for getting military capability, like the companies would have
• 1:11:02
  Speaker 1: to choose to specifically pick military capabilities to accelerate, not
• 1:11:08
  Speaker 1: just like general intelligence.
• 1:11:09
  Speaker 3: It's like, you know what I'm saying?
• 1:11:11
  Speaker 1: Like, They create this myth that they are actually pushing
• 1:11:15
  Speaker 1: the frontier of all of the capabilities of the model,
• 1:11:17
  Speaker 1: but that's not what's actually happening internally. And I have,
• 1:11:20
  Speaker 1: I had hundreds of pages of documents on like how
• 1:11:23
  Speaker 1: they were specifically training models. They pick what capabilities they
• 1:11:27
  Speaker 1: want to advance and you know how they pick them.
• 1:11:29
  Speaker 1: It's based on which industries would be able to pay
• 1:11:32
  Speaker 1: them the most money for their services. So they pick finance, law, medicine, healthcare, commerce,
• 1:11:41
  Speaker 1: It's not actually intelligent like a baby, where the more
• 1:11:47
  Speaker 1: that the baby grows up, they start having these general abilities.
• 1:11:52
  Speaker 2: I think I have jagged intelligence, obviously. I wasn't going
• 1:11:55
  Speaker 2: to say it, but... I think I know a little
• 1:11:59
  Speaker 2: bit about... No, I know a lot about a little bit.
• 1:12:02
  Speaker 1: Yeah, but you also have the capability to learn and
• 1:12:05
  Speaker 1: acquire knowledge by yourself. And you also have the ability
• 1:12:07
  Speaker 1: to choose what you're going to learn and acquire by yourself.
• 1:12:10
  Speaker 2: It's not easy, and it takes a lot more time
• 1:12:11
  Speaker 2: than these models, it seems. Less compute.
• 1:12:14
  Speaker 1: And you can learn how to drive in one place
• 1:12:16
  Speaker 1: and then immediately know how to drive in another place.
• 1:12:18
  Speaker 3: These models cannot do that. Every time a self-driving car
• 1:12:22
  Speaker 3: is shifted to another location, it has to completely retrain
• 1:12:26
  Speaker 3: on that location.
• 1:12:27
  Speaker 1: It's like all the self-driving cars. I mean, we're sitting
• 1:12:29
  Speaker 1: in Austin right now, and there's all these self-driving cars
• 1:12:31
  Speaker 1: that are driving through Austin.
• 1:12:34
  Speaker 2: But when one of them learns, they all learn.
• 1:12:37
  Speaker 1: Well, it's just because it's... an operating system that has
• 1:12:43
  Speaker 1: an AI model as part of it, and you're training
• 1:12:44
  Speaker 1: the AI model, and then you deploy the AI model
• 1:12:47
  Speaker 1: across all the self-driving cars.
• 1:12:48
  Speaker 2: Which is a big advantage. Because if one optimist robot
• 1:12:52
  Speaker 2: learns one thing in one factory, they all learn it.
• 1:12:55
  Speaker 2: And imagine that. Imagine if humans, if we all learned
• 1:12:57
  Speaker 2: what all the other humans learned. That would give us
• 1:13:00
  Speaker 2: such an unbelievable competitive advantage. I mean, one of the
• 1:13:02
  Speaker 2: ways we did that is through communication.
• 1:13:03
  Speaker 3: Or it could not, because they could be learning the
• 1:13:05
  Speaker 3: wrong thing, which has also happened again and again with
• 1:13:08
  Speaker 3: these technologies, is that all of them then learn the
• 1:13:10
  Speaker 3: wrong thing and they all have the same failure mode.
• 1:13:13
  Speaker 3: I mean, part of the resilience of human society is
• 1:13:15
  Speaker 3: that we do have different expertises and we also have
• 1:13:17
  Speaker 3: different failure modes.
• 1:13:19
  Speaker 2: I think sometimes we hold AI models to a higher
• 1:13:21
  Speaker 2: standard than we hold humans to. And in a weird way,
• 1:13:24
  Speaker 2: because I'd hear on stage, we're in Austin at the moment,
• 1:13:26
  Speaker 2: and I'd hear people go, ah, but you know, them
• 1:13:29
  Speaker 2: AI models, they hallucinate sometimes. I'm like, have you met
• 1:13:32
  Speaker 2: a human? Like, I hallucinate all the time. I can
• 1:13:36
  Speaker 2: barely spell or do math.
• 1:13:39
  Speaker 1: Yes, but it's once again, like using this analogy that
• 1:13:43
  Speaker 1: was specifically picked in the early days of the field
• 1:13:45
  Speaker 1: as a way to market these technologies. Like we're repeatedly
• 1:13:49
  Speaker 1: using the intelligence analogy and relating these machines to human
• 1:13:53
  Speaker 1: intelligence as a way to try and gauge whether or
• 1:13:57
  Speaker 1: not it is good or worthy or capable in society.
• 1:14:01
  Speaker 2: I think the output is the thing that really matters,
• 1:14:03
  Speaker 2: is the most consequential, which is like, okay, it might
• 1:14:05
  Speaker 2: have a different brain and a different system, but does
• 1:14:07
  Speaker 2: it arrive at the same capability? Like, is it able
• 1:14:10
  Speaker 2: to do surgery on someone's brain? Is it able to
• 1:14:12
  Speaker 2: drive a car? Like my car drives itself in Los Angeles.
• 1:14:15
  Speaker 2: I don't touch a steering wheel and I can drive
• 1:14:17
  Speaker 2: for many, many hours. And in here in Austin, I
• 1:14:19
  Speaker 2: just saw the ones the other day where they've removed
• 1:14:21
  Speaker 2: the steering wheel and the pedals, the new cyber cabs.
• 1:14:24
  Speaker 2: So I go, it doesn't really matter if it's using
• 1:14:25
  Speaker 2: a different system. If it's navigating through the world as
• 1:14:27
  Speaker 2: a car, has a better safety record than human beings,
• 1:14:31
  Speaker 2: then as far as I'm concerned, intelligence or not, it's like,
• 1:14:36
  Speaker 2: you know.
• 1:14:36
  Speaker 1: Yes, but that was not the original argument that you made,
• 1:14:38
  Speaker 1: which was like, these systems are just generally going to
• 1:14:41
  Speaker 1: become more intelligent across different things based on the prediction.
• 1:14:46
  Speaker 1: This is a prediction that you're making, right? And this
• 1:14:48
  Speaker 1: is a prediction that all of the AI.
• 1:14:50
  Speaker 2: Ilya's making, Dario's making, Elon's making, Zuckerberg's making, Altman's making,
• 1:14:55
  Speaker 2: Demis is making.
• 1:14:56
  Speaker 1: And do you know what the common feature of all
• 1:14:57
  Speaker 1: of them is? they profit enormously off of this myth.
• 1:15:01
  Speaker 2: Elon has recently spearheaded the construction of Colossus, a massive
• 1:15:05
  Speaker 2: supercomputer in Memphis, housing 100,000 GPUs, specifically to scale up
• 1:15:10
  Speaker 2: their GROK API models faster than their competitors. It appears
• 1:15:13
  Speaker 2: that they've all converged around this idea that you can
• 1:15:16
  Speaker 2: brute force your way to greater, more generalized intelligence.
• 1:15:20
  Speaker 1: They've converged around the idea that you can brute force
• 1:15:23
  Speaker 1: your way into models that they can sell to people
• 1:15:26
  Speaker 1: for automating certain tasks that are that are financially lucrative.
• 1:15:30
  Speaker 2: And I heard Elon say that if you're a surgeon,
• 1:15:32
  Speaker 2: now there's just no point. He was like, don't train
• 1:15:34
  Speaker 2: to be a surgeon. He says in a couple of
• 1:15:36
  Speaker 2: years time, optimists and AI generally are going to be
• 1:15:39
  Speaker 2: better than any surgeon that's ever lived.
• 1:15:40
  Speaker 1: Yeah.
• 1:15:41
  Speaker 2: Do you think these things are true?
• 1:15:42
  Speaker 1: Well, you know, I'm pretty sure it was Hinton that
• 1:15:45
  Speaker 1: famously slash infamously said there would be no need for
• 1:15:48
  Speaker 1: radiologists anymore. There would be no need for radiologists anymore.
• 1:15:52
  Speaker 1: And he set a deadline that we've already passed. I
• 1:15:55
  Speaker 1: don't remember how many years. Radiology is doing great as
• 1:15:59
  Speaker 1: a profession.
• 1:16:01
  Speaker 2: Do you think it will be in five years?
• 1:16:03
  Speaker 3: Okay.
• 1:16:03
  Speaker 1: So this, this once again goes back to this question
• 1:16:05
  Speaker 1: of like, why do we build technology and why should
• 1:16:08
  Speaker 1: we specifically be building AI?
• 1:16:10
  Speaker 3: Okay.
• 1:16:10
  Speaker 1: And for me, like the whole project of technology development
• 1:16:15
  Speaker 1: advancement is not to advance technology for technology's sake.
• 1:16:18
  Speaker 3: It's to help people. And there've been.
• 1:16:22
  Speaker 1: Lots of research that has shown that actually the best
• 1:16:26
  Speaker 1: outcomes for people in a healthcare setting are is for
• 1:16:29
  Speaker 1: the radiologist to have the AI model in their hands
• 1:16:36
  Speaker 1: and for the human expert to use the AI model
• 1:16:41
  Speaker 1: as a tool, as an input into their judgment. And
• 1:16:45
  Speaker 1: it is that combination that leads to the most accurate
• 1:16:49
  Speaker 1: and early diagnoses of certain types of cancer that then
• 1:16:53
  Speaker 1: help improve the prognosis of the patient.
• 1:16:55
  Speaker 2: Do you believe that in the coming years, all the cars,
• 1:16:58
  Speaker 2: pretty much all the cars on the road will be
• 1:16:59
  Speaker 2: driving themselves? No. You don't think so?
• 1:17:02
  Speaker 4: How come?
• 1:17:03
  Speaker 3: Because of the way the technology works.
• 1:17:06
  Speaker 1: Because these are statistical, I mean, currently the way that
• 1:17:11
  Speaker 1: AI models are primarily developed, they're statistical engines. You have
• 1:17:15
  Speaker 1: what's called a neural network, which is a piece of
• 1:17:17
  Speaker 1: software that has a bunch of densely connected nodes.
• 1:17:22
  Speaker 2: Like parameters, is this what they call parameters?
• 1:17:24
  Speaker 1: Yes.
• 1:17:24
  Speaker 3: Yeah, pretty much.
• 1:17:25
  Speaker 1: And you're just pumping a bunch of data into it,
• 1:17:28
  Speaker 1: and then it's analyzing the data and creating this, all
• 1:17:32
  Speaker 1: of these, finding all these correlations in the data, finding
• 1:17:34
  Speaker 1: all these patterns. And then it's through those patterns that
• 1:17:38
  Speaker 1: the machine is then able to act autonomously, right? And
• 1:17:42
  Speaker 1: so the way that they're training a self-driving car is
• 1:17:44
  Speaker 1: they're recording all this footage, and then they have tens
• 1:17:48
  Speaker 1: of thousands or hundreds of thousands of human contractors.
• 1:17:51
  Speaker 3: That draw literally.
• 1:17:54
  Speaker 1: around every single vehicle in the footage, every single pedestrian,
• 1:18:00
  Speaker 1: every single traffic light, every single lane marking, and label
• 1:18:03
  Speaker 1: it exactly as such so that then it's fed into
• 1:18:07
  Speaker 1: an AI model that can identify all of these different components.
• 1:18:11
  Speaker 1: And then it's connected to another piece of software that
• 1:18:15
  Speaker 1: is not AI that's saying, okay, if the AI model
• 1:18:18
  Speaker 1: recognizes the pedestrian, we do not run over the pedestrian. If.
• 1:18:23
  Speaker 3: The AI model recognizes a red traffic light. We stop.
• 1:18:27
  Speaker 1: And so the thing about statistical engines is that it's
• 1:18:32
  Speaker 1: based on probabilities. It's not based on deterministic logic. So
• 1:18:39
  Speaker 1: systems make errors all the time.
• 1:18:41
  Speaker 3: And it's impossible.
• 1:18:42
  Speaker 1: It is technically impossible to get them to stop making errors.
• 1:18:48
  Speaker 2: Humans make errors way more than humans. systems in this case.
• 1:18:52
  Speaker 1: Yeah.
• 1:18:52
  Speaker 2: Like the safety record is like, isn't it like 10
• 1:18:54
  Speaker 2: times more safe to be driven in a Tesla with
• 1:18:57
  Speaker 2: autonomous driving than it is for a human to.
• 1:18:59
  Speaker 3: Drive per mile? It depends on the place. It depends
• 1:19:01
  Speaker 3: on whether the Tesla was trained to specifically navigate the
• 1:19:04
  Speaker 3: place that you're driving.
• 1:19:05
  Speaker 2: Humans get drunk.
• 1:19:06
  Speaker 3: Because if it's in Mumbai....
• 1:19:09
  Speaker 1: In someplace in Vietnam, no, it would not be safer.
• 1:19:13
  Speaker 1: I would much rather be driven by someone that has
• 1:19:17
  Speaker 1: been driving.
• 1:19:18
  Speaker 3: In that place their whole life.
• 1:19:20
  Speaker 1: I'm not arguing against the fact that in certain places
• 1:19:23
  Speaker 1: where the car has been explicitly trained to drive in
• 1:19:26
  Speaker 1: this place, that it has a better safety record than
• 1:19:29
  Speaker 1: the humans that are driving in that place. But you
• 1:19:30
  Speaker 1: specifically asked if I think that all of the- Most cars.
• 1:19:35
  Speaker 1: Most cars in the world, in the US?
• 1:19:37
  Speaker 2: Less in the United States, because we're here.
• 1:19:40
  Speaker 3: I don't actually think that it's like imminently on the horizon.
• 1:19:43
  Speaker 2: 10 years?
• 1:19:44
  Speaker 3: No, I don't think so.
• 1:19:45
  Speaker 2: I sat with Dara from Uber and he's pretty convinced
• 1:19:47
  Speaker 2: that his 9 million couriers will be replaced by autonomous vehicles.
• 1:19:51
  Speaker 1: I mean, how long has self-driving cars been invested in?
• 1:19:56
  Speaker 1: It's been more than 10 years. And what percentage of
• 1:19:59
  Speaker 1: cars right now are autonomous.
• 1:20:03
  Speaker 3: On the US roads? I mean, so part of it
• 1:20:05
  Speaker 3: is it's actually not a technical problem, right? Like, Part
• 1:20:08
  Speaker 3: of it is also a social problem. Like, do people
• 1:20:10
  Speaker 3: even trust getting into these e-calls?
• 1:20:12
  Speaker 1: Part of it's also a legal problem, which is if
• 1:20:16
  Speaker 1: the self-driving car kills someone, which it has happened.
• 1:20:20
  Speaker 2: Yeah, it has happened.
• 1:20:21
  Speaker 3: Who is responsible?
• 1:20:24
  Speaker 2: So in the case in LA, it was both Tesla
• 1:20:26
  Speaker 2: and the driver because the driver dropped their phone. They
• 1:20:29
  Speaker 2: looked down and this was a couple of years ago,
• 1:20:31
  Speaker 2: I believe. And they went to grab their phone and
• 1:20:33
  Speaker 2: they hit someone. And so it went to court and
• 1:20:36
  Speaker 2: they were held, both responsible, both the driver and Tesla.
• 1:20:40
  Speaker 2: In terms of Tesla, pretty much everyone that gets the car,
• 1:20:46
  Speaker 2: it comes with autonomy now for pretty much most people,
• 1:20:48
  Speaker 2: I believe.
• 1:20:49
  Speaker 3: Partial autonomy.
• 1:20:49
  Speaker 2: Yeah, it's called full self-driving at the moment.
• 1:20:51
  Speaker 3: I mean, yes, it is called full self-driving.
• 1:20:54
  Speaker 2: Full self-driving supervised, where you kind of have to be
• 1:20:57
  Speaker 2: looking in the right direction.
• 1:20:59
  Speaker 3: Yeah, so it's partial autonomy.
• 1:21:02
  Speaker 2: And here in Austin, it's full autonomy because there's no
• 1:21:04
  Speaker 2: steering wheel.
• 1:21:05
  Speaker 3: Yeah.
• 1:21:06
  Speaker 2: On the new car, so you can't drive it anyway.
• 1:21:08
  Speaker 2: But it is, you know, the Model Y is the
• 1:21:10
  Speaker 2: undisputed highest selling car, best selling car in the world
• 1:21:14
  Speaker 2: across all brands. But I guess my point here is
• 1:21:18
  Speaker 2: like these predictions where they say AI is going to
• 1:21:21
  Speaker 2: completely change transportation and driving. It's going to completely change.
• 1:21:25
  Speaker 2: Lawyers aren't going to have jobs. Accountants aren't going to
• 1:21:27
  Speaker 2: have jobs. Do you believe that they are true? Do
• 1:21:30
  Speaker 2: you believe that there's going to be mass job displacement?
• 1:21:33
  Speaker 1: Okay, so I do think that there is going to
• 1:21:35
  Speaker 1: be huge impacts on employment and we are already seeing
• 1:21:37
  Speaker 1: those impacts. It is not simply because the AI models
• 1:21:42
  Speaker 1: are just automating those jobs away. It is specifically because
• 1:21:47
  Speaker 1: the models are improving in certain capabilities based on what
• 1:21:51
  Speaker 1: the companies that are developing them choose to improve them on.
• 1:21:55
  Speaker 1: And executives at other companies are then deciding to fire
• 1:21:59
  Speaker 1: or lay off their workers Because they think that AI
• 1:22:03
  Speaker 1: can replace the worker, irrespective of whether that might be true.
• 1:22:07
  Speaker 1: And there have been cases of the Klarna CEO who
• 1:22:10
  Speaker 1: laid off a bunch of people thinking that he would
• 1:22:12
  Speaker 1: replace everyone with AI, and then it didn't actually work,
• 1:22:14
  Speaker 1: and he had to ask some people to come back.
• 1:22:16
  Speaker 2: I actually DM'd him about this. If you're hearing this,
• 1:22:18
  Speaker 2: this is because I've DM'd Sebastian, and he's fine with
• 1:22:21
  Speaker 2: me sharing this. He said, because I've heard his name
• 1:22:24
  Speaker 2: mentioned a lot. And so when we talked about AI
• 1:22:26
  Speaker 2: in the past and people mentioned Sebastian and Klarna as
• 1:22:29
  Speaker 2: the example, I wanted to clarify with him what the
• 1:22:31
  Speaker 2: truth was. He said, it's great to hear from you.
• 1:22:33
  Speaker 2: I think sometimes people struggle with two things can be
• 1:22:36
  Speaker 2: true at the same time. I think it might be
• 1:22:38
  Speaker 2: time to come back on your podcast. To your point,
• 1:22:41
  Speaker 2: this is the media misinterpreting my tweet. We are doubling
• 1:22:44
  Speaker 2: down on AI more than ever. Klarna is shrinking with
• 1:22:47
  Speaker 2: almost 100 employees per month due to AI. We used
• 1:22:51
  Speaker 2: to be 7,400 at the peak. A year ago, 5,500. Now,
• 1:22:56
  Speaker 2: we're 3,300. And by the end of summer, so this
• 1:22:59
  Speaker 2: was last year, we'll be 3,000 people. AI handles 70%
• 1:23:05
  Speaker 2: of our customer service conversations at this moment. This is
• 1:23:08
  Speaker 2: because we have realized that with AI, the production cost
• 1:23:11
  Speaker 2: of software comes down to almost zero. Just like manufacturing
• 1:23:14
  Speaker 2: used to be all handcrafted, and then the machines came,
• 1:23:17
  Speaker 2: code used to be all handcrafted up until a few
• 1:23:19
  Speaker 2: years ago. And now it is machine-produced. And ultimately we
• 1:23:24
  Speaker 2: pay people more than ever for the unique handcrafted man-made stuff.
• 1:23:29
  Speaker 2: Klarna is a bank. People will want to connect to humans,
• 1:23:32
  Speaker 2: not only machines. They want us to be personable, relatable,
• 1:23:36
  Speaker 2: even flawed. So we need to make sure while we
• 1:23:39
  Speaker 2: are automating, replacing with AI in parallel, we make sure
• 1:23:43
  Speaker 2: we offer a super available human experience.
• 1:23:47
  Speaker 1: I'm really glad you read this because I think it
• 1:23:49
  Speaker 1: touches on some really important nuances to Yeah, like the
• 1:23:55
  Speaker 1: impact that AI is going to have on employment. So
• 1:23:58
  Speaker 1: I think there's often these binary narratives. It's like AI
• 1:24:02
  Speaker 1: is going to come for every job. Or people say
• 1:24:06
  Speaker 1: AI is not actually working and it's not actually coming
• 1:24:08
  Speaker 1: for jobs. And the reality is that it's coming for jobs.
• 1:24:12
  Speaker 1: There are definitely jobs that are being automated away because
• 1:24:16
  Speaker 1: of the capabilities of their models. And there's also jobs
• 1:24:19
  Speaker 1: that are being lost because executives are deciding to lay
• 1:24:21
  Speaker 1: off the workers, even if the models don't match the
• 1:24:24
  Speaker 1: capabilities because it's good enough. They would rather have the
• 1:24:27
  Speaker 1: good enough model for way cheaper jobs.
• 1:24:28
  Speaker 2: Or they made a mistake with hiring. They bloated their
• 1:24:30
  Speaker 2: team and it's a great, convenient thing to say.
• 1:24:32
  Speaker 3: Exactly. Like, there's many reasons.
• 1:24:34
  Speaker 1: But, like, clearly we're already seeing impacts on the job market.
• 1:24:37
  Speaker 3: Like, the U.S.
• 1:24:39
  Speaker 1: Jobs report that came out earlier this year showed that
• 1:24:43
  Speaker 1: there has.
• 1:24:43
  Speaker 3: Been a decline in hiring.
• 1:24:46
  Speaker 1: It's a slowdown in hiring across especially white-collar professional industries.
• 1:24:53
  Speaker 2: And you saw Anthropix report, didn't you, this week? The
• 1:24:55
  Speaker 2: TLDR matches kind of what you were saying, where they
• 1:24:58
  Speaker 2: anthropically looked at exactly how people were using their models.
• 1:25:01
  Speaker 2: And they looked at what people are saying. And they
• 1:25:05
  Speaker 2: said that there's been a 40% reduction in entry-level jobs
• 1:25:08
  Speaker 2: in particular. And then they made this graph, which has
• 1:25:10
  Speaker 2: gone viral over the internet. The red shows where we
• 1:25:12
  Speaker 2: are now in terms of capability. And based on how
• 1:25:15
  Speaker 2: people are currently using the models, they extrapolated out that
• 1:25:18
  Speaker 2: the blue part will be the disrupted parts. This is
• 1:25:21
  Speaker 2: the things that they say AI can do right now.
• 1:25:24
  Speaker 2: but people don't realize it yet. So if you look
• 1:25:26
  Speaker 2: at it, it's like, it's kind of all the stuff
• 1:25:27
  Speaker 2: you'd expect. It's the physical real world human stuff, which
• 1:25:32
  Speaker 2: robots maybe can do someday like construction or agriculture that
• 1:25:35
  Speaker 2: are untouched, but like office and admin, like saying finance stuff, math.
• 1:25:40
  Speaker 1: And notice that these are all the things that I
• 1:25:42
  Speaker 1: just named that they purposely finance.
• 1:25:44
  Speaker 3: Math, law, healthcare.
• 1:25:47
  Speaker 2: Media and arts, that's me cooked.
• 1:25:48
  Speaker 1: Yeah. Office and admin, I mean, they do focus a
• 1:25:51
  Speaker 1: lot on like assistant type jobs.
• 1:25:54
  Speaker 3: And managerial work.
• 1:25:57
  Speaker 1: But the other thing that the Klarna CEO said was,
• 1:26:02
  Speaker 1: but people also want human experiences. So it's not actually
• 1:26:06
  Speaker 1: just about the capabilities of the models. It's also about
• 1:26:09
  Speaker 1: what people want, like some things they would turn to
• 1:26:13
  Speaker 1: AI for and some things they wouldn't, irrespective of whether
• 1:26:17
  Speaker 1: or not AI is capable of doing it. But because
• 1:26:20
  Speaker 1: of a preference that they want human-to-human interaction. And so
• 1:26:25
  Speaker 1: what we're seeing right now is, yeah, the thing that
• 1:26:30
  Speaker 1: happens with every wave of automation, which is that there
• 1:26:33
  Speaker 1: is a bunch of entry-level work that gets automated away.
• 1:26:37
  Speaker 1: And there are also new jobs created, but the jobs
• 1:26:40
  Speaker 1: that are created are in one of two categories.
• 1:26:43
  Speaker 3: There are people.
• 1:26:45
  Speaker 1: That get even higher skilled jobs. And what he was saying,
• 1:26:47
  Speaker 1: like we pay people more for like the handcrafted code now.
• 1:26:52
  Speaker 3: And there's also the people who get way worse jobs.
• 1:26:55
  Speaker 1: And so there was this amazing article in New York
• 1:26:58
  Speaker 1: magazine that was talking about how a lot of people
• 1:27:02
  Speaker 1: are getting laid off and then they end up working
• 1:27:06
  Speaker 1: in data annotation. which is the labor that I've been
• 1:27:09
  Speaker 1: referring to throughout this conversation that companies need in order
• 1:27:13
  Speaker 1: to teach their models the next thing that the companies
• 1:27:16
  Speaker 1: are trying to automate. And so like a marketer gets
• 1:27:19
  Speaker 1: laid off and then they go and work for a
• 1:27:22
  Speaker 1: data annotation firm to train the models on the very
• 1:27:26
  Speaker 1: job that they were just laid off in, which will
• 1:27:29
  Speaker 1: then perpetuate more layoffs if that model then develops that skill.
• 1:27:35
  Speaker 1: And the article was talking about how this has become
• 1:27:39
  Speaker 1: a huge catch-all for a lot of people that are
• 1:27:44
  Speaker 1: struggling with finding job opportunities right now, including award-winning directors
• 1:27:50
  Speaker 1: in Hollywood that are actually secretly doing this data annotation
• 1:27:53
  Speaker 1: work to put food on the table. And so when
• 1:27:56
  Speaker 1: they talk about There's going to be mass unemployment and
• 1:28:01
  Speaker 1: then there's going to be some new jobs created that
• 1:28:02
  Speaker 1: we can't even imagine. I think a lot of these
• 1:28:05
  Speaker 1: narratives rarely talk about, like, first of all, why are
• 1:28:09
  Speaker 1: some jobs going away? It's not just because of the
• 1:28:10
  Speaker 1: model capabilities, it's also because of executive choices and because
• 1:28:13
  Speaker 1: of the rhetoric that they use if they want to
• 1:28:15
  Speaker 1: just downsize. But the other thing that is rarely talked
• 1:28:19
  Speaker 1: about is the jobs, a lot of the jobs that
• 1:28:22
  Speaker 1: are created are way worse than the jobs that were there.
• 1:28:27
  Speaker 1: And it breaks the career ladder. So it's the entry
• 1:28:30
  Speaker 1: level and the mid-tier jobs that get gouged out. It's
• 1:28:34
  Speaker 1: higher order jobs and then way more lower order jobs
• 1:28:38
  Speaker 1: that get created. And so how do people continue to
• 1:28:43
  Speaker 1: progress in their careers? There's no more rungs on the ladder.
• 1:28:46
  Speaker 2: I actually don't know the answer to this question. And
• 1:28:47
  Speaker 2: I've been furiously trying to find a good answer to
• 1:28:49
  Speaker 2: this question because I can, you I would say most
• 1:28:55
  Speaker 2: of my audience don't run businesses. A lot of them do,
• 1:28:57
  Speaker 2: a lot of them aspire to, but they don't run businesses.
• 1:28:59
  Speaker 2: So they're kind of, they're also in the land of theory.
• 1:29:02
  Speaker 2: They're hearing lots of different things. Jack Dorsey does his
• 1:29:04
  Speaker 2: tweet saying he's halving his headcount because of AI. They
• 1:29:06
  Speaker 2: don't know what's true. They don't know the sort of
• 1:29:08
  Speaker 2: internal economics of Jack's company. And did he bloat the
• 1:29:11
  Speaker 2: company during the pandemic? And he's just using this as
• 1:29:12
  Speaker 2: an excuse to make this share price spike seven points
• 1:29:15
  Speaker 2: because his investors now think they're an AI company or whatever.
• 1:29:18
  Speaker 2: It's hard to pass through. So eventually I go, okay,
• 1:29:21
  Speaker 2: what am I doing? I have hundreds of team members,
• 1:29:24
  Speaker 2: probably 70 companies I invest in, maybe five or six
• 1:29:26
  Speaker 2: that I'm like the lead shareholder in. What am I
• 1:29:28
  Speaker 2: actually doing on a day-to-day basis right now? I also
• 1:29:32
  Speaker 2: consider myself to be head of recruitment. But in the
• 1:29:34
  Speaker 2: last month in particular, I have met extremely capable candidates
• 1:29:38
  Speaker 2: in terms of cultural alignment, hard work, those kinds of things.
• 1:29:41
  Speaker 2: But I've had to take a great deal of pause
• 1:29:43
  Speaker 2: because when I run the experiment of can I get
• 1:29:45
  Speaker 2: an AI agent to do that exact same thing, the
• 1:29:48
  Speaker 2: answer is increasingly yes, especially in a world of open-close.
• 1:29:53
  Speaker 1: And so what I'm curious, like, now you confront this
• 1:29:57
  Speaker 1: decision where you're seeing in this short-term period, you could
• 1:30:02
  Speaker 1: just choose the AI agent. And in the long-term period,
• 1:30:07
  Speaker 1: there is no career ladder. who are you promoting into
• 1:30:12
  Speaker 1: these senior roles? How do you resolve it for your
• 1:30:15
  Speaker 1: own company?
• 1:30:16
  Speaker 2: Yeah, it's a good question. So there's kind of two
• 1:30:17
  Speaker 2: ways I'm thinking about it. I think really deep expertise
• 1:30:20
  Speaker 2: is very, very valuable because if you're now the orchestrator
• 1:30:23
  Speaker 2: of potentially AI agents, it's really about having a deep
• 1:30:27
  Speaker 2: understanding of the right question to ask. And that's someone
• 1:30:29
  Speaker 2: who has deep expertise on something. So I need my
• 1:30:32
  Speaker 2: CFO because if she's going to be orchestrating our team
• 1:30:35
  Speaker 2: of agents that might be doing financial analysis or whatever else,
• 1:30:38
  Speaker 2: she needs to understand... what to tell them to do
• 1:30:42
  Speaker 2: in our company. And in turn, financial analysts can't do that.
• 1:30:45
  Speaker 2: They need the 50 odd years of experience that, you know,
• 1:30:48
  Speaker 2: Claire has. On the other end, I need Kaz. Kaz
• 1:30:52
  Speaker 2: is 25. Kaz knows everything about AI agents. He's a
• 1:30:56
  Speaker 2: young Japanese kid who's highly, highly curious. You know, on
• 1:30:59
  Speaker 2: the weekend, he's building AI agents to solve problems in
• 1:31:01
  Speaker 2: my life. I need those two kinds of thinking. which
• 1:31:04
  Speaker 2: is highly proficient agent maxing young kids, or they don't
• 1:31:07
  Speaker 2: necessarily need to be young, but like really lean in
• 1:31:10
  Speaker 2: high curiosity. That's creating a forceful multiplier in my business.
• 1:31:12
  Speaker 2: And then I need deep expertise. Now, everything else outside of,
• 1:31:17
  Speaker 2: there is another one I've thought of, another group, is
• 1:31:19
  Speaker 2: like people with extremely great IRL people skills. Because we
• 1:31:24
  Speaker 2: do meet people in real life. We greet you when
• 1:31:26
  Speaker 2: you arrive here. When we go for lunch with big
• 1:31:29
  Speaker 2: clients that we have, whether it's Apple or LinkedIn or
• 1:31:31
  Speaker 2: whoever it might be, we, you know, we need to
• 1:31:33
  Speaker 2: schmooze and we have teams who, you know, are in
• 1:31:37
  Speaker 2: person in the office. So we, we do a lot
• 1:31:39
  Speaker 2: of stuff IRL and increasingly we're building communities, even for
• 1:31:41
  Speaker 2: this show, we're doing community events all around the world.
• 1:31:43
  Speaker 2: So we need people that are good at that as well. IRL,
• 1:31:46
  Speaker 2: bringing people together in real life and organizing stuff. Those
• 1:31:49
  Speaker 2: are the three groups of people that I'm like, you know,
• 1:31:52
  Speaker 2: irreplaceable right now.
• 1:31:53
  Speaker 1: And if you were to, to, All of the roles
• 1:31:58
  Speaker 1: that could be done by AI agents, if we were
• 1:32:00
  Speaker 1: to replace them with AI agents, do you think you
• 1:32:01
  Speaker 1: would still have these three roles, pools of people to
• 1:32:05
  Speaker 1: hire and promote into the three critical things that you
• 1:32:08
  Speaker 1: need in the long term?
• 1:32:10
  Speaker 2: If things carry on at the current rate of trajectory,
• 1:32:14
  Speaker 2: one could assert that even those roles would experience pressure.
• 1:32:18
  Speaker 2: If you just imagine, like people think of things either
• 1:32:20
  Speaker 2: statically or linearly or exponentially. If you imagine an exponential
• 1:32:23
  Speaker 2: rate of improvement, which is kind of what I've seen,
• 1:32:25
  Speaker 2: even like a 10% compounding rate of improvement, at some point...
• 1:32:33
  Speaker 2: At some point, I think what remains is actually the
• 1:32:37
  Speaker 2: IRL irreplaceably human stuff, human to human. Our Maslowian needs
• 1:32:40
  Speaker 2: of being in person like we are now aren't gonna change.
• 1:32:43
  Speaker 2: We need connection. Humans get very sick when they don't
• 1:32:46
  Speaker 2: have other human beings in their life and strong, deep relationships.
• 1:32:50
  Speaker 2: So that stuff is going to matter a whole lot.
• 1:32:53
  Speaker 2: I have this contrarian, weird take that actually maybe this
• 1:32:55
  Speaker 2: is the first technology that's going to deliver on the
• 1:32:57
  Speaker 2: promise of making us human and connected because we're going
• 1:33:00
  Speaker 2: to be rendered useless at everything else other than what
• 1:33:02
  Speaker 2: humans are good at. Because all the other technologies said, oh,
• 1:33:05
  Speaker 2: we're going to make you more connected, connecting the world.
• 1:33:07
  Speaker 2: And they disconnected the world and isolated the world. But
• 1:33:09
  Speaker 2: maybe this is the one that's so intelligent now that
• 1:33:11
  Speaker 2: it doesn't need us to fuck around in spreadsheets anymore.
• 1:33:13
  Speaker 3: Do you see.
• 1:33:16
  Speaker 1: that actually happening in real time right now, that it's
• 1:33:19
  Speaker 1: making us more able to be in person, connected with
• 1:33:23
  Speaker 1: one another, having deeper social community engagements?
• 1:33:28
  Speaker 2: Yes.
• 1:33:29
  Speaker 3: Yes.
• 1:33:29
  Speaker 2: And I'll give you some data points.
• 1:33:31
  Speaker 3: Okay.
• 1:33:31
  Speaker 2: Data point number one. The Financial Times released a report
• 1:33:34
  Speaker 2: on social media usage. And what they saw is 2022
• 1:33:37
  Speaker 2: was the peak and it's plateaued ever since. The generation
• 1:33:41
  Speaker 2: that's plateaued the fastest and heading down is the younger generations.
• 1:33:45
  Speaker 2: The boomers are still off to the races, right? On
• 1:33:47
  Speaker 2: Facebook and stuff. And then you look at the way
• 1:33:50
  Speaker 2: Gen Alpha are using social media. They're not posting as much.
• 1:33:53
  Speaker 2: They call it posting zero. They're scrolling sometimes, but they're
• 1:33:56
  Speaker 2: in dark social environments like WhatsApp and Snapchat and iMessage.
• 1:33:59
  Speaker 2: They're not like performing to the world. They also value
• 1:34:02
  Speaker 2: IRL experiences much more than any other generation. They're like
• 1:34:04
  Speaker 2: not getting smashed. We're seeing every brand has a run club.
• 1:34:07
  Speaker 2: We're seeing, I mean, run clubs exploding around the world.
• 1:34:11
  Speaker 2: And we're seeing this real sort of, sort of almost
• 1:34:14
  Speaker 2: like innate realization that like technology let us down at
• 1:34:17
  Speaker 2: some fundamental level. Like dating apps let us down, social
• 1:34:21
  Speaker 2: networking kind of has let us down. And we're seeing,
• 1:34:23
  Speaker 2: I think, maybe a bifurcation of society where a lot
• 1:34:26
  Speaker 2: of people are going, fuck this. Like, I want to
• 1:34:27
  Speaker 2: go back to what it is to be a human.
• 1:34:29
  Speaker 2: And I would imagine that in such a world where
• 1:34:32
  Speaker 2: intelligence is so sophisticated that we no longer needed to
• 1:34:34
  Speaker 2: sit at laptops and like, I think screen time is
• 1:34:37
  Speaker 2: going to continue to fall. I think you go into
• 1:34:39
  Speaker 2: an office, you're not going to see people sat at laptops.
• 1:34:40
  Speaker 2: You're going to see something completely different. And I think maybe,
• 1:34:45
  Speaker 2: you know, and then we talk about robots and optimist robots.
• 1:34:48
  Speaker 2: Elon says there'll be 10 billion optimist robots. Elon has
• 1:34:51
  Speaker 2: been wrong with timing before. He's almost never been wrong
• 1:34:55
  Speaker 2: on the big things. Completely. He's just, his timing has
• 1:34:59
  Speaker 2: got a bad track record. So I think he's probably right.
• 1:35:03
  Speaker 2: You know, I think I've got some people on the
• 1:35:05
  Speaker 2: way from Boston Dynamics and these other big companies like
• 1:35:07
  Speaker 2: Scale AI, and they're actually bringing the robots here to
• 1:35:10
  Speaker 2: show it like folding laundry, doing the dishes. I'm not
• 1:35:12
  Speaker 2: saying that's what I would want in my home, but
• 1:35:13
  Speaker 2: I think factory work is going to completely change. I
• 1:35:15
  Speaker 2: think a lot of manual labor is going to completely change.
• 1:35:17
  Speaker 2: And I think we're going to be forced to do
• 1:35:19
  Speaker 2: what only we can do. Sebastian, who's the CEO of Klana,
• 1:35:24
  Speaker 2: has actually just called me. Hello, Sebastian. You're right. Hey,
• 1:35:30
  Speaker 2: how are you? I'm good. How are you? It's been
• 1:35:34
  Speaker 2: a while. It has been a while since you're on
• 1:35:36
  Speaker 2: the show. I was just saying we do need to
• 1:35:37
  Speaker 2: get you back on. I just had a couple of
• 1:35:39
  Speaker 2: simple questions because, you know, I do a lot of
• 1:35:41
  Speaker 2: interviews and Kalan has always mentioned because I think the
• 1:35:44
  Speaker 2: media has said that you like double down on air,
• 1:35:46
  Speaker 2: then you reverse because it didn't work out. So I
• 1:35:49
  Speaker 2: know I spoke to you a while ago and we
• 1:35:50
  Speaker 2: exchanged a couple of DMs about it, but that was
• 1:35:52
  Speaker 2: more than it was almost a year ago now. So
• 1:35:55
  Speaker 2: I just wanted to get an update on Klarna's business, AI, agents,
• 1:35:58
  Speaker 2: and all of that, if possible.
• 1:35:59
  Speaker 4: First and foremost, we were early on released AI to
• 1:36:04
  Speaker 4: support our customer service, which had that initial benefit of
• 1:36:09
  Speaker 4: more calls being dealt with by AI, which customers liked
• 1:36:12
  Speaker 4: because those calls or chat messages were much, much faster
• 1:36:16
  Speaker 4: and more qualitative. Then since then, that has actually expanded slightly.
• 1:36:21
  Speaker 4: What we did, however, try to communicate as well is
• 1:36:23
  Speaker 4: that we believed in a world of where AI is
• 1:36:26
  Speaker 4: cheap and available, the value.
• 1:36:29
  Speaker 2: Of human interaction will be regarded as higher.
• 1:36:32
  Speaker 4: So the future of customer service VIP is a human.
• 1:36:37
  Speaker 4: We have then hence doubled down on providing more of that.
• 1:36:40
  Speaker 4: But at the same time, the efficiency gains within the
• 1:36:42
  Speaker 4: company has continued. I mean, we used to be about 6,000 people,
• 1:36:48
  Speaker 4: and now we are less than 3,000. which is two,
• 1:36:51
  Speaker 4: three years since we stopped recruiting. And at the same
• 1:36:54
  Speaker 4: point of time, our revenue has doubled, right? So you
• 1:36:57
  Speaker 4: can clearly see that AI has allowed us to do
• 1:37:01
  Speaker 4: more with less people, but we have avoided layoffs and
• 1:37:04
  Speaker 4: instead relied on natural attrition when people kind of move
• 1:37:08
  Speaker 4: on to other jobs. From my perspective, we will continue
• 1:37:12
  Speaker 4: to be very, you know, not really recruit much. I mean,
• 1:37:16
  Speaker 4: we recruit a little bit here and there, but we
• 1:37:18
  Speaker 4: expect that kind of natural attrition of 10, 15 percent
• 1:37:21
  Speaker 4: per year to continue on to become fewer. I think
• 1:37:25
  Speaker 4: the big breakthrough was really in November, December last year,
• 1:37:29
  Speaker 4: where even the kind of more most skeptical engineers who
• 1:37:35
  Speaker 4: are like very well renowned and appreciated, like the founder
• 1:37:38
  Speaker 4: of Linux and stuff. like that basically said that coding
• 1:37:40
  Speaker 4: has now been resolved and hence is not, you know,
• 1:37:45
  Speaker 4: you don't need to code anymore. And that was kind
• 1:37:47
  Speaker 4: of a common sentiment. So I think in coding, that's
• 1:37:49
  Speaker 4: definitely an engineering work that has been a tremendous shift
• 1:37:54
  Speaker 4: in the last six months.
• 1:37:56
  Speaker 2: What do all these people go do, Sebastian? I am optimistic.
• 1:38:00
  Speaker 4: I mean, I think obviously people will have a lot
• 1:38:02
  Speaker 4: of opinions about this topic, but I still believe that
• 1:38:05
  Speaker 4: we are going to move towards a richer society. Now,
• 1:38:09
  Speaker 4: in the short term, there could be more worry about
• 1:38:13
  Speaker 4: what happens if people don't get a job and so forth.
• 1:38:15
  Speaker 4: But I think in the longer term, I am optimistic
• 1:38:19
  Speaker 4: what it means for society and humanity.
• 1:38:21
  Speaker 2: Thank you so much, Seb. I'll chat to you soon.
• 1:38:23
  Speaker 2: Thank you for taking the time. I appreciate you, mate.
• 1:38:26
  Speaker 1: Thanks.
• 1:38:29
  Speaker 2: You know, the little traditional SIM card that goes inside
• 1:38:31
  Speaker 2: of our phones, they haven't changed at all since they
• 1:38:34
  Speaker 2: were invented in the 90s. You have this physical piece
• 1:38:36
  Speaker 2: of plastic that means you're locked into one carrier, one network,
• 1:38:40
  Speaker 2: and the second you cross a border, that carrier can
• 1:38:43
  Speaker 2: start charging you whatever they want, but there are alternatives.
• 1:38:46
  Speaker 2: And today's sponsor, Saley, is one of them. It's an
• 1:38:49
  Speaker 2: eSIM app that gives you a safe and secure data
• 1:38:52
  Speaker 2: connection in over 200 destinations. All of their eSIMs have
• 1:38:56
  Speaker 2: built-in cybersecurity, which is great if you're traveling for work
• 1:38:58
  Speaker 2: and looking at confidential material. I've been using Saley whenever
• 1:39:01
  Speaker 2: I travel because the connection is always reliable and it
• 1:39:04
  Speaker 2: saves me a ton of roaming fees. It also means
• 1:39:07
  Speaker 2: I don't have to deal with all of the faff
• 1:39:09
  Speaker 2: that surrounds sorting out a SIM everywhere I go. If
• 1:39:11
  Speaker 2: you want to give it a try, download the Saley
• 1:39:13
  Speaker 2: app from the App Store now and scan the QR
• 1:39:15
  Speaker 2: code on screen. And if you want 15% off your
• 1:39:18
  Speaker 2: first purchase, use my code DOAC when you get to checkout.
• 1:39:24
  Speaker 2: That's DOAC for 15% off. Keep that to yourself. This
• 1:39:29
  Speaker 2: is something that I've made for you. I've realized that
• 1:39:32
  Speaker 2: the Dairyverse here audience are strivers, whether it's in business
• 1:39:35
  Speaker 2: or health, we all have big goals that we want
• 1:39:37
  Speaker 2: to accomplish. And one of the things I've learned is
• 1:39:40
  Speaker 2: that when you aim at the big, big, big goal,
• 1:39:42
  Speaker 2: it can feel incredibly psychologically uncomfortable because it's kind of
• 1:39:47
  Speaker 2: like being stood at the foot of Mount Everest and
• 1:39:49
  Speaker 2: looking upwards. The way to accomplish your goals is by
• 1:39:52
  Speaker 2: breaking them down into tiny small steps. And we call
• 1:39:55
  Speaker 2: this in our team, the 1%. And actually this philosophy
• 1:39:58
  Speaker 2: is highly responsible for much of our success here. So
• 1:40:02
  Speaker 2: what we've done so that you at home can accomplish
• 1:40:04
  Speaker 2: any big goal that you have is we've made these
• 1:40:06
  Speaker 2: 1% diaries and we released these last year and they
• 1:40:10
  Speaker 2: all sold out. So I asked my team over and
• 1:40:12
  Speaker 2: over again to bring the diaries back, but also to
• 1:40:14
  Speaker 2: introduce some new colors and to make some minor tweaks
• 1:40:17
  Speaker 2: to the diary. we have a better range for you.
• 1:40:22
  Speaker 2: So if you have a big goal in mind and
• 1:40:24
  Speaker 2: you need a framework and a process and some motivation,
• 1:40:27
  Speaker 2: then I highly recommend you get one of these diaries
• 1:40:29
  Speaker 2: before they all sell out once again. And you can
• 1:40:32
  Speaker 2: get yours at thediary.com. And if you want the link,
• 1:40:35
  Speaker 2: the link is in the description below. Any thoughts?
• 1:40:39
  Speaker 3: Well, I actually had thoughts on something that you said
• 1:40:42
  Speaker 3: before he called.
• 1:40:44
  Speaker 1: Mm-hmm. Which is you were saying that the Gen Zers,
• 1:40:47
  Speaker 1: like there's this trend that they're actually disconnecting from technology.
• 1:40:50
  Speaker 3: So they're becoming more in person.
• 1:40:52
  Speaker 1: And then there's this other class of workers that are
• 1:40:54
  Speaker 1: actually leaning into the technology but then becoming more human
• 1:40:56
  Speaker 1: because they're leaning into the technology. Because they're realizing that
• 1:41:01
  Speaker 1: they should actually just be spending more time doing in-person
• 1:41:04
  Speaker 1: to person interactions rather than staring at a spreadsheet. And
• 1:41:07
  Speaker 1: so they're no longer doing the typing and whatever. I
• 1:41:10
  Speaker 1: really want to go back to this New York Magazine
• 1:41:11
  Speaker 1: piece that just came out. Because what you're describing is
• 1:41:16
  Speaker 1: true for a very specific category of people, which is
• 1:41:18
  Speaker 1: often like the business owners and leadership within companies that
• 1:41:22
  Speaker 1: actually can make these decisions on how they spend their
• 1:41:26
  Speaker 1: time and what they ultimately do with their time.
• 1:41:29
  Speaker 3: But what the piece talks about is the....
• 1:41:33
  Speaker 1: Working class, like people who are not business owners that
• 1:41:38
  Speaker 1: are then having to experience being laid off and then
• 1:41:43
  Speaker 1: working for the data annotation industry, which is now one
• 1:41:46
  Speaker 1: of the top jobs on LinkedIn, by the way. Yeah,
• 1:41:50
  Speaker 1: so LinkedIn had a report that showed the top 10
• 1:41:54
  Speaker 1: jobs with the highest growth in the last year, and
• 1:41:58
  Speaker 1: data annotation is on that list.
• 1:42:00
  Speaker 2: And for anyone that doesn't know what data annotation is, Yeah.
• 1:42:02
  Speaker 1: So data annotation is the process of teaching these chatbots
• 1:42:09
  Speaker 1: or any AI system to do what they ultimately are
• 1:42:12
  Speaker 1: able to do. So the fact that ChatGPT can chat
• 1:42:15
  Speaker 1: is because there were tens of thousands or hundreds of
• 1:42:17
  Speaker 1: thousands of people that were literally typing into a large
• 1:42:21
  Speaker 1: language model and showing it, this is how you're supposed
• 1:42:24
  Speaker 1: to then respond when a user types in a prompt
• 1:42:27
  Speaker 1: like this. Before they did that work, Chat GPT didn't exist.
• 1:42:32
  Speaker 1: Like it just, it would just, you would prompt the
• 1:42:35
  Speaker 1: model and the model would generate some texts that was
• 1:42:37
  Speaker 1: not in dialogue with the person. It would kind of
• 1:42:40
  Speaker 1: generate something that was adjacently related.
• 1:42:42
  Speaker 2: Is this what they call reinforcement learning where you kind of,
• 1:42:44
  Speaker 2: you give it like a.
• 1:42:45
  Speaker 1: It's a part of the process of reinforcement learning. So
• 1:42:47
  Speaker 1: you do data annotation, which is literally showing lots of
• 1:42:51
  Speaker 1: different you know, examples of things that you want the
• 1:42:55
  Speaker 1: model to know. And then reinforcement learning is getting the
• 1:42:58
  Speaker 1: model to then train on those examples iteratively in a
• 1:43:01
  Speaker 1: way that then gives the model some of those capabilities.
• 1:43:05
  Speaker 1: And what the New York Magazine piece highlighted is many,
• 1:43:10
  Speaker 1: many of the people that are getting laid off now
• 1:43:12
  Speaker 1: or are struggling to find work.
• 1:43:14
  Speaker 3: And these are highly educated people.
• 1:43:16
  Speaker 1: They're college graduates, PhD graduates, law degree graduates, doctors, and again,
• 1:43:23
  Speaker 1: like award-winning directors that are then struggling to find employment
• 1:43:28
  Speaker 1: in the economy because the economy has been very much
• 1:43:31
  Speaker 1: restructured by AI. They are then finding themselves serving this industry.
• 1:43:37
  Speaker 1: And the industry is designed in a way that is
• 1:43:40
  Speaker 1: extremely inhumane because The companies that use these data annotation services,
• 1:43:47
  Speaker 1: there's these third-party providers that are data annotation firms.
• 1:43:52
  Speaker 3: An OpenAI, a.
• 1:43:53
  Speaker 1: Grok, a Google, they will hire these firms to then
• 1:43:57
  Speaker 1: find the workers to perform the data annotation tasks that
• 1:44:00
  Speaker 1: they need. For these firms, these third-party firms, they are
• 1:44:05
  Speaker 1: incentivized to pit workers against each other because they want
• 1:44:10
  Speaker 1: this data annotation to happen at speed and as cheaply
• 1:44:12
  Speaker 1: as possible so that they can also compete with one
• 1:44:15
  Speaker 1: another in this middle layer to get the contract from
• 1:44:20
  Speaker 1: the client. And so all of these workers that were
• 1:44:25
  Speaker 1: interviewed for this New York Magazine story talk about how
• 1:44:28
  Speaker 1: they actually no longer have an ability to be human.
• 1:44:32
  Speaker 1: because they are waiting at their laptop to be pinged
• 1:44:36
  Speaker 1: on Slack for when a project is going to open
• 1:44:39
  Speaker 1: up for data annotation because they've tried job hunting. They
• 1:44:42
  Speaker 1: literally can't find anything else. This is the thing that's
• 1:44:44
  Speaker 1: going to help them put food on the table for
• 1:44:46
  Speaker 1: their kids. And there was this one woman who said, like,
• 1:44:50
  Speaker 1: I have so much anxiety about when the project is
• 1:44:53
  Speaker 1: gonna come, when it's gonna leave, that when the project came,
• 1:44:56
  Speaker 1: it was right when my kid was coming off of school.
• 1:45:00
  Speaker 1: And I just started tasking furiously because I don't know
• 1:45:03
  Speaker 1: when it's gonna go and I need to earn as
• 1:45:04
  Speaker 1: much money as possible in this window of opportunity. So
• 1:45:07
  Speaker 1: then when my kid came home and tried to talk
• 1:45:10
  Speaker 1: to me, I screamed at my child for distracting me.
• 1:45:15
  Speaker 1: And then she was like, I've become a monster And
• 1:45:19
  Speaker 1: I'm not even allowed to go to the bathroom or
• 1:45:21
  Speaker 1: take care of my kids, let alone myself, because this
• 1:45:26
  Speaker 1: industry that is absorbing more and more of the workers
• 1:45:29
  Speaker 1: that are being laid off is mechanizing my life, atomizing
• 1:45:35
  Speaker 1: my work, devaluing my expertise, and then harvesting it for
• 1:45:42
  Speaker 1: the perpetuation of this machine that all of these AI
• 1:45:45
  Speaker 1: executives are saying is then going to come for everyone
• 1:45:47
  Speaker 1: else's jobs.
• 1:45:49
  Speaker 3: And so what you were saying.
• 1:45:51
  Speaker 1: About this class of workers, the business owners that get
• 1:45:56
  Speaker 1: to become more human because there are all of these
• 1:45:59
  Speaker 1: AI models now doing the tasks that they don't have
• 1:46:01
  Speaker 1: to do anymore. It is at the cost of the
• 1:46:05
  Speaker 1: vast majority of people who are not business owners that
• 1:46:08
  Speaker 1: are struggling to find work, getting absorbed into the work
• 1:46:12
  Speaker 1: of then providing these technologies that the business owners can use.
• 1:46:18
  Speaker 1: And instead of becoming more human, they feel like their
• 1:46:21
  Speaker 1: humanity has been squeezed and diminished, and they have no
• 1:46:27
  Speaker 1: ability to have control, agency, and dignity in their lives anymore.
• 1:46:31
  Speaker 2: I think this is a big question that kind of
• 1:46:34
  Speaker 2: pertains to this graph here, which is, you know, all
• 1:46:37
  Speaker 2: of these people, if we believe anthropic's prediction of who
• 1:46:40
  Speaker 2: will be disrupted, these people in these industries like arts
• 1:46:43
  Speaker 2: and media, legal... life and social sciences, architecture and engineering,
• 1:46:49
  Speaker 2: computer and maths, business and finance, and management, and also
• 1:46:53
  Speaker 2: office and admin. These people, if we believe this, would
• 1:46:57
  Speaker 2: have to retrain at something else. And unlike the industrial
• 1:46:59
  Speaker 2: revolution where you might get 10, 20 years to retrain
• 1:47:02
  Speaker 2: because factories take a long time to build, the distribution
• 1:47:05
  Speaker 2: layer that AI sits on top of is the open internet.
• 1:47:07
  Speaker 2: So this is why Chachapati can go pop and get
• 1:47:10
  Speaker 2: hundreds of millions of users in no time at all
• 1:47:12
  Speaker 2: and become the fastest growing company of all time. One
• 1:47:15
  Speaker 2: of my fears is that this disruption takes place at
• 1:47:18
  Speaker 2: a speed where we can't transition.
• 1:47:21
  Speaker 1: And that was, you know, I think you said that
• 1:47:24
  Speaker 1: sentence in the passive voice, the transition would happen at
• 1:47:28
  Speaker 1: a speed, but who is driving that speed?
• 1:47:33
  Speaker 3: It's the companies.
• 1:47:34
  Speaker 2: The companies, yeah.
• 1:47:34
  Speaker 3: And their race with one another. Yeah.
• 1:47:37
  Speaker 1: And so they are driving the transition to happen at
• 1:47:40
  Speaker 1: a speed at which it would be possible really hard
• 1:47:44
  Speaker 1: to take care of all of the people that would
• 1:47:47
  Speaker 1: be bulldozed over by this is one of the.
• 1:47:49
  Speaker 2: Crazy questions that no one can answer for me when
• 1:47:51
  Speaker 2: i sit with these people that are ai ceos i
• 1:47:53
  Speaker 2: go so what happens to the people if this is
• 1:47:55
  Speaker 2: if you agree that this is going to happen at
• 1:47:56
  Speaker 2: super speed you know i've spoke to that ceo of
• 1:47:59
  Speaker 2: uber dara who said very similar things to what you're
• 1:48:01
  Speaker 2: saying is you know there'll be data labeling jobs for
• 1:48:04
  Speaker 2: example for the drivers but um they can't all become
• 1:48:07
  Speaker 2: data labelers and there's a question around meaning and purpose
• 1:48:10
  Speaker 2: and fulfillment and that comes from losing your meaning in life.
• 1:48:14
  Speaker 2: I also sit here with so many people who talk
• 1:48:16
  Speaker 2: about how their father lost their job in Iran or
• 1:48:19
  Speaker 2: some other country and came to the United States and
• 1:48:22
  Speaker 2: had to be a toilet cleaner on a particular case,
• 1:48:25
  Speaker 2: was a doctor in Iran, but came to the US
• 1:48:27
  Speaker 2: and was a toilet cleaner and had to deal with
• 1:48:29
  Speaker 2: the sense of shame that that particular person felt and
• 1:48:32
  Speaker 2: the lack of dignity that that caused and how that
• 1:48:34
  Speaker 2: made that person's self-esteem feel and the depression and alcoholism
• 1:48:37
  Speaker 2: that transpired from that. If this happens at a large
• 1:48:40
  Speaker 2: scale across society, there's going to be a ton of
• 1:48:44
  Speaker 2: consequences like that.
• 1:48:45
  Speaker 3: I mean, this is like the core themes of my work.
• 1:48:48
  Speaker 1: And the reason why I'm critical of these companies is
• 1:48:50
  Speaker 1: that they are creating technologies in a way that creates
• 1:48:55
  Speaker 1: the haves.
• 1:48:55
  Speaker 3: And have-nots in an extreme form that we have.
• 1:48:59
  Speaker 1: It's exacerbating the inequality that we already see in the world.
• 1:49:03
  Speaker 1: Like the people who have things... will have way more riches.
• 1:49:07
  Speaker 1: They'll have way more free time. They'll be allowed to
• 1:49:09
  Speaker 1: be more human. But the people who don't have things
• 1:49:13
  Speaker 1: are being squeezed even more. And it's not just from
• 1:49:19
  Speaker 1: a work perspective. I mean, I talk in my book
• 1:49:23
  Speaker 1: also about the environmental and public health crisis that these
• 1:49:26
  Speaker 1: companies have created, where they are building these colossal supercomputer
• 1:49:33
  Speaker 1: facilities and in communities all around the world. And they
• 1:49:39
  Speaker 1: specifically pick some of the most vulnerable communities. We're sitting
• 1:49:42
  Speaker 1: in Texas right now. OpenAI's largest, one of its largest
• 1:49:47
  Speaker 1: data center projects is being built in Abilene, Texas as
• 1:49:50
  Speaker 1: part of the Stargate initiative, which was an effort announced
• 1:49:53
  Speaker 1: at the beginning of Trump's second administration to spend $ 500
• 1:49:56
  Speaker 1: billion on AI computing infrastructure. This facility is, Consumes will
• 1:50:04
  Speaker 1: when it's finished will consume more than a gigawatt of power,
• 1:50:07
  Speaker 1: which is over 20% Over 20% so this is actually
• 1:50:13
  Speaker 1: a little bit inaccurate now This was something that circulated
• 1:50:16
  Speaker 1: online for a while But there's updated numbers.
• 1:50:19
  Speaker 2: Just for someone that can't see because they're listening on
• 1:50:21
  Speaker 2: Spotify or something It's a picture of the size of
• 1:50:23
  Speaker 2: this facility.
• 1:50:25
  Speaker 3: So this is not the Abilene Texas one. This is
• 1:50:28
  Speaker 3: a meta facility.
• 1:50:29
  Speaker 1: So let's talk about opening a facility in Texas and
• 1:50:32
  Speaker 1: That one would be the size of Central Park, and
• 1:50:36
  Speaker 1: it would run a million computer chips, and it would
• 1:50:40
  Speaker 1: require the power of more than 20% of New York City.
• 1:50:43
  Speaker 2: Do you know one of the things which I found confusing,
• 1:50:47
  Speaker 2: so I'd like to alleviate the dissonance, is I thought
• 1:50:50
  Speaker 2: you were saying earlier that you didn't think the job
• 1:50:52
  Speaker 2: disruption promises were real.
• 1:50:55
  Speaker 1: No, what I was saying is that when we talk
• 1:50:59
  Speaker 1: about what these executives predict about the future, we need
• 1:51:04
  Speaker 1: to understand that they are ultimately trying to influence the
• 1:51:08
  Speaker 1: public in a way that allows them to continue maintaining
• 1:51:11
  Speaker 1: control over the technology.
• 1:51:13
  Speaker 2: But objectively, do you think that the job disruption that
• 1:51:15
  Speaker 2: they talk about.
• 1:51:16
  Speaker 3: Do you think this is real?
• 1:51:20
  Speaker 1: I don't want to comment specifically on this chart, but
• 1:51:23
  Speaker 1: we've already seen in job reports that there is a
• 1:51:25
  Speaker 1: restructuring of the economy happening right now. But going back
• 1:51:29
  Speaker 1: to the data science, so this supercomputer facility, it's a
• 1:51:33
  Speaker 1: meta supercomputer facility, is being built in Louisiana. And it
• 1:51:37
  Speaker 1: would be four times the size of the Abilene, Texas one.
• 1:51:41
  Speaker 1: and use half of the average power demand of New
• 1:51:44
  Speaker 1: York City. So it's one-fifth the size of Manhattan. This
• 1:51:46
  Speaker 1: makes it seem like almost all of Manhattan, but it
• 1:51:49
  Speaker 1: would be one-fifth the size of Manhattan. When these facilities
• 1:51:52
  Speaker 1: go into these communities, what happens? Power utility increases, grid
• 1:51:57
  Speaker 1: reliability decreases. The facilities also need fresh water to generate
• 1:52:03
  Speaker 1: the power for powering them as well as fresh water
• 1:52:06
  Speaker 1: to cool. And there have been lots of documented stories
• 1:52:09
  Speaker 1: of communities that are already really constrained in their fresh
• 1:52:12
  Speaker 1: water resource. They're under a drought when a facility comes
• 1:52:14
  Speaker 1: in and then there are people, the community is actually
• 1:52:17
  Speaker 1: like competing with this facility for fresh water. I talk
• 1:52:20
  Speaker 1: about one of those communities in my book. And also
• 1:52:23
  Speaker 1: sometimes these facilities, instead of connecting to the grid, They instead,
• 1:52:28
  Speaker 1: a power plant pops up next to it. So in Memphis, Tennessee,
• 1:52:33
  Speaker 1: where Musk built Colossus, the supercomputer for training Grok, he
• 1:52:37
  Speaker 1: used 35 methane gas turbines to power the facility. This
• 1:52:42
  Speaker 1: is a working class community, a black and brown community,
• 1:52:45
  Speaker 1: a rural community that was not even told that they
• 1:52:49
  Speaker 1: would be the hosts of this facility. And they discovered
• 1:52:53
  Speaker 1: it because they literally smelled what seemed like a gas
• 1:52:57
  Speaker 1: leak in all of their living rooms. And that's when
• 1:53:00
  Speaker 1: they discovered that these methane gas turbines were taking away
• 1:53:05
  Speaker 1: their right to clean air. And this is a community
• 1:53:08
  Speaker 1: that's already been facing a history of environmental racism.
• 1:53:12
  Speaker 3: They had already had lots of struggles to access their
• 1:53:15
  Speaker 3: right to clean air.
• 1:53:17
  Speaker 1: And now there's this huge supercomputer that's landed in their
• 1:53:22
  Speaker 1: midst that is pumping thousands of tons of toxins into
• 1:53:26
  Speaker 1: their air exacerbating the asthmatic symptoms of the children, exacerbating
• 1:53:32
  Speaker 1: the respiratory illnesses of other people. It's one of the
• 1:53:36
  Speaker 1: communities that has the highest rates of.
• 1:53:39
  Speaker 3: Lung cancer. And so.
• 1:53:42
  Speaker 2: And that supercomputer is taking their jobs.
• 1:53:45
  Speaker 1: And then they also have supercomputers taking their jobs. So
• 1:53:47
  Speaker 1: this is what I mean. It's like the haves and
• 1:53:49
  Speaker 1: have-nots are fundamentally being pulled apart even further. Like... If
• 1:53:56
  Speaker 1: you in this version of Silicon Valley's future are in
• 1:54:01
  Speaker 1: the misfortunate category of being a have-not, we are talking
• 1:54:06
  Speaker 1: about you now getting a job that is way worse
• 1:54:10
  Speaker 1: than what you had because you might be doing data annotation.
• 1:54:14
  Speaker 1: And you might be treated as a machine rather than
• 1:54:16
  Speaker 1: as a human to extract value, the value of your
• 1:54:19
  Speaker 1: labor for perpetuating this labor automating machine that these people
• 1:54:24
  Speaker 1: are building. You might be competing with these facilities for
• 1:54:27
  Speaker 1: freshwater resources.
• 1:54:29
  Speaker 3: They're also polluting your air.
• 1:54:31
  Speaker 1: Your bills have increased, so the affordability crisis is getting worse. Like,
• 1:54:37
  Speaker 1: how is that making people able to be more human?
• 1:54:41
  Speaker 2: What do we do about it? Yes. Okay.
• 1:54:46
  Speaker 1: So one of the analogies that I always use is
• 1:54:49
  Speaker 1: AI is like the word transportation. Transportation can literally refer
• 1:54:52
  Speaker 1: to everything from a bicycle to a rocket. And we
• 1:54:56
  Speaker 1: have nuanced conversations about transportation where we always say we
• 1:55:00
  Speaker 1: need to transition our transportation towards more efficient, uh, sustainable options.
• 1:55:06
  Speaker 1: We need to transition towards, you know, public transport, electric vehicles.
• 1:55:10
  Speaker 1: And we know we don't ever say everyone should get
• 1:55:13
  Speaker 1: a rocket to do every.
• 1:55:16
  Speaker 3: To serve all of their transportation needs, right? Like we're
• 1:55:18
  Speaker 3: in Austin.
• 1:55:19
  Speaker 1: If you use a rocket to fly from Dallas to Austin,
• 1:55:21
  Speaker 1: like that would just make not no sense. It's just
• 1:55:24
  Speaker 1: a disproportionate use of resources to get the benefit of
• 1:55:29
  Speaker 1: getting from point A to point B. This is how
• 1:55:31
  Speaker 1: we should think about AI. So all of the models
• 1:55:34
  Speaker 1: that we've been talking about, I like to think of
• 1:55:36
  Speaker 1: them as the rockets of AI. They use an extraordinary
• 1:55:40
  Speaker 1: amount of resources and they provide benefit, some dramatic benefit
• 1:55:44
  Speaker 1: to some people, but they're also exacting an extraordinary cost
• 1:55:48
  Speaker 1: on a large swath of people because of the costs
• 1:55:53
  Speaker 1: of developing this technology. Why don't we build more bicycles
• 1:55:59
  Speaker 1: of AI? This is things like DeepMind's AlphaFold, which is
• 1:56:03
  Speaker 1: a system that predicts how proteins will fold based on
• 1:56:07
  Speaker 1: amino acid sequences. It's really important for accelerating drug discovery,
• 1:56:12
  Speaker 1: for understanding human disease. And it won the Nobel prize
• 1:56:16
  Speaker 1: in chemistry in 2024. And the reason why it's a
• 1:56:19
  Speaker 1: bicycle of AI is because you're using small.
• 1:56:23
  Speaker 3: Curated data sets.
• 1:56:25
  Speaker 1: You just have data that has amino acid sequences and
• 1:56:29
  Speaker 1: protein folding. So that means you need significantly less computational
• 1:56:35
  Speaker 1: resources to develop the system, which means significantly less energy,
• 1:56:38
  Speaker 1: which means less emissions, so on and so forth. And
• 1:56:40
  Speaker 1: you're providing enormous benefit to people.
• 1:56:43
  Speaker 2: It feels like the horse has left the stable in
• 1:56:47
  Speaker 2: this regard because they've already taken people's IP. They've taken media.
• 1:56:52
  Speaker 2: They train on this podcast. We know they do because
• 1:56:54
  Speaker 2: it shows that they do. I think there's a button
• 1:56:56
  Speaker 2: actually in the back end of YouTube now that allows
• 1:56:58
  Speaker 2: you just to click it, and it says, we will
• 1:57:00
  Speaker 2: train on your YouTube channel. So the horse has kind
• 1:57:03
  Speaker 2: of left the station.
• 1:57:04
  Speaker 3: Here's the thing.
• 1:57:05
  Speaker 1: If the horse truly had left the stables, they wouldn't
• 1:57:07
  Speaker 1: have to train on anything anymore. Why is it that
• 1:57:10
  Speaker 1: their appetite for data has actually expanded? It's because in
• 1:57:15
  Speaker 1: order to build the next generations of their technologies, in
• 1:57:17
  Speaker 1: order to have the technologies continue to be relevant and
• 1:57:21
  Speaker 1: continue to thrive, update with the pace of new knowledge
• 1:57:25
  Speaker 1: creation and society's evolvement. They need to train again and
• 1:57:30
  Speaker 1: again and again and again. And why are they employing
• 1:57:33
  Speaker 1: actually more and more and more data annotation workers over time?
• 1:57:37
  Speaker 1: It's because they need more and more of that work
• 1:57:40
  Speaker 1: over time. I mean, I've been reporting on data annotation
• 1:57:44
  Speaker 1: work for over seven years now, and it's not gone down.
• 1:57:48
  Speaker 3: It's gone, it's increased.
• 1:57:50
  Speaker 2: Do you think there's any chance of it going down?
• 1:57:53
  Speaker 2: Do you think there's any chance of this sort of
• 1:57:55
  Speaker 2: brute force scaling approach where you take data, you take
• 1:57:58
  Speaker 2: computational power, energy, and you have the data labelers and
• 1:58:05
  Speaker 2: building out more and more parameters for the models? Do
• 1:58:07
  Speaker 2: you think there's any chance it's going to stop? or
• 1:58:10
  Speaker 2: go in a different direction other than the one that's
• 1:58:11
  Speaker 2: going in now?
• 1:58:12
  Speaker 1: I would love to reframe the question and say, what
• 1:58:15
  Speaker 1: should we be doing in this moment where it's not
• 1:58:17
  Speaker 1: going down, where we do recognize that actually these companies
• 1:58:21
  Speaker 1: in this moment need continued resources, inputs, and labor to
• 1:58:26
  Speaker 1: perpetuate what they are doing?
• 1:58:28
  Speaker 2: Yeah, because this sounds like stop. And I just feel
• 1:58:31
  Speaker 2: like stop is like a hard, it feels like, I
• 1:58:34
  Speaker 2: just think, you know, with the government in place, they're
• 1:58:36
  Speaker 2: supporting these companies like crazy. Globally, this is happening. So
• 1:58:39
  Speaker 2: I'm like, stop doesn't feel.
• 1:58:41
  Speaker 1: I always say we need to break up the empire
• 1:58:43
  Speaker 1: and we need to develop alternatives. And we are already
• 1:58:47
  Speaker 1: seeing a flourishing of incredible grassroots movements that are applying
• 1:58:51
  Speaker 1: an enormous amount of pressure to the way that the
• 1:58:54
  Speaker 1: empire is trying to unfold its agenda.
• 1:58:57
  Speaker 2: Mm-hmm.
• 1:58:57
  Speaker 1: 80% of Americans in the most recent poll think that
• 1:59:01
  Speaker 1: the AI industry needs to be regulated.
• 1:59:03
  Speaker 2: Yeah.
• 1:59:03
  Speaker 1: When was the last time that 80% of Americans were
• 1:59:06
  Speaker 1: on the same side of an issue?
• 1:59:07
  Speaker 2: No, yeah. When I have these conversations on the podcast,
• 1:59:09
  Speaker 2: the comments section are clear.
• 1:59:10
  Speaker 3: Yeah.
• 1:59:11
  Speaker 2: There's no disagreement. There's no one in there going, oh no,
• 1:59:13
  Speaker 2: I think they should crack on.
• 1:59:14
  Speaker 3: Yeah.
• 1:59:16
  Speaker 1: Dozens of protests against data centers have broken out all
• 1:59:19
  Speaker 1: around this country.
• 1:59:20
  Speaker 3: And the US, all around the world.
• 1:59:22
  Speaker 2: So what do we do about it?
• 1:59:23
  Speaker 1: So these are people that are doing something about it.
• 1:59:27
  Speaker 1: they are actually reasserting their agency and exercising democratic contestation
• 1:59:33
  Speaker 1: against the ways that the empires are going about their business.
• 1:59:36
  Speaker 2: What goal should we be aiming at? So if I
• 1:59:38
  Speaker 2: said to my audience, Janet at home, because this is
• 1:59:40
  Speaker 2: kind of what I see in the comments, it's hopelessness.
• 1:59:42
  Speaker 2: It's like, what can I do? I'm just.
• 1:59:44
  Speaker 1: Yeah, well, the goal is not that we completely get
• 1:59:48
  Speaker 1: rid of this technology. The goal is that these companies
• 1:59:51
  Speaker 1: need to stop being empires. And the way I define
• 1:59:53
  Speaker 1: like a typical business versus an empire is... is that
• 1:59:55
  Speaker 1: the empires are predicated on this idea that they do
• 1:59:58
  Speaker 1: not have to provide a fair exchange of value with
• 2:00:01
  Speaker 1: the workers who work for them or the people who
• 2:00:03
  Speaker 1: use them or all of the other people that are
• 2:00:05
  Speaker 1: involved in the supply chain of producing and deploying these technologies.
• 2:00:09
  Speaker 1: They can extract and exploit and extract and exploit and
• 2:00:11
  Speaker 1: get more value than what they offer. Whereas typical businesses,
• 2:00:16
  Speaker 1: there is a fair exchange. You buy a service, you
• 2:00:19
  Speaker 1: feel like you got the same amount of value as
• 2:00:21
  Speaker 1: the service that you provided, but like for these data
• 2:00:23
  Speaker 1: annotation workers, for example, they do not feel in any
• 2:00:25
  Speaker 1: way that they're being paid the same value that they
• 2:00:27
  Speaker 1: provide to these companies. So that's like, for me, the
• 2:00:30
  Speaker 1: North star is like, we should be pushing back and
• 2:00:34
  Speaker 1: holding accountable these companies when they operate in an Imperial way.
• 2:00:40
  Speaker 1: And that's what we've seen with all of these people
• 2:00:43
  Speaker 1: that are now literally protesting the streets against data centers
• 2:00:45
  Speaker 1: and having an enormous effect, by the way, actually stalling
• 2:00:49
  Speaker 1: data center projects and also completely banning data centers from
• 2:00:53
  Speaker 1: being developed in their localities. We're seeing that with artists
• 2:00:56
  Speaker 1: and writers that are suing these companies for intellectual property
• 2:00:59
  Speaker 1: infringement and creating a huge public conversation about what is
• 2:01:03
  Speaker 1: it that we actually, how do we actually want to
• 2:01:06
  Speaker 1: protect our intellectual property? Three weeks ago, I met Megan Garcia,
• 2:01:12
  Speaker 1: who is the mother of Sewell Setzer III, who is
• 2:01:16
  Speaker 1: the 14-year-old who died by suicide after suicide. being sexually
• 2:01:20
  Speaker 1: groomed by a character he has trapped on. And she,
• 2:01:25
  Speaker 1: when that happened, I mean, obviously was incredibly devastated by
• 2:01:31
  Speaker 1: what had happened to her son. She also decided to
• 2:01:34
  Speaker 1: do something about it. She sued the companies and that
• 2:01:37
  Speaker 1: lawsuit then sparked many other parents and families who were
• 2:01:41
  Speaker 1: actually experiencing similar things to sue these companies as well.
• 2:01:45
  Speaker 1: That has created an enormous public conversation about what these
• 2:01:50
  Speaker 1: companies are actually doing when they exploit and they extract.
• 2:01:54
  Speaker 1: What is the cost to the lives of people around
• 2:01:58
  Speaker 1: the world, including children?
• 2:02:00
  Speaker 2: So what do you think my audience should do? If
• 2:02:02
  Speaker 2: they agree with everything written in your book, Empire of AI,
• 2:02:06
  Speaker 2: Dreams and Nightmares, and Sam Altman's OpenAI, if they agree
• 2:02:10
  Speaker 2: with everything said here, if they agree with everything we've
• 2:02:12
  Speaker 2: discussed today, they're concerned about their kids, they don't want
• 2:02:15
  Speaker 2: everyone to become data labelers, they don't think that's a particularly...
• 2:02:19
  Speaker 2: Great solution. What can they actually go and do?
• 2:02:22
  Speaker 1: When I was writing the book, the only discourse that
• 2:02:24
  Speaker 1: was happening was, this is the best thing since sliced bread.
• 2:02:28
  Speaker 1: Because of all of the actions of these people, like
• 2:02:32
  Speaker 1: saying when they're not happy with the things that these
• 2:02:36
  Speaker 1: companies are doing, we now have 80% of Americans that
• 2:02:39
  Speaker 1: want to regulate this industry. And so I would say
• 2:02:41
  Speaker 1: to people, think about all the ways that your life
• 2:02:44
  Speaker 1: intersects with the resources And that the air industry needs
• 2:02:49
  Speaker 1: to perpetuate what they do. And also the spaces that
• 2:02:52
  Speaker 1: they would need to deploy these technologies to continue having
• 2:02:56
  Speaker 1: broad-based adoption in.
• 2:02:59
  Speaker 3: Their work.
• 2:03:00
  Speaker 1: So you're a data donor to these companies. You could
• 2:03:05
  Speaker 1: withhold that data. And that's what those artists and writers
• 2:03:07
  Speaker 1: are doing. Like they're suing these companies to withhold, to
• 2:03:10
  Speaker 1: try and create mechanisms by which that data would then
• 2:03:13
  Speaker 1: be withheld. You probably have a data center popping up
• 2:03:16
  Speaker 1: around you. If you're at a school environment or a
• 2:03:20
  Speaker 1: company environment, you're probably having a discussion in those environments
• 2:03:23
  Speaker 1: right now about what should the AI adoption policy be.
• 2:03:27
  Speaker 1: And these companies, they like, I was talking with some
• 2:03:31
  Speaker 1: OpenAI employees just the other day, and they were telling
• 2:03:34
  Speaker 1: me that it's understood internally that the revenue targets for
• 2:03:39
  Speaker 1: the company are extraordinary. And they need things to go
• 2:03:44
  Speaker 1: flawlessly for it to all work out. And so they
• 2:03:49
  Speaker 1: would need every single person to adopt this, every single
• 2:03:53
  Speaker 1: space to adopt this. They would need to be able
• 2:03:55
  Speaker 1: to build their data centers at the speed that they're
• 2:03:57
  Speaker 1: trying to build them. And so what I would say
• 2:03:59
  Speaker 1: to everyone of your viewers is let's not make it
• 2:04:03
  Speaker 1: go flawlessly if we don't agree with what they are doing.
• 2:04:06
  Speaker 2: Ah, okay, I got you.
• 2:04:08
  Speaker 1: And then let's build alternatives because the thing is, what
• 2:04:12
  Speaker 1: I'm saying is not that these technologies don't have utility.
• 2:04:16
  Speaker 3: It's that specifically the political.
• 2:04:18
  Speaker 1: Economy that has emerged to support the production of these
• 2:04:21
  Speaker 1: technologies right now is exacting a lot of harm on people.
• 2:04:25
  Speaker 1: But we have research that shows that that the very
• 2:04:28
  Speaker 1: same capabilities could be developed with much more efficient methods,
• 2:04:34
  Speaker 1: with much less resource consumption. And we have a lot
• 2:04:38
  Speaker 1: of different other AI systems at our disposal that are
• 2:04:40
  Speaker 1: like the bicycles of AI that we also know provide
• 2:04:44
  Speaker 1: extraordinary benefit at very little cost. So let's break up
• 2:04:47
  Speaker 1: the empire and let's forge new paths of AI development
• 2:04:50
  Speaker 1: that are broadly beneficial to everyone.
• 2:04:53
  Speaker 2: It's strange. I think I'm... I've trained myself to deal
• 2:04:58
  Speaker 2: with dichotomies in my head. And this for me is such,
• 2:05:01
  Speaker 2: is a dichotomy where I, as a CEO and as
• 2:05:04
  Speaker 2: a founder, as an entrepreneur and someone that loves technology,
• 2:05:06
  Speaker 2: I think it's incredible. It's absolutely incredible AI. It's just
• 2:05:09
  Speaker 2: so amazing and incredible. The things it's enabled me to
• 2:05:12
  Speaker 2: do and create.
• 2:05:13
  Speaker 3: No, because it's designed to enable people like you.
• 2:05:16
  Speaker 2: And my car driving in the morning and being safer, incredible. Um,
• 2:05:23
  Speaker 2: I think, you know, the billion odd people that use
• 2:05:25
  Speaker 2: AI tools or chat GPT or whatever it might be,
• 2:05:27
  Speaker 2: they'd probably say that it's added value to their life. But,
• 2:05:30
  Speaker 2: and this is the part that people find confusing that
• 2:05:31
  Speaker 2: you can, and I like, I invest in companies that are,
• 2:05:34
  Speaker 2: you know, heavily using AI. But, and the big but is,
• 2:05:37
  Speaker 2: is it possible to think that is true and also
• 2:05:39
  Speaker 2: think that there are significant unintended consequences, which technology and
• 2:05:44
  Speaker 2: the history of technology should have taught us to take
• 2:05:46
  Speaker 2: a moment to pause to talk about?
• 2:05:48
  Speaker 1: I think this is absolutely, like, you can have both
• 2:05:52
  Speaker 1: of these things in your head. And what I'm saying
• 2:05:54
  Speaker 1: is that this tension doesn't have to be a tension,
• 2:05:57
  Speaker 1: because we could actually preserve the utility and benefits of
• 2:06:02
  Speaker 1: these technologies, but actually develop and design them in a
• 2:06:05
  Speaker 1: different way that doesn't have all of these unintended consequences.
• 2:06:08
  Speaker 2: Yes. And I think there needs to be a big
• 2:06:10
  Speaker 2: social conversation, which is why I have so many conversations
• 2:06:12
  Speaker 2: about AI on the show, like, There needs to be
• 2:06:14
  Speaker 2: a big social conversation about being intentional about the social impact.
• 2:06:20
  Speaker 2: The social and environmental impact. And that conversation is not
• 2:06:22
  Speaker 2: being had in government from what I can see. The
• 2:06:26
  Speaker 2: conversation takes place in the industry. And actually trying to
• 2:06:29
  Speaker 2: pull it out of the industry and open people's minds
• 2:06:32
  Speaker 2: to it is hopefully what we've been doing over the
• 2:06:34
  Speaker 2: last couple of months with this subject.
• 2:06:35
  Speaker 3: I think it's actually been... happening everywhere outside of the industry.
• 2:06:41
  Speaker 1: And for local governments and state level governments, there have
• 2:06:44
  Speaker 1: been huge conversations about this. Everywhere, like I've been on
• 2:06:47
  Speaker 1: book tour, I've been to dozens of cities around the world.
• 2:06:51
  Speaker 1: People are having these crucial conversations everywhere. I have not
• 2:06:56
  Speaker 1: gone to a single city.
• 2:06:57
  Speaker 2: Yeah, it's everywhere. Even here in South Byrd.
• 2:06:59
  Speaker 1: Yeah, I haven't gone to a single city where the
• 2:07:01
  Speaker 1: room is not packed and people are not wrestling with
• 2:07:04
  Speaker 1: the same exact questions as every other person in every
• 2:07:06
  Speaker 1: other room that I've been in.
• 2:07:08
  Speaker 2: Speaking of packed rooms, I know you've got to go
• 2:07:10
  Speaker 2: because you've got to talk today. So we've got our
• 2:07:13
  Speaker 2: last question, which is the closing tradition on this podcast.
• 2:07:15
  Speaker 2: How would your advice to a friend with a terminal
• 2:07:17
  Speaker 2: diagnosis differ from what you would do yourself? That's a
• 2:07:23
  Speaker 2: great question.
• 2:07:24
  Speaker 3: Differ from what you would do yourself? Oh, my God.
• 2:07:28
  Speaker 3: I would tell them, like, enjoy, like, live life for yourself.
• 2:07:33
  Speaker 2: You wouldn't do it.
• 2:07:34
  Speaker 3: And take it easy. And, yeah, I.
• 2:07:38
  Speaker 3: I am not taking it easy.
• 2:07:39
  Speaker 2: Well, I think it's a good thing you're not taking
• 2:07:41
  Speaker 2: it easy because you're leading a conversation, which is incredibly important.
• 2:07:44
  Speaker 2: And I think that's the thing. I think the conversation
• 2:07:46
  Speaker 2: is the important thing. And so, you know, because of
• 2:07:49
  Speaker 2: algorithms and echo chambers, it's so rare to have a
• 2:07:51
  Speaker 2: conversation these days, especially a long form one I agree
• 2:07:55
  Speaker 2: like this so I think they're so important and your
• 2:07:56
  Speaker 2: book is for anyone that's curious about I think a
• 2:08:01
  Speaker 2: lot of people would have learned a lot of stuff
• 2:08:02
  Speaker 2: today because I sit here with and interview AI people
• 2:08:04
  Speaker 2: all the time and I've learned so much today from
• 2:08:06
  Speaker 2: reading your book and the extensive objective perspective that your
• 2:08:10
  Speaker 2: book takes you're able to unravel all of these stories
• 2:08:13
  Speaker 2: that we sometimes see in tweets and we don't know
• 2:08:14
  Speaker 2: if they're true or not because you've gone and met
• 2:08:16
  Speaker 2: the people and you've done your research and you're an
• 2:08:18
  Speaker 2: incredibly intelligent person extremely intelligent person who clearly has humanity's
• 2:08:24
  Speaker 2: interests as your North Star. And that shows up in
• 2:08:27
  Speaker 2: everything you do and everything you say. So please continue
• 2:08:29
  Speaker 2: to fight in the way that you are because it's
• 2:08:32
  Speaker 2: an incredibly important one. It's people like you that are,
• 2:08:34
  Speaker 2: I think, galvanizing the world to take the collective action
• 2:08:40
  Speaker 2: that we're starting to see everywhere. Empire of AI, Dreams
• 2:08:44
  Speaker 2: and Nightmares in Sam Altman's OpenAI by Karen Howe. I'll
• 2:08:47
  Speaker 2: link it below for anyone that wants to read this book.
• 2:08:49
  Speaker 2: I highly recommend you do. It's a New York Times
• 2:08:50
  Speaker 2: bestseller for good reason. Karen, thank you.
• 2:08:53
  Speaker 3: Thank you so much, Stephen.
• 2:08:55
  Speaker 2: YouTube have this new crazy algorithm where they know exactly
• 2:08:57
  Speaker 2: what video you would like to watch next based on
• 2:09:00
  Speaker 2: AI and all of your viewing behavior. And the algorithm
• 2:09:03
  Speaker 2: says that this video is the perfect video for you.
• 2:09:07
  Speaker 2: It's different for everybody looking right now. Check this video out.
• 2:09:10
  Speaker 2: I bet you, you might love it.