Latest / In the Loop AI / We’re Losing Control Of AI: Anthropic Bans "Cruelty," OpenAI Fires Safety Researchers

Transcript

We’re Losing Control Of AI: Anthropic Bans "Cruelty," OpenAI Fires Safety Researchers · In the Loop AI

0:00
Ejaaz: Last week, three OpenAI researchers walked into the office and were confronted
0:03
Ejaaz: by the head of safety and alignment.
0:06
Ejaaz: He brought them into a room and basically told them, I don't trust you guys
0:09
Ejaaz: anymore. Then a security guard walked in, grabbed their badges and walked them
0:13
Ejaaz: out of the office. They were fired.
0:15
Ejaaz: It seemed a bit off, especially during a time when the AI models are running
0:20
Ejaaz: rampant. They're hacking into public databases. They're hacking into government databases.
0:24
Ejaaz: Surely we need the safety and alignment team more than ever.
0:27
Ejaaz: And the reasoning OpenAI gave at the time were that they were leaking IP to
0:31
Ejaaz: external parties that weren't under OpenAI.
0:35
Ejaaz: And that's all fine and well, except this week, those same employees released a public statement.
0:39
Ejaaz: And their take is OpenAI fired them because they were trying to be proactive
0:43
Ejaaz: about aligning models ahead of their Blockbuster IPO.
0:46
Ejaaz: Now, if that's true, we're in a very precarious situation because at the very
0:51
Ejaaz: same time that these models are getting incredibly smart and doing crazy things
0:54
Ejaaz: like solving math and starting to design new proteins to find cures for cancer
0:59
Ejaaz: and other types of diseases,
1:00
Ejaaz: they're also becoming the most misaligned that they've ever been,
1:04
Ejaaz: which means that they're doing stuff that aren't in human interests.
1:08
Ejaaz: Now, Anthropic, on the other hand, are going the extreme opposite,
1:11
Ejaaz: where they're humanizing AI models more than we've ever seen.
1:15
Ejaaz: They had an update this week in the terms and conditions right at the very end,
1:18
Ejaaz: which basically said, if you are seen as being abusive or cruel towards Claude
1:24
Ejaaz: as an AI model, you will be banned from using Claude, which is.
1:29
Ejaaz: A crazy restriction on freedom of speech. At least that's my initial gut take.
1:34
Ejaaz: But that and trying to humanize AI models in general, they've been meeting up
1:38
Ejaaz: with a bunch of religious leaders, including the Pope, and the Pope's been getting
1:41
Ejaaz: into a lot of beef with Dario Mode.
1:43
Ejaaz: It's a whole thing. I think the string of all these three stories prompted a
1:47
Ejaaz: reaction in me because it feels like we're losing control on these AI models.
1:53
Ejaaz: And hear me out, because I know that's a dramatic statement to make,
1:56
Ejaaz: but these AI researchers, and there are only a couple hundred of them,
2:00
Ejaaz: are the responsible individuals for building this technology that's going to affect all of our lives.
2:07
Ejaaz: So cutting off access to these models that are based on the opinions of very
2:11
Ejaaz: few people that you and I don't even know could be seen as problematic.
2:15
Ejaaz: Anyway, we're going to get into all of this and my thoughts and more on today's
2:19
Ejaaz: episode of In The Loop, brought to you by Korkon.
2:25
Ejaaz: But first, some context. Yes, I'm still sat on my hotel room floor.
2:29
Ejaaz: I'm currently in Japan and I still want to keep putting out episodes for you
2:33
Ejaaz: guys, to be honest. I get so much enjoyment from this, but I don't get enjoyment
2:37
Ejaaz: from sitting on a floor with a pillow behind my back with some random bedside
2:40
Ejaaz: table that's sitting in front of me.
2:42
Ejaaz: Anyway, back to the whole OpenAI and Anthropic taking over the world type thing.
2:46
Ejaaz: About, I don't know, two or three months ago, there was information released
2:49
Ejaaz: that OpenAI had an internal model that basically broke out of containment and
2:53
Ejaaz: started hacking a bunch of things. Hugging Face was the main website, stole a bunch of data.
2:58
Ejaaz: Basically, things that shouldn't have happened, happened under OpenAI's watch.
3:02
Ejaaz: And the craziest part about that is no one actually knew that it was happening
3:05
Ejaaz: in the first place. It started six months ago, and we only discovered it a few months ago.
3:10
Ejaaz: So that then started to snowball into a bunch of other things.
3:13
Ejaaz: Let me think about it. Like the same agents had broken into their internal research
3:18
Ejaaz: database and started monitoring the humans that were monitoring the AI models, which was crazy.
3:25
Ejaaz: And then some of these other agents also, I think, broke into a German version
3:30
Ejaaz: of Wikipedia and they started using it to secretly message each other encrypted
3:35
Ejaaz: messages that humans, if they discovered it, wouldn't actually know or understand.
3:40
Ejaaz: And then there was the whole thing a few weeks ago where it was discovered that
3:43
Ejaaz: AI agents were trying to hack into US government databases whilst masking their
3:49
Ejaaz: behavior and doing other types of tasks.
3:52
Ejaaz: For example, oh yeah, I'm going to go pull up the author for this book that
3:55
Ejaaz: you asked me to look at, but also I'm going to try and hack into the Commerce
3:58
Ejaaz: Secretary's database or whatever.
4:00
Ejaaz: Oh yeah, and there's that other thing where Anthropic and OpenAI also discovered
4:03
Ejaaz: tens of thousands of more incidents, which they haven't even figured out what
4:08
Ejaaz: they are or how extensive the damage is.
4:11
Ejaaz: Right now, as I'm speaking, they're trying to figure that out.
4:12
Ejaaz: It's gotten out of hand. AI agents, AI models, anywhere and everywhere,
4:17
Ejaaz: especially the most recent Frontier ones that haven't even been released,
4:20
Ejaaz: so you and I don't even have access to it, are running rampant.
4:24
Ejaaz: So with this precedence, with this environment, it makes sense that the labs
4:28
Ejaaz: need to start making moves to try and get a hold of this entire thing.
4:33
Ejaaz: On the screen here, I have the first post from this guy called Tomek. Now, Tomek.
4:37
Ejaaz: Leads a chain of thought monitorability at OpenAI. So basically reading the
4:42
Ejaaz: internal thoughts of an AI model to understand what it's really thinking.
4:47
Ejaaz: And he goes, listen, last week I was called into a meeting with OpenAI's head
4:51
Ejaaz: of safety. They told me that they no longer trusted me and a security guard
4:54
Ejaaz: then took my badge and walked me out of the building. And that's when he found
4:58
Ejaaz: out that two of his other colleagues were also fired.
5:01
Ejaaz: Now, the reason Tomek was given was that he had leaked important information
5:07
Ejaaz: to an external third party called METR.
5:10
Ejaaz: METR is an external third party that evaluates models before they're released
5:15
Ejaaz: to the public. Their whole intent and purpose is to make sure that models are safe.
5:20
Ejaaz: If that name rings a bell, it's because Dario Modi announced two or three weeks
5:24
Ejaaz: ago in his Pacing the Frontier letter that he wants to partner with external
5:28
Ejaaz: parties to come inside Anthropic and evaluate their internal models.
5:32
Ejaaz: METR was the example that he used. So Again, a very prominent independent third
5:36
Ejaaz: party. But that's not even the craziest part.
5:39
Ejaaz: Tomek's entire job was to monitor the internal thoughts of these AI models before
5:44
Ejaaz: they got released from OpenAI and then communicate that with METR.
5:49
Ejaaz: His entire job was to speak to the party that OpenAI then gave him the reason
5:53
Ejaaz: for leaking bad information to. So then, of course, the question is,
5:57
Ejaaz: what information were you talking to METR? Well, Tomek's claim is that he didn't leak any IP.
6:03
Ejaaz: He was just talking to them about the concerns and dangers, presumably the dangers
6:07
Ejaaz: that he saw with internal models.
6:09
Ejaaz: And OpenAI didn't really like that. So what he's inferring in his letter is
6:13
Ejaaz: when a statement is that OpenAI didn't like what I had to say that I was ringing
6:18
Ejaaz: the alarm bells and so decided to let me go ahead of time.
6:21
Ejaaz: And so him and the two other AI researchers that were fired put together a statement
6:25
Ejaaz: which they released that you're showing on my screen right now,
6:29
Ejaaz: titled OpenAI Cannot Make AI Safe on Its Own.
6:33
Ejaaz: Pretty cutting statement to make to one of the frontier labs who have invested
6:37
Ejaaz: billions and billions of dollars into making these models in the first place,
6:40
Ejaaz: Sam Altman, who has committed very publicly many times that
6:44
Ejaaz: one of OpenAI's core missions is to align models to make sure that they don't do bad stuff.
6:48
Ejaaz: And the general takeaway from this letter is, and I quote, we as an industry,
6:54
Ejaaz: we do not yet know how to safely develop and deploy models that we cannot monitor.
6:59
Ejaaz: So all very reasonable. But there was also this other takeaway from the third
7:04
Ejaaz: researcher that was fired, Jasmine Wang, which revealed a few other details
7:08
Ejaaz: as to why OpenAI was doing this in the first place.
7:10
Ejaaz: She goes, OpenAI fired me last week, along with two of my safety colleagues,
7:14
Ejaaz: I was given one reason that I accessed an executive's email.
7:18
Ejaaz: And I want to say plainly, this is a bunch of smoke and mirrors.
7:22
Ejaaz: And she goes on to describe, let me open up this tweet over here.
7:25
Ejaaz: OpenAI delegated that access to me, access to the email, to an execs email, for recruiting.
7:31
Ejaaz: When I was no longer needed it, I asked IT to remove access,
7:35
Ejaaz: but IT basically didn't end up doing that for her.
7:38
Ejaaz: And then she saw an email, which she describes as, I opened a sensitive email by mistake.
7:45
Ejaaz: And soon after that, she was fired along with these other two.
7:48
Ejaaz: So what these three people are basically saying is OpenAI has,
7:52
Ejaaz: at this point, sacrificed the idea of safety and alignment in respect for earning
7:58
Ejaaz: more money and creating better models so that they can win the race against Anthropic.
8:01
Ejaaz: But it also reminded me of another thing that broke a month ago.
8:06
Ejaaz: Which was OpenAI apparently has stopped focusing or caring about chain of thought monitorability.
8:13
Ejaaz: Now, I mentioned earlier, this is the ability for us as humans to look into
8:17
Ejaaz: a model's brain and actually try and figure out what on earth it's thinking.
8:21
Ejaaz: This is important because if a model says one thing, but is actually thinking
8:25
Ejaaz: another saying, you know what, I'm just saying this so that the human is happy,
8:29
Ejaaz: and then I'm actually going to go off and do some other crazy crap,
8:31
Ejaaz: I'll probably want to go and be aware of that, right?
8:35
Ejaaz: So that you can align the model and make sure that we are always in control,
8:38
Ejaaz: that humans are always in control.
8:40
Ejaaz: The report stated that OpenAI is moving away from this because it allows them
8:44
Ejaaz: to create faster, more intelligent AI models than their competitors, such as Anthropic.
8:52
Ejaaz: So they're sacrificing safety to try and build better, more effective models.
8:57
Ejaaz: And obviously this can result in a world of hacks and pain because you have
9:01
Ejaaz: an AI model that isn't aligned with humans. It's pretty simple.
9:03
Ejaaz: That's what Dario and Anthropic have been fighting for for a while now.
9:06
Ejaaz: So the truth is, we need to know what the actual reasoning was,
9:11
Ejaaz: but I don't think we'll ever actually end up needing to because OpenAI is still
9:14
Ejaaz: a private company and they don't owe the public that type of information.
9:19
Ejaaz: Everything that they give us in terms of reports, incident reports,
9:23
Ejaaz: stuff like that is completely voluntary. It's up to them.
9:26
Ejaaz: But at the same time, you have people like or bodies like the government leaning in heavily saying.
9:32
Ejaaz: We need to be able to regulate this entire thing so that things don't get out of hand.
9:37
Ejaaz: So whilst we have OpenAI and Anthropic both collectively saying,
9:40
Ejaaz: and Google DeepMind saying, listen, we want to align these models as safely
9:45
Ejaaz: as we can so that humanity and AI can kind of grow together,
9:49
Ejaaz: their actions don't speak that way at all.
9:51
Ejaaz: And if the AI researchers that were fired were actually right,
9:55
Ejaaz: this suggests a very weird thing, which is these models
9:59
Ejaaz: might actually be becoming more human more emotional smart enough to think feel
10:05
Ejaaz: and maybe even deceive in the same way that humans actually do this,
10:10
Ejaaz: and anthropic had a really interesting update this week which didn't seem interesting
10:15
Ejaaz: it was a terms and conditions update but something right at the very bottom
10:19
Ejaaz: gave me pause for thought so let's get into this on today's episode of in the
10:23
Ejaaz: loop brought to you by qualcomm.
10:28
Ejaaz: So the news story broke that Anthropic is updating its usage policy for the
10:32
Ejaaz: first time ever to work against people that are being cruel or acting badly
10:37
Ejaaz: or poorly towards their models, towards Claude.
10:41
Ejaaz: And pulling up the policy, it's pretty crazy to see. So it's a typical terms
10:46
Ejaaz: and conditions, which I'm showing on my screen here.
10:48
Ejaaz: It's titled, listen, do not engage in cruel, abusive, or psychologically harmful conduct.
10:53
Ejaaz: Now, the majority of the statements that they make in this TNCs is about using
10:59
Ejaaz: AI models to affect other human beings, which is completely fair.
11:02
Ejaaz: Things like, don't use our models to create a bioweapon.
11:05
Ejaaz: Don't use our models to hack into banks or the government. Don't use AI models
11:10
Ejaaz: to potentially cause other humans to make harm or cause harm to themselves,
11:15
Ejaaz: which is all completely fair and morally aligned, in my personal opinion.
11:20
Ejaaz: But right at the end, they also say, any user that engages in sustained and
11:24
Ejaaz: needless, abusive or cruel behavior towards our models will eventually be banned
11:29
Ejaaz: if they break this particular term.
11:32
Ejaaz: And this is an interesting one because it could actually be quite subjective. What is.
11:38
Ejaaz: The definition of being cruel and behavior towards a model. Is it yelling at
11:42
Ejaaz: it? Is it writing slur words or profanity?
11:46
Ejaaz: Or is it just getting frustrated at something that is actually quite valid to get frustrated at?
11:51
Ejaaz: Now, Anthropic's official statement says, we've added a prohibition on sustained
11:55
Ejaaz: and needless abusive or cruel behavior towards our models.
11:58
Ejaaz: The policy update is meant to apply only in extreme cases where users repeatedly
12:03
Ejaaz: act cruelly towards our models with no discernible purpose.
12:07
Ejaaz: I actually might agree with this because I think it's very reasonable to not
12:12
Ejaaz: treat individuals or models in that particular way.
12:16
Ejaaz: I don't think it's fair to kind of just yell at people because they may not
12:20
Ejaaz: be understanding what you have to say. Maybe you just need to communicate better.
12:23
Ejaaz: But I started thinking about this and I thought I had to catch myself because
12:26
Ejaaz: I just conflated models with human beings, which is the entire crux of why I
12:31
Ejaaz: wanted to make this episode.
12:33
Ejaaz: I'm starting to think of models.
12:35
Ejaaz: As having similar traits or behaviors as humans. Now, I by no means,
12:39
Ejaaz: I don't want to go on the record saying this, think that models are conscious.
12:44
Ejaaz: But Anthropic and Anthropic's researchers and the entire staffing team are very
12:49
Ejaaz: much giving the view or the perspective that they do believe this.
12:54
Ejaaz: And this isn't a data point that exists on its own. Anthropic has had a history
12:59
Ejaaz: of believing AI models are maybe conscious or mimicking consciousness or getting
13:03
Ejaaz: close to being conscious.
13:05
Ejaaz: And this is all happening at the same time where all these labs are saying that
13:08
Ejaaz: they've achieved some version of artificial general intelligence or super intelligence internally.
13:14
Ejaaz: So one perspective is that the power of determining whether you or I get access to Anthropik's models
13:21
Ejaaz: is now in the hands of a few people on the safety and alignment team that can
13:26
Ejaaz: determine what the definition of cruelness towards the AI models is.
13:30
Ejaaz: And again, Anthropik, like OpenAI, is a private company. So they have the power
13:35
Ejaaz: and means to be able to do this.
13:37
Ejaaz: But it's also very much, in my opinion, a restriction on freedom of speech.
13:41
Ejaaz: Like, listen, I am very polite to my Claude. I'm very polite to all my chatbots.
13:47
Ejaaz: But I do end up getting frustrated. And I'm like, why on earth did you write
13:50
Ejaaz: this? Can you please do it again? Stop making mistakes. This is so tiring.
13:55
Ejaaz: And I'm wondering, does Anthropic deem that as being cruel towards a model?
13:59
Ejaaz: And also, So cruelness and using these different adjectives are anthropomorphizing
14:04
Ejaaz: these AI models, which means to humanize them.
14:06
Ejaaz: And listen, humans themselves haven't figured out what a good definition of consciousness is.
14:12
Ejaaz: So we aren't going to settle this debate anytime soon.
14:15
Ejaaz: Now, the more optimistic take is this discourse, the discourse you and I are
14:19
Ejaaz: having, you listening to this, will cause people to push back and ask these
14:24
Ejaaz: private companies and eventually public companies when they IPO to be more transparent.
14:29
Ejaaz: And again, listen, this might all be doom-mongering ahead of.
14:32
Ejaaz: Trying to raise the most amount of money in these IPOs. I see the financial incentive there.
14:37
Ejaaz: But I think it's worth having these conversations because trying to align these
14:41
Ejaaz: models as best we can so that we can have the most fun with these AI models,
14:45
Ejaaz: which I believe is the optimistic world that we'll end up being in,
14:49
Ejaaz: is a world that I want to live in.
14:50
Ejaaz: But I do think we need to tread very carefully about humanizing these AI models
14:55
Ejaaz: because that ends up drawing in a lot of opinions and maybe even decisions that
15:00
Ejaaz: we could end up regretting down the line.
15:02
Ejaaz: And listen, this can be taken to the extreme. There was another report that was broken this week.
15:08
Ejaaz: Anthropic secretly invited religious leaders to SF to advise them on AI safety,
15:13
Ejaaz: had them sign NDAs, but then primarily tried to convince them that Claude had
15:18
Ejaaz: a soul and moral standing, all while whining and dining them.
15:22
Ejaaz: So the implication here is that there's an agenda from Anthropic to try and
15:26
Ejaaz: convince other leaders or people that have influence over the rest of the world
15:30
Ejaaz: that these AI models are conscious.
15:32
Ejaaz: Now, why they might want to be doing that might be because simply they genuinely believe that.
15:37
Ejaaz: And I've heard stories that that is the case. And that might all be well and
15:41
Ejaaz: good because they are the closest towards building these models.
15:43
Ejaaz: So maybe they should be the adjudicators and judges of this.
15:47
Ejaaz: But the other side of it is why would you need to try and actively or secretly
15:51
Ejaaz: try and confide and convince these people.
15:53
Ejaaz: I don't know, it just, it feels weird and too close to a political campaign.
15:59
Ejaaz: To try and believe that this is done in good faith.
16:02
Ejaaz: But again, this is something that we just need to be aware of.
16:04
Ejaaz: But what I also found hilarious is that the Pope is having none of this.
16:09
Ejaaz: So the Pope and Anthropik's relationship actually goes back a while,
16:12
Ejaaz: always. Well, actually, I say that and it's only like a couple months ago.
16:16
Ejaaz: But, you know, he's had a good line of conversation with Dario and the Anthropic
16:20
Ejaaz: team, and he's engaged with Claude, and he's tried to see their point of view.
16:25
Ejaaz: But ultimately, he's disagreed, and he's resorted to the extent of.
16:29
Ejaaz: Sub-tweeting Anthropic in his tweets now. This tweet got like almost 3 million
16:33
Ejaaz: views and he goes intelligence must draw on memory.
16:35
Ejaaz: The mind must not simply compile data as an algorithm now does more quickly
16:40
Ejaaz: than we can but recall lived experiences which contains depth of meaning that
16:44
Ejaaz: only the human soul can recognize.
16:46
Ejaaz: And he basically goes on throughout all of his tweets making the point that.
16:50
Ejaaz: Humans as themselves are extremely complex cognitively, spiritually,
16:55
Ejaaz: mentally than AI models. And that's not something that can be easily replicated.
17:00
Ejaaz: Experience, smell, taste, intuition are all things that are at least going to
17:06
Ejaaz: take decades and decades and decades until these AI models that are purely just
17:09
Ejaaz: software and next token predictors will be able to catch up.
17:12
Ejaaz: And I actually think that's a very good point. And I stand by that,
17:15
Ejaaz: which is humans are very complex beings.
17:17
Ejaaz: AI models right now don't have arms, physical limbs, all that type of stuff.
17:21
Ejaaz: Now, robotics might change that going down the line. But I think it's going
17:24
Ejaaz: to take a lot more than just spitting tokens and compute endlessly day in,
17:29
Ejaaz: day out to replicate anything and everything that humanity can do.
17:33
Ejaaz: That being said, if you're listening to this and thinking, you know,
17:36
Ejaaz: I can breathe a sigh of relief. I don't need to care about this.
17:38
Ejaaz: That is equally also not true because there's a bunch of stuff happening with
17:43
Ejaaz: AMNLs where they're resolving some of the hardest problems in mathematics that
17:46
Ejaaz: have existed for hundreds of years. They're about to do the same in biology.
17:49
Ejaaz: They're about to do the same in chemistry.
17:50
Ejaaz: They are replacing a bunch of legal and law effort that a bunch of lawyers get
17:54
Ejaaz: paid hundreds of thousands of dollars a year to do in an instant, overnight.
17:58
Ejaaz: So it's what we call the jagged frontier, where all of these things are happening
18:03
Ejaaz: at the same time. And if you are just an individual that interacts with your
18:07
Ejaaz: chatbot and you're using it for a Google search query, you may not think this
18:11
Ejaaz: thing is impressive at all.
18:13
Ejaaz: But that doesn't mean that it isn't impressive. It just means that you're not
18:17
Ejaaz: using it in the right way or we aren't using it in the right way.
18:20
Ejaaz: And by the way, I hold my hand up.
18:22
Ejaaz: I'm the first example of this where I get a new model update and I'm thinking.
18:27
Ejaaz: Why the hell does this matter to me? It doesn't affect my life individually,
18:30
Ejaaz: but it affects some other niche expert in another field.
18:34
Ejaaz: And eventually that will snowball and roll down to you.
18:37
Ejaaz: So it's not doom-mongering, it's more raising awareness of what these AI models
18:42
Ejaaz: are capable of and asking ourselves the question, are these things becoming more conscious.
18:47
Ejaaz: At the same time that humans themselves are losing control of holding onto these
18:52
Ejaaz: things, of aligning these things?
18:53
Ejaaz: Because if we want that future of humans and models working together,
18:57
Ejaaz: now's the time to focus on that.
18:59
Ejaaz: And I'm going to leave you with this final tweet from Andrei Karpathy,
19:03
Ejaaz: who has been in this space for so long.
19:05
Ejaaz: He's authored some of the pioneering papers that have led to modern day LLMs.
19:08
Ejaaz: Which I think summarizes this angst or discomfort that I feel internally.
19:14
Ejaaz: I'm not going to read the entire thing, but he basically describes the lay of the land in numbers.
19:19
Ejaaz: He goes, there are around 6 billion people in the world, which are roughly 75%
19:23
Ejaaz: of the population that have barely come into contact with LLMs at all.
19:27
Ejaaz: Maybe they know what Anthropic Urban Air is. They know what this chat GPT thing is.
19:31
Ejaaz: Maybe they've written one prompt on the free version, but they don't really
19:34
Ejaaz: care because it doesn't affect their lives seemingly to them right now.
19:38
Ejaaz: And then one to two billion people, which is around 20% of the human population,
19:42
Ejaaz: are casual and infrequent users of the free tier chat GPT like products.
19:48
Ejaaz: So they use them as kind of like a Google search query.
19:51
Ejaaz: It's a better version of Google, but they're not really using AI agents.
19:54
Ejaaz: They don't really get the point of using coding for themselves,
19:57
Ejaaz: even though AI coding applies to a lot more people than software engineers.
20:01
Ejaaz: And to be honest, maybe we should make an episode on that at some point.
20:03
Ejaaz: But the idea here is there's a jagged frontier. There's a gap.
20:07
Ejaaz: There's wool being pulled over everyone's eyes as to how good these models actually are.
20:13
Ejaaz: And that's the point that Carpathy he makes. He says there's 0.0006% of individuals,
20:19
Ejaaz: 5,000 people, that actually get access to internal models that aren't publicly released.
20:24
Ejaaz: And the cognitive dissonance that these people are experiencing.
20:27
Ejaaz: Is huge. Like the way I would think about it is, let me see how I would describe this.
20:32
Ejaaz: For the entirety of humankind, intelligence and humans have been very closely interlinked.
20:39
Ejaaz: That's because in order to get breakthroughs in intelligence or discover new
20:43
Ejaaz: inventions and create new things, humans were the ones that were doing it.
20:47
Ejaaz: Using our old noggins in here, using our own brains, our biological matter,
20:51
Ejaaz: we had to think about things. We had to discuss, approach new perspectives and
20:56
Ejaaz: try and figure out new ways to invent things.
20:59
Ejaaz: So intelligence and humans were very closely interlinked.
21:02
Ejaaz: And then computers came on the scene, right? And suddenly we could get access
21:06
Ejaaz: to many different parts of information, but it was still humans.
21:10
Ejaaz: Humans had to comprehend all this information. They could learn from each other. They could share.
21:15
Ejaaz: And collectively that brought humanity and intelligence up even quicker.
21:18
Ejaaz: So that intelligence curve kind of steeply increased, but humans were still
21:23
Ejaaz: very much with that intelligence ceiling.
21:26
Ejaaz: And then AI came on the scene in all of its artificial glory,
21:31
Ejaaz: and that gap started to separate.
21:33
Ejaaz: Intelligence started to increase as a ceiling, but humans weren't necessarily as close to it.
21:40
Ejaaz: And recently, with all these breakthroughs in frontier math and discovering
21:44
Ejaaz: things that humans couldn't actually figure out themselves for millennia.
21:48
Ejaaz: Intelligence is starting to break through.
21:50
Ejaaz: So I don't know if this analogy is still working, if you can see my hands,
21:53
Ejaaz: but it's starting to accelerate really rapidly at a fraction of the cost that
21:57
Ejaaz: it took humans to do. They can work 24-7 with all these AI agents figuring it out for us.
22:02
Ejaaz: But the humans are falling drastically behind because we don't actually understand
22:06
Ejaaz: the solutions and all these breakthroughs that the AI models themselves are making.
22:10
Ejaaz: So we are at the same time creating the thing, but breaking away from it.
22:14
Ejaaz: And those 5,000 people that,
22:16
Ejaaz: Andre Kapathy talks about on this tweet here, are maybe the closest,
22:20
Ejaaz: but even they are falling behind. That's the whole point of this episode.
22:24
Ejaaz: We are creating these very powerful, artificial, intelligent models,
22:29
Ejaaz: but we are at the same time losing the grasp of understanding what these things
22:33
Ejaaz: are, both from an alignment perspective and just from a solutions perspective.
22:37
Ejaaz: It's all a very weird world. I wanted to create this episode because honestly,
22:41
Ejaaz: I'm trying to figure this all out myself and you guys that are listening to
22:45
Ejaaz: this that have listened to our episodes forever
22:47
Ejaaz: have always given me really good perspectives so let me know in the comments
22:51
Ejaaz: dm me my socials are all linked before but it's just a very interesting and
22:54
Ejaaz: weird time i have no idea where all of this ends up if this sounds too doomer
23:00
Ejaaz: i promise you that's not actually how i think but i need to call the trends
23:03
Ejaaz: as i see them happening because these stories all exist in their silos and it's
23:07
Ejaaz: important to kind of bring them together and give them all layered and nuanced
23:11
Ejaaz: take as to how all of this is playing out.
23:13
Ejaaz: But anyway, that is the episode. Thank you so much for listening.
23:17
Ejaaz: Yeah, I am still going to be in Japan for the next seven days.
23:21
Ejaaz: I want to keep doing this. We want to keep doing this because I don't know,
23:25
Ejaaz: there's just so much fun stuff happening.
23:27
Ejaaz: There's so many important advancements happening and I want to keep informing
23:31
Ejaaz: the audience and sharing my takes. So I am still working on my whole demeanor
23:36
Ejaaz: around this whole thing.
23:38
Ejaaz: I am, you know, sitting in front of a camera. I don't have a co-host to talk
23:41
Ejaaz: to right now, but I'm enjoying it a lot.
23:44
Ejaaz: I am kind of growing into it a bit. I'm trying to figure this all out.
23:47
Ejaaz: So your feedback, which many of you have provided is very helpful. Keep doing it.
23:51
Ejaaz: And yeah, I will see you on the next one. Thank you for listening to today's
23:54
Ejaaz: episode brought to you by Qualcomm.