AI Whistleblower: We Are Being Gaslit By AI Companies, They’re Hiding The Truth! - Karen Hao

0:00 So much of what's happening today in the AI industry is extremely inhumane.

0:04 But this is me playing devil's advocate.

0:06 And logically, it could be the case that the civilization that accelerate

0:09 their research with AI is going to be the superior civilization.

0:13 No, it's not.

0:13 This is a prediction that you're making, right?

0:16 Making Zuckerberg's making.

0:18 And do you know what the common feature of all of them is?

0:20 They profit enormously off of this myth.

0:22 You know, I have all these internal

0:23 documents showing that they're purposely trying to create

0:26 that feeling within the public so that they

0:28 can extract and exploit and extract and exploit.

0:31 So, what do we do about it?

0:32 We need to break up the empires of AI.

0:35 You know, I've been covering the tech industry for over 8 years,

0:38 interviewed over 250 people,

0:40 including former or current OpenAI employees and executives.

0:42 And I can tell you that there are many parallels

0:45 between the empires of AI and the empires of old, right?

0:47 like Lelay claimed the intellectual property of artists, writers,

0:49 and creators in the pursuit of training these models.

0:52 Second, they exploit an extraordinary amount of labor,

0:54 which breaks the career ladder because someone gets laid off

0:57 and then they work to train the models on the very

1:00 job that they were just laid off in, which will

1:02 then perpetuate more layoffs if that model then develops that skill.

1:05 And when they talk about that there's going to be

1:07 some new jobs created that we can't even imagine,

1:09 a lot of the jobs that are created are way worse than the jobs that were there.

1:14 And then there's the environmental and public health crisis that these companies

1:17 have created and how they're able to also spend hundreds of millions

1:21 to try and kill every possible piece of legislation that gets

1:24 in their way and will censor researchers

1:27 that are inconvenient to the empire's agenda.

1:29 But what I'm saying is not that these technologies don't have utility.

1:33 It's that the production of these technologies right

1:35 now is exacting a lot of harm on people.

1:38 But we have research that shows that the very same capabilities could be

1:43 developed in a different way that doesn't

1:45 have all of these unintended consequences.

1:48 So let's talk about all of that.

1:53 This is super interesting to me.

1:54 My team given me this report to show me

1:55 how many of you that watch this show subscribe.

1:57 And some of you have told us according

1:59 to this that you are unsubscribed from the channel randomly.

2:02 So favor to ask all of you.

2:03 Please could you check right now if you've hit the subscribe button if you

2:06 are a regular viewer of the show and you like what we do here.

2:08 We're approaching quite a significant landmark

2:10 on this show in terms of a subscriber number.

2:12 So if there was one simple free thing that you could do to help us,

2:16 my team, everyone here to keep this show free,

2:18 to keep it improving year over year and week over week,

2:21 it is just to hit that subscribe button and to double check if you've hit it.

2:24 Only thing I'll ever ask of you, do we have a deal?

2:27 If you do it, I'll tell you what I'll do.

2:28 I'll make sure every single week, every single month,

2:31 we fight harder and harder and harder and harder to bring

2:33 you the guests and conversations that you want to hear.

2:35 I've stayed true to that promise since the very beginning

2:37 of the D of Sio and I will not let you down.

2:40 Please help us.

2:41 Really appreciate it.

2:42 Let's get on with the show.

2:47 Karen, how you've written this book in front of me here called Empire of AI:

2:52 Dreams and Nightmares in Sam Alman's Open AI.

2:55 I guess my first question is what is the research and the journey you went

3:00 on in order to write this book we're going

3:02 to talk about and the subjects within it today

3:04 I took a strange route into journalism I studied

3:07 mechanical engineering at MIT and so when I graduated

3:10 I moved to San Francisco I joined a tech

3:12 startup I became part of Silicon Valley and I

3:16 basically received an education in what Silicon Valley is

3:19 about because a few months into joining a very

3:21 missiondriven startup that was focused on building technologies

3:23 that would help facilitate the fight against climate change.

3:27 The board fired the CEO because the company was not profitable.

3:31 And this was in hindsight a very pivotal moment for me

3:35 because I thought if this hub is ultimately geared towards building

3:40 profitable technologies and many of the problems in the world

3:44 that I think need solved are not profitable problems like climate change.

3:48 Then what are we actually doing here?

3:51 like what how did we get to a point where innovation is not actually necessarily

3:56 working in the public benefit and sometimes even

3:58 undermining the public benefit in pursuit of profit.

4:01 In that moment, I had a bit of a crisis where I thought, well,

4:06 I just spent 4 years trying to set myself up

4:10 for this career that I now don't think I am cut out for.

4:15 And I thought, well, I might as well just try something totally different.

4:19 I've always liked writing and that's how after 2 years

4:23 I landed at a role at MIT technology review covering

4:27 AI full-time and that gave me a space to then

4:31 explore all of these questions of who gets to decide what

4:34 technologies we build how does money and ideology also drive

4:38 the production of those technologies and how do we ultimately

4:41 make sure that we actually reimagine the innovation ecosystem to work

4:46 for a broad base of people all around the world.

4:50 And so that is kind of how I then

4:52 set off on this journey of ultimately writing a book.

4:56 I didn't realize that I was working towards writing a book,

4:59 but starting in 2018 when I took that job was essentially the moment

5:04 in which I began researching the story that I I document in it.

5:09 A very timely time to start working in artificial intelligence.

5:12 For anyone that doesn't know,

5:12 this is pre OpenAI chat GPT launch moment that shook the world.

5:19 But in writing this book,

5:20 you interviewed a lot of people and went to a lot of places.

5:22 Can you give me a flavor of how many people you've interviewed,

5:25 where it's taken you around the world, etc.

5:27 I interviewed over 250 people.

5:29 So over 300 interviews,

5:30 over 90 of those people were former or current OpenAI employees and executives.

5:36 So the book covers the inside story of opening eyes's first

5:40 decade and how it ultimately got to where it is today.

5:44 But I didn't want to write a corporate book.

5:46 I felt very strongly that in order to help

5:49 people understand the impact of the AI industry,

5:52 we would also have to travel well beyond Silicon Valley.

5:55 These companies tell us that AI is

5:58 going to benefit everyone and that's their mission.

6:00 But you really start to see that rhetoric break down when

6:04 you go to the places that look nothing like Silicon Valley,

6:08 that speak nothing like Silicon Valley,

6:10 and that have a history and culture that are fundamentally different as well.

6:14 And that's where you start to really understand the true

6:17 reality of how this industry is unfolding around us.

6:23 Karen, I often try and steer conversations, but in this situation,

6:26 I feel like it's probably my responsibility to follow.

6:29 So with that in mind,

6:30 I'm going to ask you where does this journey begin and where should

6:33 we be starting if we're talking about the subjects of empire of AI,

6:37 AI generally artificial intelligence and also I'd say one

6:40 thing I'm really keen to do in this conversation which

6:42 is I often see in conversations is left out

6:45 is let's assume that our viewers know nothing about AI.

6:48 Yeah.

6:48 So they don't know what scaling laws are or GPUs or comput

6:51 or whatever and let's try and keep this as simple as we possibly

6:55 can in terms of language or explain all the complicated language so

6:59 that we can bring as much people with us as we possibly can.

7:02 Yes.

7:02 Where should we start?

7:03 I think we should start with when AI started as a field.

7:08 So this was back in 1956 and there were a group

7:12 of scientists that gathered at Dartmouth University to start a new discipline,

7:16 a scientific discipline to try and chase an ambition.

7:20 And specifically an assistant professor at Dartmouth University,

7:23 John McCarthy decided to name this discipline artificial intelligence.

7:28 This was not the first name that he tried.

7:30 The previous year he tried to name it Automata Studies.

7:34 And the reason why some of his colleagues were concerned about this name

7:37 was because it pegged the idea

7:39 of this discipline to recreating human intelligence.

7:44 And back then, as is true today,

7:46 we have no scientific consensus around what human intelligence is.

7:51 There's no definition from psychology, biology, neurology.

7:55 And in fact, every attempt in history to quantify

8:00 and rank human intelligence has been driven by nefarious motives.

8:05 It's been driven by a desire to prove scientifically that certain

8:10 groups of people are inferior to other groups of people.

8:14 There are no goalposts for this field and there

8:18 are no goalposts for the industry when they

8:20 say that they are ultimately trying to recreate

8:22 AI systems that would be as smart as humans.

8:26 How do we even define what that means?

8:28 And when are we going to get there

8:30 if we don't know how to define the destination?

8:34 And what that effectively means is that these companies

8:38 can just use the term artificial general

8:40 intelligence which is now the term to refer

8:42 to this ambitious um goal to recreate human intelligence.

8:47 They can use it however they want to and they can

8:50 define and redefine it based on what is convenient for them.

8:52 So in OpenAI's history, it has defined and redefined it many times.

8:56 When Sam Alman is talking with Congress,

8:59 AGI is a system that's going to cure cancer, solve climate change, cure poverty.

9:04 When he's talking with consumers that he's trying to sell his products

9:08 to, it's the most amazing digital assistant that you're ever going to have.

9:13 When he was talking with Microsoft, you know,

9:15 in the deal that OpenAI and Microsoft

9:17 struck where Microsoft invested in the company,

9:21 it was defined as a system that will generate hundred billion of revenue.

9:25 And on OpenAI's own website, they define it as highly autonomous systems

9:30 that outperform humans in most economically valuable work.

9:35 This is like not a coherent vision of one technology.

9:39 These are very different definitions that are spoken out loud to the audience

9:43 that needs to be mobilized to ward off regulation or get

9:48 more consumer buy in into the the industry's quest or to get

9:54 more capital more resources for continuing

9:56 on this journey with ambiguous definitions.

10:00 I mean, speaking about different definitions through time, in 2015,

10:04 in a blog post that Sam Waltman wrote before open air was officially announced,

10:08 he explicitly outlined the existential risk by saying,

10:12 "Development of superhuman machine intelligence is probably

10:15 the greatest threat to the continued existence of humanity.

10:18 There are other threats that I think are more certain to happen, for example,

10:22 an engineered virus,

10:23 but AI is probably the most likely way to destroy everything

10:28 in general." When Alman is writing for the public or speaking for the public,

10:33 he does not just have the public as the audience in mind,

10:37 there are other people that he is trying

10:39 to motivate or mobilize when he says these things.

10:43 And in that particular moment,

10:45 Alman was trying to convince Elon Musk to join him on co-founding OpenAI.

10:51 And Musk in particular was spending all of his time sounding the alarm

10:56 on what he saw as a huge existential threat that AI could pose.

11:01 And so in that blog post, if you look at the the language that Alman uses side

11:06 by side with the language that Musk was using at the time,

11:09 it mirrors all the things that Musk was saying identical.

11:12 I mean, 10 years ago, Musk was going on podcast saying,

11:15 tweeting, whatever, that the greatest existential risk to humanity was AI.

11:20 Yeah.

11:20 And so you know like his parenthetical there are other things

11:23 that we that might actually be more likely to happen like engineered viruses.

11:27 It's because up until then Alman had been talking just about engineered viruses.

11:34 And so now that he needs to pivot to speak to an audience of one to Musk.

11:39 He needs to kind of resolve the contradiction

11:41 between what he's now elevating as his new central

11:45 fear to be the same as Musk's new

11:47 central fear with what he had previously been saying.

11:50 So that's why he's like I think this is now even though before I said this

11:56 and are you saying that Sam Alman

11:57 manipulated Musk because Elon did end up donating

12:01 a huge amount of money to um open

12:04 AAI and co-founding it I believe with Sam Alman.

12:06 Elon Musk did end up co-ounding it with Altman.

12:08 And certainly from Musk's perspective, he does feel manipulated because he

12:14 feels like Alman was engineering his language

12:19 in a way that would make Musk trust him as a a partner in this endeavor.

12:26 And of course then Musk is leaves.

12:29 Um and through some of the documents that came out during

12:32 the the lawsuit that Musk and Altman are engaged in now,

12:36 it has become clear that there was a degree

12:39 to which Musk was actually muscled out a little bit.

12:43 And so that's why he's left

12:45 with this very intense personal vendetta against Altman,

12:49 saying that somehow Alman tricked him into being part of this.

12:53 So in in 2015, Sam Alman is writing these blog posts saying this is,

12:57 you know, one of the greatest existential threats.

12:59 At the same time, in 2015,

13:01 Musk is doing some very famous speeches at the time at MIT.

13:05 He said that AI was the biggest existential

13:07 threat and compared developing AI to summoning the demon.

13:11 And what you're saying here is you're saying that Samman was just mirroring

13:14 the language that Elon was using to get Elon involved in open open AAI.

13:18 And later it appears and again there's a legal case taking

13:21 place now that Sam might have muscled Elon out in some capacity.

13:26 Yeah.

13:26 So we know from the lawsuit and the documents that have come out

13:28 in the lawsuit that Ilia Sgver who is the chief scientist of OpenAI at the time

13:34 and Greg Brockman chief technology officer

13:36 at the time when they were deciding whether

13:40 or not to maintain OpenAI as a nonprofit

13:43 because it was originally founded as a nonprofit.

13:45 They decided okay we need to create a for-profit entity

13:47 but the question was who should be the CEO of this for-profit entity.

13:50 Should it be Musk or should it be Alman?

13:52 because it's they were the two co-chairmen of the nonprofit.

13:56 And in the emails, it became clear that Ilia

13:59 and Greg first chose Musk to be the CEO.

14:05 But through my reporting,

14:07 I discovered that Altman then appealed personally to Greg Brockman,

14:12 who was a friend of his that they had known,

14:14 they had known each other for many years through the Silicon Valley scene,

14:17 and said, "Don't you think that it would be a little

14:22 bit dangerous to have Musk be the CEO of this company,

14:26 this new for-profit entity, because, you know, he's a famous guy.

14:30 He has a lot of pressures in the world.

14:32 He could be threatened.

14:34 He could act erratically.

14:36 He could be unpredictable.

14:37 And do we really want a technology that could be super powerful

14:41 in the future to end up in the hands of this man?

14:45 And that convinced Greg and Greg then convinced Ilia,

14:49 you know, I think there's a point here.

14:51 Do we really want to give this much power to Musk?

14:55 And that is why Musk then leaves

14:58 because then they the two switch their allegiances.

15:01 They say, "Actually, we want Altman to be the CEO." And then Musk is like,

15:04 "If I'm not CEO, I'm out."

15:06 So, it sounds like Sam again managed to persuade someone to do something.

15:12 Mhm.

15:12 I guess this begs the question, what do you think of Sam Orman?

15:17 I think he's a very controversial figure.

15:19 You did an interesting pause.

15:21 It's a pause where someone tries to select their words.

15:26 Well, this is this is this is what's so interesting

15:30 about those interviews is people are extremely polarized on Alman there.

15:35 No one has in between feelings about him.

15:39 Either they think he's the greatest tech leader

15:42 of this generation akin to the Steve Jobs

15:44 of the modern era or they think that he's

15:47 really manipulative and an abuser and a liar.

15:52 And what I realized because I interviewed so many people is it really comes

15:57 down to what that person's vision of the future is and what their goals are.

16:02 So if you align with Altman's vision of the future,

16:07 you're going to think he's the greatest asset ever

16:09 to have on your side because this man is really persuasive.

16:12 He's incredible at telling stories.

16:14 He's incredible at mobilizing capital, at recruiting talent,

16:17 at getting all the inputs that you need to then make that future happen.

16:22 But if you don't agree with his vision of the future,

16:25 then you begin to feel like you're being manipulated by him

16:30 to support his vision even if you fundamentally don't agree with it.

16:34 And this is the story especially of Daria Amade,

16:39 CEO of Enthropic, who was originally an executive at OpenAI.

16:43 So for people that don't know,

16:44 Dario now runs anthropic which is the maker of Claude.

16:47 A lot of people probably are more familiar with Claude.

16:49 Yeah.

16:50 And it's one of the biggest competitors to OpenAI.

16:53 And Amade at the time when he was an ex executive at OpenAI,

16:59 he thought that Alman was on the same page

17:03 with him and then over time began to feel

17:06 that Altman was actually on exactly the opposite page

17:10 of him and felt that Altman had used Amade's intelligence,

17:16 capabilities, skills to build things and bring about a vision

17:22 of the future that he actually fundamentally didn't agree with.

17:25 And so that's why people end up with this bad taste in their mouths.

17:29 And so, you know, I've been covering the tech

17:31 industry for over eight years and covered many companies.

17:35 I've covered Meta, Google, Microsoft in addition to Open AI.

17:38 and OpenAI and Altman is it's the only figure that I've seen this degree

17:45 of polarization with where people cannot decide

17:50 whether he's the greatest or the worst.

17:53 You mentioned Dario there and I found it really what I found really

17:57 interesting is to look at how people's

17:58 quotes evolve over time with their incentives.

18:00 So I was looking at all of the all of the things

18:03 they've said on the record on podcasts in their blog

18:05 post to see how it's evolved over time and Dario who

18:08 was the former VP of research open AAI and has now moved

18:11 on to enthropic who are taking a slightly different approach to developing

18:14 AI said back in 2017 while he was still at open AI

18:19 that this is a quote I think at the extreme end is

18:22 the Nick Bostonramm style of fear that an AGI could destroy humanity.

18:26 I can't see any reason in principle why that couldn't happen.

18:30 My chance that something goes really quite catastrophically wrong on the scale

18:34 of human civilization might be somewhere between 10% and 25%.

18:40 And also you mentioned Ilia who was a co-founder of OpenAI and then left.

18:45 I guess the first question I'd ask is why did I leave?

18:49 It's a great question.

18:52 So he was instrumental in trying to get Sam

18:55 Alman fired and he's another one of the people who

18:59 over time began to feel like he was being manipulated

19:02 by Alman towards contributing something that he didn't believe in.

19:06 And for you know because I interviewed a lot of people Ilia

19:10 in particular had two pillars that he cared about deeply.

19:16 One is making sure we get to so-called AGI

19:20 and the other is making sure that we get to it safely.

19:24 And he felt that Altman was actively undermining both things.

19:29 He felt that Alman was creating a very

19:31 chaotic environment within the company where he

19:34 was pitting teams against each other where

19:36 he was telling different things to different people.

19:39 Have you ever spoken to him?

19:40 I have.

19:41 So, so I interviewed him in 2019 for a profile

19:44 that I did of OpenAI um for MIT Technology Review

19:49 and back in 2019, he has a quote where he says,

19:51 "The future's going to be good for AIs regardless.

19:53 It would be nice if it was also good for humans as well.

19:56 It's not that it's going to actively hate humans or want to harm them,

19:59 but it's just going to be so powerful.

20:01 And I think a good analogy would be the way that humans treat animals.

20:04 It's not that we hate animals.

20:05 I think humans love animals, and I have a lot of affection for them.

20:08 But when the time comes to build a highway between two cities,

20:11 we are not asking the animals for permission.

20:14 We just do it because it's important to us.

20:16 And I think by default,

20:17 that's the kind of relationship that's going to be between us and AI,

20:22 which are truly autonomous and operating on their own behalf.

20:26 And that was in 2019, the year that you interviewed him.

20:29 One of the things that I I feel like we should take a step back to examine is

20:32 going back to this idea of what even is

20:34 artificial intelligence and what do we mean by intelligence?

20:37 And a huge part of the views of the different people and the quotes that you're

20:42 reading derives from a specific belief that they

20:45 each have in this question of what is intelligence,

20:49 what constitutes intelligence.

20:52 For Ilia, he has throughout his research career

20:56 felt that ultimately our brains are giant statistical models.

21:03 This is not something that you know we actually know but this is his own

21:07 hypothesis also the hypothesis of his mentor

21:10 Jeffrey Hinton who also was on this podcast.

21:13 This is why they have such a strong conviction

21:15 in the idea of building AI systems that are

21:18 statistical models and that this particular approach is going

21:22 to lead to intelligent systems as we are intelligent.

21:26 It's a hypothesis that they have.

21:28 It's not one that has been proven by science.

21:32 And some people vehemently disagree with them on this particular thing.

21:36 But if you step into their shoes and take

21:39 on that hypothesis and assume that it's true,

21:42 that our brains are in fact statistical engines

21:46 and that these systems that they're building are also statistical engines,

21:50 that they're making bigger and bigger and bigger

21:52 until they become the size of the human brain.

21:55 That's why they say that making

21:58 this comparison where the system will become equal

22:02 to human intelligence and then maybe exceed

22:04 human intelligence is relevant in their framework.

22:08 And um Ilia gave a talk at one point at this really

22:12 prominent AI research conference that happens

22:14 every year called neural information processing systems.

22:18 It's a mouthful, but he gave this keynote where he shows

22:22 this chart of the size of brains and the intelligence of a species.

22:28 And it's roughly linear.

22:30 The bigger the size of the brain, the more intelligent the species.

22:34 And so for him, he thinks he's building a digital

22:38 brain because he he thinks brains are just statistical engines.

22:42 So from that logic it's like okay if we then build

22:46 a bigger statistical engine than the human brain then based on this chart

22:52 it will be more intelligent and then we will be subjected

22:55 to the same treatment that we've subjected

22:58 animals but it's really important to understand

23:00 that these are scientific hypotheses

23:02 of specific individuals within the AI research

23:05 community and there's a lot a lot of debate about whether this is

23:09 in fact the case and some of The biggest critics say it's

23:14 very reductive to think of our brains as simply just statistical engines.

23:18 Why why does it matter to know the mechanism?

23:23 Is it not just important to know the outcome

23:25 which is that it's going to be able to do

23:28 make a video for me or agents are going to be able to do the work that I do.

23:32 Does it does it really really matter for us to know the mechanism behind it?

23:36 Yes and no.

23:37 So it matters because these companies they are

23:42 driving their future actions based on this hypothesis.

23:47 So they have decided we think that this hypothesis is true like we should

23:53 just continue building larger and larger statistical

23:56 models in the pursuit of artificial general intelligence.

23:59 And that's then having global consequences like in order

24:03 to continue doing that they're hoovering up more and more data.

24:07 They're building more and more data centers.

24:09 They are having uh they're, you know,

24:11 exploiting more and more labor in order to continue on this path.

24:15 Here's a question that I think is important to ask is why

24:19 are we trying to build AI systems that are duplicative of humans?

24:23 We're kind of having this conversation right now where we've

24:25 just taken the premise of this industry as a good thing.

24:30 Like they said that we should be building AGI,

24:33 so we say that we should be building AGI.

24:35 I would like to ask like why are we doing that?

24:38 Why is it that we are building a technology

24:41 that is ultimately designed to replace and automate people away?

24:46 That is not the enterprise of technology.

24:49 Like we should be building technology and the purpose

24:53 of technology throughout history has been to improve human flourishing,

24:58 not to replace people.

25:00 And so this is like a a critical part

25:04 of my critique of these companies and and these scientists that have

25:07 just adopted this goal and have relentlessly pursued it

25:11 and have had enormous capital and enormous resources to pursue it.

25:14 Is is this the right goal?

25:16 What like why are we doing this?

25:18 Why can't we just build AI systems that do things

25:22 like accelerate drug discovery and improve people's health care outcomes,

25:27 which are systems that have nothing to do with the statistical

25:30 engines that they're trying to build to duplicate the human brain?

25:34 So why are they doing it?

25:35 I mean, you've interviewed all these people.

25:36 I think it's what, 300 people in total,

25:39 80 or 90 of them from OpenAI, the maker of CHACHBC.

25:42 Why do you think they're doing it?

25:44 I think it's because they're driven by an imperial agenda.

25:47 And that is why I call these companies empires of AI.

25:50 What do you mean by an imperial agenda?

25:52 What does that term mean?

25:53 Empire is the only metaphor that I've ever found

25:57 to fully encapsulate all of the dimensions of what

26:01 these companies do and the scale that they operate

26:04 and what motivates them to do what they do.

26:07 And there are many parallels that you see between what

26:11 I call the empires of AI and the empires of old.

26:13 They lay claim to resources that are not

26:16 their own in the pursuit of training these models.

26:18 That's the data of individuals,

26:20 the intellectual property of artists, writers, and creators.

26:23 Their land grabbing in order to build

26:25 these supercomputer facilities for training the next generation models.

26:28 Second, they exploit an extraordinary amount of labor.

26:31 They contract hundreds of thousands of workers all around

26:34 the world including in the US to ultimately make these technologies.

26:40 We can talk about that more.

26:42 And they also design their tools to be

26:45 labor automating so that when the technologies are deployed,

26:48 it also affects labor rights because it erodess away labor rights.

26:53 And this is a political choice that they have.

26:56 Third, they monopolize knowledge production.

26:59 And so they project this idea that they're

27:00 the only ones that really understand how the technology works.

27:03 And so if the public doesn't like it,

27:04 it's because they don't actually know enough about this technology.

27:08 They do this to the public.

27:09 They do this to policy makers.

27:11 And they've also captured the majority of the scientists

27:14 that are working on understanding the limitations and capabilities of AI.

27:18 You think they're gaslighting the public in a way?

27:21 They are.

27:21 Yeah.

27:21 So if most of the climate scientists

27:24 in the world were bankrolled by fossil fuel companies,

27:28 do you think we would get an accurate picture of the climate crisis?

27:33 No.

27:32 And in the same way they employ and bankroll the AI

27:36 industry employs and bankrolls most of the AI researchers in the world.

27:40 So they set the agenda on AI research in soft ways simply by funneling

27:45 money to their priorities so that only

27:48 certain types of AI research are produced.

27:50 But they also will censor researchers when they

27:54 do not like what the researcher has found.

27:57 And so I talk about the case of Dr.

27:59 Timmy Gabru in my book who was the ethical AI team co-lead at Google when

28:05 she was literally hired to critique the types

28:09 of AI systems that Google was building.

28:11 She then co-wrote a critical research paper that was showing how

28:16 large language models specifically were leading

28:18 to certain types of harmful outcomes.

28:22 And in an attempt to try and stop this research from being published,

28:26 Google ended up firing Gabru and then fired her other co-lead Margaret Mitchell.

28:33 And so they control and quash the research

28:39 that is inconvenient to the empire's agenda.

28:42 Did you have an example where this is happening to journalists

28:45 as well that are asking questions of their team members?

28:49 I think I was watching a video of yours where there was

28:52 a young man that was saying he had someone show up at his door,

28:54 knocked on his door and asked for information, emails, text messages,

28:58 and this person was from one of the big AI companies.

29:01 This was opening.

29:02 I started subpoenaing some of its critics.

29:04 Yeah.

29:04 Um as a as part of a what's what appears to be a campaign of intimidation,

29:11 but also what appeared to be a campaign of fishing for more

29:14 information to figure out to map out the network of critics further.

29:20 But this was a man who runs a small watchdog

29:24 nonprofit and they had been doing a lot of work

29:27 during that time to try and ask questions about

29:31 OpenAI's attempt to convert from a nonprofit to a for-profit.

29:35 Ultimately, OpenAI was successful in that conversion.

29:37 But during the period where it was sort

29:40 of existential for open AI to complete this conversion,

29:44 there were a lot of civil society groups and watchdog groups like MIDAS

29:48 who were trying to prevent the process from happening in the dead of night.

29:54 They were trying to get more transparency.

29:57 They were trying to have more

29:58 public debate about this because it's unprecedented.

30:01 And it was then that um there was a knock on his door and he was served papers.

30:08 What did the papers say?

30:10 The papers asked him to reproduce every single piece

30:13 of communication that he had had that might have involved Musk.

30:17 So this was like this strange paranoia that OpenAI had

30:19 that Musk was somehow funding these people to block the conversion.

30:24 None of them were actually funded by Musk.

30:26 So in this particular case their request he simply was just answered

30:30 you know I I don't have any documents because this doesn't exist.

30:33 So going back to this point of empires you were saying that one

30:36 of the factors of an empire is a land grab and then the next one was

30:40 was labor exploitation labor exploitation.

30:43 The third one, controlling knowledge production.

30:47 And one of the other ones that's really important to understand about the AI

30:52 empires in particular is empires always have this narrative that they they say

30:59 to the public like we're the good empire and we need to be

31:03 an empire in the first place because there are also bad empires in the world.

31:07 And if you allow us to take all the resources and use all of the labor,

31:14 then we promise we will bring you progress and modernity for everyone.

31:19 We will bring you to this utopic state akin to an AI heaven.

31:23 But if the evil empire does it first, we will descend into a hell.

31:28 And the evil empire being in this case,

31:30 in this case, most often it's China.

31:33 But actually in the early days, Open AI evoked Google as the evil empire.

31:38 So all of their decisions were about we

31:40 need to do it first because otherwise Google,

31:43 this evil corporation that's driven by profit, us as a benevolent nonprofit.

31:48 Like this is a this is a critical contest of who wins.

31:54 Do you think the people building these AI companies believe

31:58 that the outcome is going to be all good now?

32:02 Do you think they think that it's going to be it's going to serve everyone?

32:05 It's going to be the age of abundance.

32:06 Everything's going to go up well.

32:07 What do you think they believe?

32:08 What do you think Sam believes?

32:10 So, so this is so funny is such a core part of the mythology

32:15 that they create around the AI industry includes

32:19 the belief that it could go very badly.

32:22 It goes hand in hand.

32:24 like they need that part of the myth in order

32:27 to then say and that's why we need to be in control

32:30 of the technology because that's the only way that it's

32:32 going to go really really well and Alman has said publicly

32:35 you know the worst case lights out for everyone but best

32:40 case we cure cancer we solve climate change and there's abundance

32:44 and Dario Amade same kind of rhetoric was like worst case

32:49 catastrophic or existential harm for humanity best case mass human flourishing.

32:56 So this is like two sides of the same coin.

32:58 Like they have to use both of these narratives

33:02 in order to continue justifying an extremely

33:06 anti-democratic approach to AI development where there should

33:09 not be broad participation in developing this technology.

33:12 They must be the ones controlling it at every step of the way.

33:16 Sam Orman did a tweet saying,

33:18 "There are some books coming out about open AI and me.

33:21 We only participated in two of them.

33:23 one by Kesh Hegy

33:27 Keegy Khaggy focused on me and one by Ashley Vance on OpenAI.

33:31 Um he went on to say no book will get everything right

33:34 especially when some people are so intent on twisting things but these two

33:38 authors are trying to you quote retweeted that tweet from Sam

33:44 Alman and you said the unnamed book empire of AI is mine.

33:51 Do you believe that tweet from Sam Alman was in reference to your book?

33:55 100%.

33:55 Because there's only three books coming out about him

33:58 and he had caught wind that your book was coming out and

34:00 he knew my book was coming out because I had contacted OpenAI

34:04 from the very beginning of my process and said I'm working on a book now.

34:07 Will you participate in it?

34:08 And actually initially they said yes even though so my history

34:12 with OpenAI I profiled the company for MIT technology review.

34:16 I embedded within the office for 3 days in 2019.

34:19 my profile comes out in 2020, the leadership are very unhappy.

34:25 And in my book, I actually quote an email that I received

34:28 that Sam Alman sent to the company about my profile saying, "Yeah,

34:33 this is not great." And from then on, the company's stance to me was,

34:43 "We are not going to participate in anything that you do.

34:47 we are not going to respond to anything any of the questions that you receive.

34:50 And this was, you know, this was things that they explicitly articulated.

34:55 It wasn't like me inferring.

34:57 Um, so I I had a a colleague at MIT Technology Review that also covered AI.

35:02 And at one point opening, I sent him this press release being like,

35:05 "We would love for you to cover this story." And he was like, "I'm really busy.

35:08 Will you send it to Karen?" And they were like, "Oh, no.

35:12 We have a history.

35:13 You understand?" And so, so for three years they they refused to talk to me,

35:20 but then I ended up at the Wall Street Journal where if they felt

35:24 a a bit compelled because it was

35:26 the journal to reopen the lines of communication.

35:30 And so I I I started having, you know, more dialogue with them.

35:35 Every time I wrote a piece,

35:36 I would always send them here's my request for comment.

35:38 I would always ask them like, will you sit for interviews?

35:41 And we did get to a more productive relationship.

35:45 And then I embarked on the book.

35:46 So I I left the journal to focus on the book full-time.

35:49 And I told them right away, I'm working on this book.

35:53 I want to continue this productive conversation where I

35:57 make sure I reflect OpenAI's perspective in the book.

36:02 And so they were like, we can arrange interviews for you.

36:05 You can come back to the office.

36:07 We'll set up some conversations.

36:10 And then as we were going back and forth on this, the board fired Sam Alman.

36:17 And that's when things started going kind of south

36:19 because the company started becoming very sensitive to scrutiny.

36:24 And so then they started pushing kicking the can down the road,

36:27 down the road, down the road.

36:28 And I kept saying, "Hey, when are we rescheduling this?

36:30 What's going on?" And then I get an email saying,

36:33 "We are not going to participate at all.

36:35 You are not coming to the office.

36:36 You're not doing interviews." and I had actually already booked my tickets.

36:40 So, I was already going to fly to San Francisco to have the the interviews.

36:46 And so, then I told them I was like, "That's fine.

36:50 I will still engage in the process

36:53 where I'll give you extensive requests for comment.

36:55 I'll ask through my reporting,

36:56 I'll keep you updated on all the things that I'm finding so that you

37:00 can choose to still comment." I gave them 40 pages of requests for comment.

37:06 and I gave them over a month to respond to all of that.

37:09 So, this was when the tweet came out was we were doing

37:12 all this back and forth trying to and that's when Alman tweeted this.

37:21 H and they never responded to a single one of the one of the 40 pages.

37:25 Sam Alman does a lot of interviews.

37:28 Yeah.

37:28 You know, he's doing a lot of interviews all the time.

37:29 He's done every podcast.

37:31 I've seen him on everything from Tucker Carlson to I think he's done Theo,

37:35 Joe Rogan, um podcasts all over the world.

37:39 I wonder why he won't do mine.

37:45 Well, maybe.

37:46 I don't know why.

37:47 I I I don't know.

37:47 I think I'm fair with everyone.

37:48 I just ask I just ask questions I genuinely care about.

37:51 I don't come in with huge preconceptions

37:53 or at least meet people for the first time.

37:55 But I've heard through the grape vine um that he doesn't want to do mine.

38:00 I mean, going back to what you were saying earlier

38:02 that with this the way that OpenAI and these companies control research,

38:08 you asked, do they also do this with journalists?

38:12 I mean, yes, the answer is yes.

38:14 And apparently they they also do it with anyone who has,

38:17 you know, a broad mass communications platform.

38:21 It's not just about the conversation that you're going to have with them.

38:24 It's about who you also choose to platform.

38:28 And there's this huge problem in technology journalism where companies know

38:34 that a really big carrot that they can give to technology journalists is access.

38:39 Yeah.

38:39 Yeah.

38:39 Yeah.

38:40 And they will withhold that access at the drop of a hat if they

38:46 catch wind that you're speaking to someone

38:47 that they didn't want you to speak to.

38:49 This is so true.

38:50 And I don't think the average person really truly understands this.

38:54 Yeah.

38:54 So, this kind of sounds like theory as you say it,

38:56 but I'm not going to name names here because I don't think it's important,

39:00 but there is a particular person in AI who um whose team

39:06 have basically dangled the carrot of them coming here for like 18 months.

39:10 And I'm like, you don't you don't have to dangle the carrot.

39:12 I'm going to speak to whoever I want to regardless of the carrot or not.

39:15 And when this person comes, if they want to come,

39:17 I'll I'll give them a fair shot.

39:18 I'll ask them all genuinely curious

39:20 questions about what they're doing, their incentives.

39:22 I won't gotcha them.

39:24 I don't have a history of ever gotchering anybody.

39:26 Even if I dis like even if I have a different of opinion, I'll ask the question.

39:30 Yeah.

39:29 But they dangle carrots and they say, "Well,

39:31 if you know he he's thinking about it,

39:33 let's think about a date." And what what the strategy is,

39:35 and I don't think they they think those people don't understand,

39:38 is if we just dangle it for long enough,

39:40 then they will um perform in the way that we want them to do and they'll be

39:47 they'll be pleasant about us.

39:48 They won't be critical.

39:49 They won't give a give a critics.

39:52 Our critics.

39:53 And I think a lot of their game is just dangle the carrot forever.

39:57 Yes.

39:57 Yeah.

39:57 That's like the optimal outcome is if we just dangle it.

40:00 If we just tell them, yeah, look, we're just trying looking at the schedule.

40:03 It just doesn't work.

40:03 I think in the modern world,

40:04 you just have to go there and give your opinion

40:06 and allow the clash of ideas in the public forum,

40:08 let the viewers un decide for themselves.

40:11 Yeah.

40:11 What they think.

40:13 Yeah.

40:12 Um, but this is a Yeah.

40:14 This is such a huge part of their machinery is

40:17 the way that they use these tactics to massage the public image

40:22 of these companies and make sure that information that they don't want

40:24 out and even opinions that they don't want out there go out there.

40:30 Mhm.

40:29 And so this is this is you know I feel very

40:33 lucky now that opening I shut the door early on me

40:38 at the time I didn't feel lucky.

40:40 I felt like I had screwed myself over.

40:42 I was nicer access to a journalist, right?

40:48 Like you're supposed to report the truth and you're

40:51 always supposed to report in the interest of the public.

40:54 Like that is the point of journalism.

40:56 And in that moment it I I was like relatively junior in my career.

41:00 I was like, did I misunderstand what journalism about is is about?

41:05 Like

41:06 should I have actually been playing the access game?

41:10 Mhm.

41:10 But it was too late.

41:10 I had the door shut to me and so I had to build

41:14 my career understanding that the door the front door was never going to be open.

41:20 Yeah.

41:19 And that actually really strengthened my own ability

41:24 to just tell it like it is like objective.

41:27 Yeah.

41:27 And just report what I see are the facts being presented

41:31 to me irrespective of whether the company likes it or not.

41:34 And most often the company really does not like it but

41:38 I can continue to do the work.

41:40 They don't need to open the front door for me.

41:41 I was still able to do more than 300 interviews.

41:46 So Sam Alman gets kicked off the OpenAI executive team.

41:55 Did you find out why that happened?

41:58 Yeah, there's a scene by scene recounting from who?

42:03 I can't remember the exact number of sources,

42:05 so I don't want to misquote myself,

42:07 but it was around six or seven people that were directly

42:10 involved or had spoken to people

42:11 directly involved in the decision-making process.

42:15 So, Ilia Satskever is seeing these serious concerns about the way that Altman's

42:25 behavior is leading to bad research outcomes and poor decision-m at the company.

42:35 He then approaches a board member, Helen Toner.

42:39 Ilia, for anyone that doesn't know, is the the co-founder we mentioned earlier.

42:42 The co-founder of OpenAI we mentioned earlier.

42:45 Yes.

42:46 And he kind of does a bit of a sounding

42:50 board thing to Helen just because Ilia is freaking out.

42:54 He's like he's been like sitting on this these these concerns

42:57 for a while and he's like if I tell this to someone,

43:01 this could also be really bad for me if Alman finds out.

43:06 And so he asks for a meeting with Toner and in that first

43:13 meeting he's like re like he barely says a thing.

43:17 He's just like dancing around trying to figure out hey is

43:22 this someone that I can maybe trust to divulge more information.

43:25 And Toner's role and responsibilities at OpenAI were she was a board member.

43:29 Just a board member.

43:30 Yeah.

43:30 And and specifically an independent board member.

43:32 So opening eye when it was a nonprofit

43:35 the board was split between people who had

43:37 a stake financial stake in the company and then

43:40 people who were fully independent and this was meant

43:42 to be a structure that would balance the decision-m

43:46 to be in the benefit of the public

43:47 interest rather than to be in the benefit

43:49 of the for-profit entity that opening I then created

43:55 and Ilia as a non-independent board member was approaching toner

44:00 as an independent board member her to try and see whether or not

44:06 she was potentially seeing or hearing the same things that he

44:09 was about the effect that Alman was having on the company.

44:12 This then sets off a series of conversations first between Ilia

44:18 and Helen and then between Amir Moratti and some of the board members.

44:23 Samir Moratti was at that point the chief

44:26 technology officer of OpenAI where these two

44:29 senior leaders essentially through these conversations and through

44:31 documentation that they're pulling together like email,

44:34 Slack messages and so forth,

44:35 they convey to the independent board members, three independent board members,

44:40 we are very concerned about Altman's leadership

44:45 like he is creating too much instability

44:49 at the company and it is like he is the root of the problem.

44:54 It's not they they they were trying

44:57 to say to these independent board members like

44:59 the problem will not be fixed unless Alman

45:02 is removed because of the way that he's pitting

45:06 teams against each other and creating this environment

45:09 where people are unable to trust each other

45:11 anymore and they're competing rather than collaborating on what's

45:14 supposed to be this really really important technology.

45:17 When you say instability, that's a that's quite a vague term.

45:21 That could mean lots of things.

45:22 Like instability could mean pushing people hard to work harder, right?

45:26 What do you mean by instability in spec

45:28 as specific terms as you can possibly say them?

45:31 When chat GBT came out in the world, OpenAI was wholly unprepared.

45:37 They didn't think that they were launching a gangbusters product.

45:41 Yeah.

45:41 They thought they were releasing a research preview

45:44 that would help them get the data flywheel going,

45:47 collect a bunch of data from users that would then

45:50 inform what they thought would be the gang busters product,

45:54 which was a chatbot using GPT4 and chat GBT was using GPT 3.5.

46:01 And because of that, there were servers crashing all the time

46:08 because they they weren't they had to scale their their infrastructure,

46:11 you know, faster than any company in history.

46:13 And there were um there were all of these outages.

46:17 They were trying to also hire faster than any

46:19 company in history to try and have more personnel there.

46:22 And they were then sometimes hiring people that they were like,

46:24 "Actually, we made a mistake.

46:26 We shouldn't have hired you." So they were firing people left and right.

46:29 and people were just disappearing off of Slack and that's how

46:32 their colleagues would learn that they were no longer at the company.

46:35 And so it was yes like many fast

46:39 growing companies a very chaotic environment and a particularly

46:43 chaotic environment because it was extra fast like

46:48 they had to accelerate more than any other startup.

46:52 And on top of that mirror Morati and Ilasgiver felt that Alman was making

46:58 it worse like he was not actually

47:00 effectively ameliorating the circumstances of the chaos.

47:04 He was actually sewing more chaos, getting these teams to be more divided.

47:10 And this is where it's important to understand

47:13 that the executives and the independent board members,

47:18 they're all operating under this idea that they're building AGI

47:22 and that AGI could either be devastating or utopic to humanity.

47:29 And so it's not yes it's like any other

47:32 company and no it's not like any other company.

47:35 You cannot have like in their view you

47:38 cannot have this degree of chaos as the pressure

47:42 cooker for creating a technology that they

47:44 in their conception could make or break the world.

47:48 And so that is basically what the independent

47:51 board members also begin to reflect on.

47:54 They have these conversations amongst themselves where they're like, "Well,

47:59 based on what we're hearing about Altman's behavior,

48:01 like if this was an Instacart,

48:03 would that warrant firing him?" And they concluded, "Maybe not,

48:08 but this is not Instacart." And that's why they were like, "Well, crap.

48:13 Maybe this is actually this does rise

48:16 to the to the bar where we should consider replacing him because

48:20 we are ultimately building a technology that we think could

48:25 have transformative impacts either in the positive or negative direction.

48:30 And so that is what happens.

48:31 It's like these two executives and then

48:33 the independent board members also they were hearing

48:35 other feedback as well from their connections within

48:38 the company with other people in the industry.

48:40 At one point, Adam D'Angelo,

48:41 who is one of the independent board members and the CEO of Kora, uh,

48:45 which is, you know, start a tech startup in the valley,

48:48 he is at a party in San Francisco,

48:51 and he starts to hear some of these rumors that there's something

48:56 weird about the way that OpenAI has structured its OpenAI startup fund,

49:02 which was this fund that they the company

49:05 had created to start investing in other startups.

49:10 Mhm.

49:09 and he realizes they'd never really seen documentation about

49:14 how the startup fund had been set up from Alman.

49:16 And finally they get the documents and it turns

49:18 out that OpenAI startup fund is not OpenAI's startup fund.

49:22 It's Altman's startup fund.

49:24 And this was something like one of several experiences that the independent

49:29 board members were also having where

49:31 they're like there's something not right about

49:34 the fact that there continuously are

49:37 inconsistencies inconsistencies between the way that Altman

49:40 is portraying what is being done versus what is actually being done.

49:46 And so when these two executives approach the board

49:49 or the independent board members, then they're like,

49:52 "Okay, this lines up with also the experiences

49:55 that we've been having." And at that point,

50:00 they then have this series of very intense discussions where they're meeting

50:04 almost every day talking about should

50:07 we actually really consider removing Altman?

50:12 And in the end they conclude, yes, we should.

50:16 And if we're going to do it, we need to do it quickly.

50:19 Because they were very concerned that the moment that Alman found out,

50:22 his persuasive abilities would make it impossible to do.

50:28 And so they end up firing Altman without telling anyone.

50:32 You know, they don't talk to any stakeholders to get them on the same page.

50:37 Microsoft gets a call right before they execute the action saying,

50:41 "We're going to fire Altman." And Microsoft, for anyone that doesn't know,

50:43 are a lead investor in OpenAI at the time.

50:47 Yes.

50:47 One of the only investors in OpenAI at the time.

50:52 And that is what then devolves the whole thing because every single person

50:58 that is affected by this decision is

50:59 now extremely angry that they were not involved.

51:04 And that is what then creates this campaign to bring Altman back.

51:09 And then Alman is reinstalled as CEO days later.

51:13 This company that I've just invested in, it's grown like crazy.

51:16 I want to be the one to tell you about it because

51:17 I think it's going to create such a huge productivity advantage for you.

51:20 Whisper Flow is an app that you can get on your computer and on your phone

51:23 on all your devices and it allows you to speak to your technology.

51:26 So, instead of me writing out an email,

51:27 I click one button on my phone and I can just speak the email

51:31 into existence and it uses AI to clean up what I was saying.

51:34 And then when I'm done,

51:35 I just hit this one button here and the whole email is written for me.

51:38 And it's saving me so much time in a day because Whisper learns how I write.

51:43 So on WhatsApp, it knows how I am a little bit more casual.

51:46 On email, a little bit more professional.

51:47 And also, there's this really interesting thing they've just done.

51:49 I can create little phrases to automatically do the work for me.

51:52 I can just say Jack's LinkedIn and it copies Jack's LinkedIn

51:56 profile for me because it knows who Jack is in my life.

51:58 This is saving me a huge amount of time.

51:59 This company is growing like absolute crazy.

52:02 And this is why I invested in the business

52:03 and why they're now a sponsor of this show.

52:05 And Whisper Flow is frankly becoming the worstkept secret in business,

52:09 productivity, and entrepreneurship.

52:10 Check it out now at whisperflow spelled w i s p r l o w.ai/steven.

52:17 It will be a game changer for you.

52:20 There's a phase a lot of companies hit where they're

52:22 no longer doing the most important thing, which is selling.

52:24 And they get really bogged down with admin.

52:26 And it's often something that creeps up slowly

52:28 and you don't really notice until it's happened.

52:30 Slowly momentum starts to leak out.

52:33 This happened to us and our sponsor Pipe Drive

52:35 was a fix I came across 10 years ago.

52:37 And ever since my teams across my different companies have continued to use it.

52:40 Pipe Drive is a simple but powerful sales CRM

52:43 that gives you the visibility on any deals in your pipeline.

52:46 It also automates a lot of the tedious,

52:48 repetitive, and time-conuming parts of the sales process,

52:50 which in turn saves you so many hours every single month,

52:53 which means you can get back to selling.

52:55 Making that early decision to switch to Pipe Drive was a real gamecher,

52:58 and it's kept the right things front of mind.

53:00 My favorite feature is Pipe Drive's ability to sync your CRM with multiple

53:04 email inboxes so your entire team can work together from one platform.

53:08 And we aren't the only ones benefiting.

53:10 Over 100,000 companies use Pipe Drive to grow their business.

53:13 So, if something I've said resonates,

53:15 head over to pipedive.com/ceeo where you can get a 30-day free trial.

53:21 No credit card or payment required.

53:24 How does a CEO of a major company get fired by the board?

53:28 Because board members,

53:30 there's a quote in your book on page 357 where you say about Ilia saying,

53:33 "I don't think Sam is the guy who should

53:35 have the finger on the button for AGI." Now, I I asked myself this question.

53:40 You know, I work with lots of people here.

53:41 We have 150 people that work in this business and those people know me best.

53:49 Yeah.

53:48 They see me on camera.

53:49 They see me off camera.

53:50 So if they said that we don't think Steven is the right person to host the direc

53:56 Yeah.

53:56 It would take a lot for them to say that.

53:59 Yeah.

53:59 They must have seen some off camera for them to go

54:02 we don't think he's the right person to be on camera.

54:04 Yeah.

54:04 Or for whatever reason.

54:05 And in the case of AI, which is much more consequential than a podcast that is,

54:08 you know, filmed in my old kitchen.

54:10 Um it almost sends a chill down one's body to think

54:13 that the co-founder of a business has gone to the board

54:16 and said this isn't the guy to lead this consequ I mirror

54:20 Marotti then also said I don't think Alman is the right guy

54:23 and then they both left later.

54:26 So then Altman comes back and lo and behold Ilia never comes back.

54:30 So his concerns about the fact that Alman

54:33 founding out would be bad for him manifested.

54:36 He ended up not coming back and Miriam Marotti then left shortly thereafter.

54:41 Quite a lot of these people leave, don't they?

54:43 Open AAI they do.

54:45 So if you consider one of the origin stories of open

54:52 AI is this dinner that happened at the Rosewood Hotel,

54:57 which is a very swanky hotel um right

54:59 right in the heart of Silicon Valley that uh

55:02 was one of Elon Musk's favorites whenever he

55:04 was coming up from LA to the Bay Area.

55:07 And there was this dinner that was there where Altman

55:10 was intending to recruit the OG team that would start OpenAI.

55:15 So he's kind of telling everyone you might have a chance

55:19 to meet Musk because Musk is going to come to this dinner dinner.

55:22 And he cold emails Ilia and gets Ilia to then come because

55:27 and Ilia specifically wants to come because he wants to meet Musk.

55:30 And he also emails all these other people including Greg Brockman, Dario Amade.

55:35 These are all people that ended up working at Open

55:37 and they all almost all of them not not every one

55:40 of them but almost all of them end up working at OpenAI

55:45 and leaving

55:46 almost all of them end up leaving specifically after they clash with Alman

55:52 and Ilia he left and launched a company called Safe Super Intelligence.

56:01 Yeah.

56:00 Which is I mean that's an indirect if I've ever heard one.

56:04 Do you know what I mean?

56:05 Do you know what I mean?

56:06 If someone like co-ounded this podcast with me and then

56:10 they left and started a podcast called Safe Podcasting,

56:15 I I'd take that as a slight.

56:19 I' I'd have people knocking on their door and asking for their texts.

56:24 One of the things that is happening here is

56:30 it is not a coincidence that every

56:32 single tech billionaire has their own AI company.

56:38 Mhm.

56:38 They want to create AI in their own

56:40 image and that's why they keep not getting along.

56:45 And in fact, it's not just don't get along,

56:47 they end up hating each other after working together.

56:51 Mhm.

56:51 and then splinter off into their own organizations.

56:55 So after Musk leaves, he starts XAI.

56:58 After Dario leaves, he starts Anthropic.

57:00 After Ilia leaves, he starts Safe Super Intelligence.

57:03 After Meera leaves, she starts thinking machines lab.

57:07 They want to have control over their own vision of this technology.

57:16 And the best way that they have derived from their experiences

57:23 of trying to put their vision into the arena

57:27 is by creating a competitor and then competing

57:30 with OpenAI and with all the other companies out there.

57:33 Do you think some of these AICOs realize that they are

57:35 quite literally summoning the demon as Elon said 10 years ago,

57:38 but they don't really care because being the person that summoned the demon is

57:44 makes you consequential and powerful and historical

57:48 even if the outcome is potentially horrific.

57:51 Even if there's like a 20% outcome of it being horrific.

57:53 I remember I think it was Dario,

57:56 he's the one that said there's somewhere between a 10% and 25%

58:00 chance of things going catastrophically wrong

58:04 on the scale of human civilization.

58:06 25% is a one in4 chance.

58:10 If you put bullets in a fourchamber revolver and said Steven,

58:16 the upside is you could become a multi-gazillionaire and be remembered forever.

58:20 The downside is that there would be a bullet in your head.

58:22 There is no chance that I would take take that bet

58:25 with a 25% potential chance of things going catastrophically wrong.

58:31 So, I have a very long answer to this because

58:36 do they know if they're summoning the demon?

58:37 It really depends on what we define as summoning the demon.

58:40 And in this particular case, to go back to what we were saying before,

58:46 there's a mythology that the AI industry uses

58:50 where summoning the demon is an integral part

58:53 of convincing everyone that therefore they can be

58:58 the only ones that are developing this technology.

59:01 I got it.

59:01 So on one end, you got to say if we don't, China will and that's terrible.

59:06 Yeah.

59:06 But if we let anyone else do it other than me, then we're as well.

59:11 Exactly.

59:11 So that means that I have to do it and you have to give me money and support.

59:14 Exactly.

59:14 So when they're saying these things,

59:18 we should understand it as not as like a genuine

59:21 prediction based on what they're seeing because first of all,

59:24 we don't predict the future.

59:25 We make it.

59:26 We should understand this as an act of speech to persuade

59:30 other people into believing that they should seed more power,

59:35 more resources to these individuals.

59:37 And so, do they know that they're summoning the demon?

59:41 I mean, they are purposely trying to create this this feeling within

59:48 the public that they are because it is a crucial part of their power.

59:53 But do they if we were to define just do they realize that the things

59:59 that they are doing are having already really

1:00:02 harmful impacts all around the world on vulnerable people,

1:00:06 vulnerable communities, vulnerable countries.

1:00:09 That's where I'm like maybe yes, maybe no.

1:00:11 and they don't really care because in the frame of mind like

1:00:19 I sometimes use the analogy that the AI world is like Dune.

1:00:22 Dune for anyone that doesn't know Dune

1:00:24 science fiction epic written by Frank Herbert and it's set in this intergalactic

1:00:29 era where there are all these houses and they're fighting each other for spice.

1:00:33 So it's a call back to colonialism and empire

1:00:36 and they all are trying to control the spice.

1:00:38 But one of the features of this story is that there

1:00:41 are these myths that are seated on the different planets

1:00:45 about a a religious myth basically about the coming

1:00:48 of the Messiah that are used as ways to control the people.

1:00:52 And Paul at Trades when he arrives

1:00:56 at the planet Iraqis uh with with the intention

1:00:59 of um trying to then fight against the empire and um avenge his father's death.

1:01:07 He steps into a myth that has been seated on this planet that says

1:01:12 that one day there will be a Messiah that comes and saves the planet.

1:01:15 So he steps into the role of the Messiah and leans into this idea in order

1:01:21 to better control the people and rally them behind

1:01:24 him as a leader to help with this quest.

1:01:29 He knows that it's a myth in the beginning,

1:01:32 but because he lives and breathes and embodies it,

1:01:36 it kind of starts to blur in his mind whether

1:01:38 this is really a myth or whether he's really the messiah.

1:01:42 And this is what I think happens in the AI world.

1:01:46 On one hand, there are all these executives

1:01:49 that actively engage in mythmaking because, you know,

1:01:54 I have all these internal documents that I write about in the book where

1:01:57 they are very keenly aware of how to bring the public along with them

1:02:03 by showing them dazzling demonstrations of the technology

1:02:07 by using crafting a mission that will

1:02:10 sound really good uh and and and make

1:02:13 people give more leniency to their companies.

1:02:18 So they know they're doing the mythmaking

1:02:20 and also I think many of them lose themselves

1:02:23 in the myth because they have to live

1:02:27 and breathe and embody it day in and day out.

1:02:29 And so when you know Daario says he thinks that 10 to 25%

1:02:34 of the future could be catastrophic or or whatever the probability is 10 to 25%.

1:02:40 He is actively engaging in the mythmaking

1:02:42 but also he's losing himself in the myth.

1:02:44 Like I think if you were to ask him,

1:02:46 "Do you genuinely believe that?" He would be like, "Yes,

1:02:48 I genuinely believe that." Because there's been a blurring

1:02:52 of when he's saying something just to say something versus when

1:02:57 he actually believes what is he's required to believe

1:03:03 in order to then continue doing the things that he's doing.

1:03:09 And this is the whole psychology of cognitive dissonance, right?

1:03:12 where you the brain struggles to hold

1:03:14 two conflicting worldviews at the same time.

1:03:16 So it's it's incentivized or it endeavors to dismiss one.

1:03:19 So if you you know if you wanted to be a healthy person but also a smoker.

1:03:23 Um and I pointed out that smoking is bad for you.

1:03:25 The first words out of your mouth are going to be yes but

1:03:28 smoking helps me with stress.

1:03:30 Yeah, but I only do it when I think I don't

1:03:33 know I kind of see that at the moment because these companies

1:03:36 have to raise extortionate like huge amounts of money to fund

1:03:40 their AI research and they're building out all of these data centers.

1:03:44 So when they're out in the public, they're always fundraising.

1:03:47 All of these major companies are fundraising all the time at the moment.

1:03:50 So you can't be fundraising and saying,

1:03:51 "I'm going to destroy your children's future potentially.

1:03:53 There's 25% chance that your children aren't going

1:03:55 to have a great life." Which might be the truth.

1:03:59 I mean that is actually what they say Dario.

1:04:01 This is what famously Dario Amade does.

1:04:03 He's like he does that but the others Sam's not doing that as much anymore.

1:04:06 Yes.

1:04:06 And it's because you know it goes back to like each of them kind

1:04:11 of distinguish themselves a little bit

1:04:13 as as the brand that they need to project.

1:04:17 Do you think any of them are more have a stronger moral compass than others?

1:04:21 cuz I think Dario often gets the credit for having more of a, you know,

1:04:26 more of a backbone and being more conscious of implications.

1:04:31 He does get a lot of credit for that.

1:04:33 He's from Claude and Anthropic.

1:04:34 For anyone that doesn't know,

1:04:37 I don't think it truly matters that question,

1:04:41 the answer to that question, because to me,

1:04:44 even if you were to swap all the CEOs for someone

1:04:48 that people would say is better at running these companies,

1:04:52 it doesn't fix the problem that I identify in the book,

1:04:55 which is that there is a system

1:04:57 of power that has been constructed where these companies

1:04:59 and the people running these companies get

1:05:02 to make decisions that affect billions of people's lives.

1:05:04 lives around the world and those billions of people

1:05:07 do not get any say in how it goes.

1:05:11 Those people, they can go to the polls, right?

1:05:13 So, if the public are sufficiently educated,

1:05:15 they can go to the polls and pick a leader that says

1:05:18 they're going to legislate or pass laws or try and pass laws.

1:05:24 Yes.

1:05:23 But at the speed and pace at which

1:05:26 these companies operate and at the sheer scale and size,

1:05:30 they're able to also spend extraordinary amounts of money,

1:05:33 hundreds of millions in this upcoming midterms

1:05:35 to try and kill every possible piece of legislation that gets in their way

1:05:38 and craft legislation that would codify their advantage.

1:05:42 And so to me, I think sometimes as a society,

1:05:46 we obsess a little bit with are these leaders good or bad people?

1:05:53 And to me the bigger question is is the governance

1:05:57 structure that we've created a sound one or that allows broad

1:06:01 participation or an anti-democratic one

1:06:04 that has consolidated this decision-making power

1:06:06 in the hands of the few because no person is perfect.

1:06:09 It does I don't I don't care who is on at the top of these companies.

1:06:13 they're not going to have the ability to make decisions

1:06:17 on behalf of so many people around the world who live

1:06:20 and talk and um and and have a culture and history

1:06:24 that are fundamentally different from them without things going wrong.

1:06:29 And so that is why throughout history we've moved from empires to democracy.

1:06:36 It's because empire as a structure is inherently unound.

1:06:41 it does not actually maximize the chances of most

1:06:45 people in the world being able to live dignified lives.

1:06:49 I'm going to try and take on their point of view.

1:06:51 So, this is me playing devil's advocate.

1:06:53 Okay.

1:06:54 But Karen, if the US don't continue

1:06:59 to accelerate their research with AI, at some point,

1:07:02 China's model is going to become so smart and intelligent that we're basically

1:07:07 going to have to rent it off them and we're going to be,

1:07:09 you know, they'll get the scientific discoveries.

1:07:10 They'll discover the new era of autonomous

1:07:13 weapons and we will be their backyard.

1:07:17 And like logically that argument does appear to be pretty true.

1:07:22 No, it's not.

1:07:23 If we scale up, if we just imagine any rate of change with this intelligence,

1:07:26 at some point we're going to come to a weapon

1:07:29 that could theoretically disable um all of the United States electricity,

1:07:34 their weapons systems.

1:07:36 It would know exactly how to disable the United States

1:07:39 from a cyber perspective because it would be that smart.

1:07:42 All you've got to imagine is any rate of improvement

1:07:44 of any period any sort of long period of time.

1:07:47 So this is a theory that might be true and if it's true

1:07:52 I mean yeah any theory might be true

1:07:55 but but if but but you know again going to this point of like even if

1:07:58 it's a small percentage it's worth paying attention

1:08:00 to on the other side of the foot.

1:08:02 This is a theory that people talk about.

1:08:04 It could be the case that the most

1:08:07 intelligent civilization is going to be the superior civilization.

1:08:12 Logically, that's a pretty sound thing to say.

1:08:13 No.

1:08:14 So, there's a lot of a lot of fundamentals in this argument that would

1:08:19 need to be true in order for this to be a viable argument.

1:08:22 And let's knock them down one by one.

1:08:24 So the first one is that these systems are intelligent

1:08:31 and that just scaling them is going to bring us more intelligence.

1:08:34 So far so true.

1:08:36 No, it's actually not because first of all again we don't actually know if

1:08:42 these systems are like intelligence is not

1:08:45 it's not like the right analogy almost.

1:08:47 It's sort of like it's like is a calculator

1:08:51 a calculator can do math problems faster than a human.

1:08:54 Does that make it intelligent?

1:08:56 It has a narrow intelligence because they're solving a narrow

1:08:58 problem which is like 1 plus 1 equals 2.

1:09:00 But and these systems, they actually also are

1:09:04 quite narrowly intelligent in the sense that even

1:09:07 though these companies say that they're everything

1:09:09 machines that can do anything for anyone,

1:09:11 they actually can only do some things for some people.

1:09:13 This is like the jagged frontier of these AI

1:09:16 models like some of the capabilities are quite good,

1:09:18 other capabilities are not that good.

1:09:20 You know why that happens?

1:09:22 is because the company can only

1:09:23 focus on advancing certain types of capabilities.

1:09:25 It can't literally focus on advancing all types of capabilities.

1:09:29 They have to actually set their mind to advancing a certain

1:09:31 by gathering the data that is needed for that capability

1:09:34 by taking uh you know getting a bunch of human contractors

1:09:39 to annotate and train the model to do that exact thing.

1:09:43 And so scaling these models is actually a perpendicular question to are

1:09:51 we actually getting more cyber capabilities

1:09:55 specifically and more military capabilities specifically.

1:09:58 I would argue that most of the most of the top people in AI

1:10:01 believe that the intelligence is going to continue to scale for some time.

1:10:05 a lot of them do like Jeffrey Hinton does.

1:10:07 And again, it's it's back to his hypothesis about how human

1:10:11 intelligence works and what the appropriate model of the brain is.

1:10:15 His hypothesis throughout his career has been the brain is a statistical engine.

1:10:20 But that's his hypothesis and that is not universally agreed

1:10:24 upon especially among people that are not in the AI world.

1:10:27 When you talk with neuroscientists and psychologists,

1:10:29 people who actually study human intelligence in the human brain,

1:10:32 that is where you start to get a lot

1:10:34 of debate and disagreement about this particular view that Hinton has.

1:10:40 And so this is kind of like one of the one of the things is like AI

1:10:46 is already being used in the military and has

1:10:49 been used in the military for a long time.

1:10:51 But ex specifically accelerating large language models isn't

1:10:58 just the only path for getting military cap.

1:11:01 like the companies would have

1:11:02 to choose to specifically pick military capabilities

1:11:06 to accelerate not just like general intell it's like you know what I'm saying

1:11:11 like they create this myth that they are actually pushing the frontier

1:11:15 of all of the capabilities of the model but that's not what's actually happening

1:11:19 internally and I have I had hundreds of pages of documents on like

1:11:22 how they were specifically training models they pick what capabilities they want

1:11:27 to advance and you know how they pick them it's based on which

1:11:31 industries countries would be able to pay

1:11:32 them the most money for their services.

1:11:34 So they pick finance, law, medicine, healthcare, commerce.

1:11:41 It's not actually intelligent like a like a a baby where

1:11:46 you the the more that you that the baby grows up,

1:11:48 they start having this like general these general abilities.

1:11:52 I think I have jagged intelligence.

1:11:54 I'll be honest.

1:11:54 I wasn't going to say it,

1:11:55 but I think I know a little I know a little bit about uh No,

1:12:00 I know a lot about a little bit.

1:12:02 Yeah, but if but you also have

1:12:03 the capability to learn and acquire knowledge by yourself.

1:12:06 And you also have the ability to choose

1:12:07 what you're going to learn and acquire by yourself.

1:12:10 It's not easy and it takes a lot more time than these models.

1:12:12 It seems less compute, but and you can learn how to drive in one place

1:12:16 and then immediately know how to drive in another place.

1:12:18 These models cannot do that.

1:12:20 Every time a self-driving car is shifted to another location,

1:12:25 it has to completely retrain on that location.

1:12:27 It's like all the self-driving cars.

1:12:29 I mean, we're sitting in Austin right now and there's

1:12:30 all these self-driving cars that are driving through Austin.

1:12:34 But when one of them learns, they all learn which is which

1:12:37 well it's just because it's a it's an operating

1:12:41 system that is has an AI model as part

1:12:44 of it and you're training the AI model and then

1:12:46 you deploy that AI model across all the self-driving

1:12:48 a big advantage because if one optimist robot learns one thing

1:12:53 in one factory they all learn it and imagine that imagine if

1:12:56 humans if we all learned what all the other humans learned

1:12:59 that would be that would give us such an unbelievable competitive advantage.

1:13:02 I mean one of the ways we did that is through communication.

1:13:04 They could not because they could be learning

1:13:05 the wrong thing which has also happened again and again

1:13:07 with these technologies is that all of them then learn

1:13:10 the wrong thing and they all have the same failure mode.

1:13:12 I mean part of the resilience of human society is that we

1:13:15 do have different expertises and we also have different failure modes.

1:13:19 I think sometimes we hold AI models to a higher standard than we hold humans to.

1:13:23 And in a weird because I I' I'd hear on stage we're in we're in Austin

1:13:26 at the moment and I'd hear people go ah

1:13:28 but you know them AI models they hallucinate sometimes.

1:13:31 I'm like, "Have you met a human?" Like, I I hallucinate all the time.

1:13:36 I can barely spell or do math.

1:13:40 So, yes, but it's it's once again like

1:13:41 using this analogy that was specifically picked

1:13:44 in the early days of the field as a way to market these technologies.

1:13:48 like we're repeatedly using the intelligence analogy and relating

1:13:52 these machines to human intelligence as a a way

1:13:56 to try and gauge whether or not it is good or worthy or capable in society.

1:14:01 I think the output is the thing that really

1:14:02 m is the most consequential which is like

1:14:04 okay it might have a different brain and a different

1:14:06 system but does it arrive at the same capability like does it is it able to do

1:14:10 surgery on someone's brain is it able to drive a car like my car drives itself

1:14:14 in in Los Angeles I don't touch the steering wheel

1:14:16 and I can drive for many many hours and in here in Austin I just saw the ones

1:14:20 the other day where they've removed the steering wheel

1:14:22 and the pedals the new cyber cabs so I go it doesn't really matter if it's using

1:14:25 a different system if it's navigating through the world

1:14:27 as a car it has a better safety record

1:14:29 than human beings Um then as far as I'm concerned,

1:14:34 intelligence or not, it's like yes, you know,

1:14:37 but that was not the original argument that you made,

1:14:38 which was like these systems are just generally going

1:14:41 to become more intelligent across different things based on the prediction.

1:14:46 This is a prediction that you're making, right?

1:14:47 Like that and this is a prediction that all the AI um

1:14:50 Ilia's making, Dario's making, Elon's making,

1:14:53 Zuckerberg's making, man's making, Dennis is making.

1:14:56 And do you know what the common feature of all of them is?

1:14:58 They profit enormously off of this myth.

1:15:02 Elon has recently spearheaded the construction of Colossus,

1:15:05 a massive supercomputer in Memphis housing a 100,000 GPU specifically

1:15:09 to scale up their API models faster than their competitors.

1:15:13 It appears that they've all converged around this idea

1:15:15 that you can brute force your way to greater, more generalized intelligence.

1:15:20 They've converged around the idea that you can brute force your way into models

1:15:24 that they can sell to people for automating

1:15:27 certain tasks that are that are financially lucrative.

1:15:30 And I heard Elon say that if you're a surgeon, there's just no point.

1:15:33 He was like, don't train to be a surgeon.

1:15:35 He says in a couple of years time, Optimus and AI generally are going to be

1:15:39 better than any surgeon that's ever lived.

1:15:41 Yeah.

1:15:41 You know, do you think these things are true?

1:15:42 Well, you know, I I'm pretty sure it was Hinton that famously

1:15:45 slash infamously said there would be no need for radiologists anymore.

1:15:50 There would be no need for radiologists anymore

1:15:52 in he set a deadline that we've already passed.

1:15:55 I don't remember how many years.

1:15:58 Radiology is doing great as a profession.

1:16:00 Do you think it will be in 5 years?

1:16:03 Okay.

1:16:03 So, this this once again goes back to this question of like

1:16:06 why do we build technology and why should we specifically be building AI?

1:16:10 Okay.

1:16:10 And for me like the whole project of technology

1:16:14 development advancement is not to advance technology for technologies sake.

1:16:19 It's to help people.

1:16:21 And there have been lots of research that has shown that actually the best

1:16:26 outcomes for people in a healthcare setting is for the radiologist to have

1:16:31 the AI model in their hands and for the for the human expert

1:16:39 to use the AI model as a tool as an input into their judgment.

1:16:44 And it is that combination that leads to the most accurate and early diagnoses

1:16:51 of certain types of cancer that then help improve the prognosis of the patient.

1:16:55 Do you believe that in the coming years all the cars

1:16:58 pretty much all the cars on the road will be driving themselves?

1:17:01 No.

1:17:01 You don't you don't think so?

1:17:02 Mm-m.

1:17:02 How come?

1:17:03 Because of the way the technology works.

1:17:06 Because because these are statistical I mean currently

1:17:10 the way that AI models are primarily developed.

1:17:13 They're statistical engines.

1:17:14 You have what's called a neural network,

1:17:16 which is a piece of software that has a bunch of densely connected nodes and

1:17:22 like parameters.

1:17:23 Is this what they call parameters?

1:17:25 Yeah, pretty much.

1:17:25 And you're just pumping a bunch of data into it and then it's analyzing

1:17:29 the data and creating this all

1:17:32 of these finding all these correlations in the data,

1:17:34 finding all these patterns and then it's through those patterns

1:17:37 that the machine is then able to act autonomously, right?

1:17:41 And so the way that they're training a self-driving car

1:17:44 is they're they're recording all this footage and then they

1:17:48 have tens of thousands or hundreds of thousands of human

1:17:50 contractors that draw literally around every single vehicle in the footage,

1:17:58 every single pedestrian, every single traffic light,

1:18:01 every single lane marking and label it exactly as such.

1:18:06 So that then it's fed into an AI model that can identify

1:18:09 all of these different components and then it's connected to another piece

1:18:14 of software that is not AI that's saying okay if you if

1:18:18 the AI model recognizes the pedestrian we do not run over the pedestrian.

1:18:23 If the AI model recognizes a red traffic light we stop.

1:18:27 And so the like the thing about

1:18:31 statistical engines is that it's based on probabilities.

1:18:34 It's not based on deterministic logic.

1:18:37 So systems make errors all the time and it's impossible.

1:18:42 It is technically impossible to get them to stop making errors.

1:18:48 Humans make errors way more than systems in this case.

1:18:52 Like the safety record is like isn't it like 10 times more safe to be

1:18:55 driven in a Tesla with autonomous driving than it is to for a human to drive?

1:18:59 It depends on the place.

1:19:01 It depends on whether the Tesla was trained

1:19:03 to specifically navigate the place that you're driving.

1:19:05 Get drunk because if it's in Mumbai,

1:19:10 in some place in Vietnam, no, it would not be safer.

1:19:13 I WOULD MUCH RATHER be driven

1:19:16 by someone that has been driving in that place their whole life.

1:19:20 I'm I'm not arguing against like the fact

1:19:21 that in certain places where the car has been explicitly trained

1:19:25 to drive in this place that it has a better

1:19:28 safety record than the humans that are driving in that place.

1:19:30 But you specifically asked if I think that all of the most cars

1:19:35 most cars in the world in the US in the United States cuz we're here.

1:19:40 I don't actually think that it's like imminently on the horizon 10 years.

1:19:44 No, I don't think so.

1:19:45 I sat with Dra from Uber and he's pretty convinced

1:19:47 that his 9 million couriers will be replaced by autonomous vehicles.

1:19:51 I mean, how long have has self-driving cars been invested in thus far?

1:19:56 It's been more than 10 years.

1:19:58 And what percentage of cars right now are autonomous on the US roads?

1:20:04 I mean, so part of it is it's actually not a technical problem, right?

1:20:07 Like part of it is also social problem

1:20:10 like do people even trust getting into these vehicles?

1:20:12 Part of it is also a legal problem which

1:20:15 is if the car the self-driving car kills someone, which it has happened.

1:20:20 Yeah, it has happened.

1:20:22 Who is responsible?

1:20:24 So, in the case in LA,

1:20:25 it was both Tesla and the driver because the driver dropped their phone,

1:20:29 they looked down, and this was a couple of years ago, I believe.

1:20:32 Um, and they went to grab their phone and they hit someone,

1:20:35 and so it went to court,

1:20:36 and they were held both responsible, both the driver and Tesla.

1:20:40 Um, in terms of Tesla, pretty much everyone that gets the car,

1:20:46 it comes with autonomy now for pretty much most people, I believe.

1:20:49 Partial autonomy.

1:20:50 Yeah, it's called full self-driving at the moment where it's like

1:20:52 I mean, yes, it is called full self-driving.

1:20:54 Full self-driving supervised where you kind of have to be looking in the d.

1:20:57 You have to be looking in the right direction, but Yeah.

1:20:59 So, it's partial autonomy.

1:21:01 And here in Austin, it's full autonomy cuz there's no steering wheel.

1:21:06 Yeah.

1:21:06 On the new car.

1:21:06 Um, so you can't drive it anyway.

1:21:08 But it is, you know, the Model Y is the undisputed highest selling car,

1:21:12 bestselling car in the world across all brands.

1:21:16 Well, I guess my point here is like these predictions where

1:21:19 they say AI is going to completely change transportation and driving.

1:21:24 It's going to completely change lawyers aren't going to have jobs.

1:21:26 Accountants aren't going to have jobs.

1:21:28 Um, do you believe that they are true?

1:21:30 Do you believe that there's going to be mass job displacement?

1:21:33 Okay, so I do think that there is going to be

1:21:35 huge impacts on employment and we already seeing those impacts.

1:21:39 It is not simply because the AI models are just automating those jobs away.

1:21:44 It is specifically because the models

1:21:48 are improving in certain capabilities based

1:21:50 on what the companies that are developing them choose to improve them on.

1:21:55 And executives at other companies are then

1:21:58 deciding to fire or lay off their workers because they think that AI can replace

1:22:04 the worker irrespective of whether that might be true.

1:22:07 And there, you know, there have been cases of like the CLA CEO who

1:22:10 laid off a bunch of people thinking that he would

1:22:12 replace everyone with AI and then it didn't actually work

1:22:14 and he had to ask some people to come back.

1:22:16 I actually DM'd him about this.

1:22:18 If you're hearing this, this is because I've

1:22:19 DM'd Sebastian and he's fine with me sharing this.

1:22:22 He said, because I've heard his name mentioned a lot and so when I when

1:22:25 we talked about AI in the past

1:22:26 and people mention Sebastian and Cler as the example,

1:22:30 I wanted to clarify with him what the truth was.

1:22:32 He said, "It's great to hear from you.

1:22:33 Um, I think sometimes people struggle with two

1:22:35 things can be true at the same time.

1:22:38 I think it might be time to come back on your podcast.

1:22:41 To your point, this is the media misinterpreting my tweet.

1:22:44 We are doubling down on AI more than ever.

1:22:46 Cler is shrinking with almost 100 employees per month due to AI.

1:22:50 We used to be 7,400 at the peak.

1:22:53 A year ago, 5,500.

1:22:56 Now we're 3,300.

1:22:58 And by the end of summer, so this was last year, will be 3,000 people.

1:23:03 AI handles 70% of our customer service conversations at this moment.

1:23:08 This is because we have realized that with AI,

1:23:10 the production cost of software comes down to almost zero.

1:23:13 Just like manufacturing used to be all handcrafted and then the machines came.

1:23:17 Code used to be all handcrafted up until a few years ago.

1:23:20 And now it is machine produced.

1:23:23 And ultimately we pay people more than

1:23:26 ever for the unique handcrafted man-made stuff.

1:23:29 China is a bank.

1:23:30 People will want to connect to humans not only machines.

1:23:33 They want us to be personable, relatable, even flawed.

1:23:37 So we need to make sure while we are automating replacing with AI in parallel,

1:23:42 we make sure we offer a super available human experience.

1:23:47 I'm really glad you read this because I think

1:23:49 it touches on some really important nuances to the AI.

1:23:55 Yeah.

1:23:55 Like the impact that AI is going to have on employment.

1:23:57 So I think the there's often these binary narratives.

1:24:01 It's like AI is going to come for every job.

1:24:05 Mhm.

1:24:05 Or people say AI is not actually working and it's not actually coming for jobs.

1:24:09 And like the reality is it's coming for jobs.

1:24:12 There are definitely jobs that are being automated

1:24:15 away because of the capabilities of their models.

1:24:18 And there's also jobs that are being lost because executives are deciding to lay

1:24:21 off the workers even if the models

1:24:23 don't match the capabilities because it's good enough.

1:24:25 Like they would rather have the good enough model for way cheaper

1:24:28 or they made a mistake with hiring.

1:24:30 They blowed their team and it's a great convenient thing to say.

1:24:33 Exactly.

1:24:33 Like there's there's there's many reason but like

1:24:35 clearly we're already seeing impacts on the job market.

1:24:37 Like the um US jobs report that came out earlier this year showed that there has

1:24:44 been a decline in hiring is a slowdown

1:24:47 in hiring across especially white collar professional industries.

1:24:53 And you saw Anthropic's report the new this week.

1:24:55 The TLDDR is it matches kind of what you were saying where they Anthropic looked

1:24:58 at exactly how people were using their models

1:25:01 and they looked at like what people are saying.

1:25:05 Yeah.

1:25:04 And they said that there's been a 40% reduction in entry- level jobs

1:25:08 in particular and then they made this graph

1:25:09 which has gone viral over the internet.

1:25:11 The red shows where we are now in terms of capability

1:25:14 and based on how people are currently using the models they prediction

1:25:18 extrapolated out that the blue part will be the disrupted parts.

1:25:20 This is the things that they say AI can do right now,

1:25:24 but people don't realize it yet.

1:25:25 So, if you look at it, it's like it's kind of all the stuff you would expect.

1:25:30 Yeah.

1:25:29 It's the physical real world human stuff which robots maybe can do someday

1:25:33 like construction or agriculture that are untouched,

1:25:36 but like office and admin, um like saying finance stuff, math,

1:25:41 and notice that these are all the things that I just named that they purposely

1:25:45 finance, math, law, media and arts.

1:25:47 That's me cooked.

1:25:50 Yeah.

1:25:50 office and admin.

1:25:50 I mean they do focus a lot on like assistant type and managerial work.

1:25:57 So but but the the other thing that the CLO

1:25:59 CEO said was but people also want human experiences.

1:26:05 So it's not actually just about the capabilities of the models.

1:26:08 It's also about what people want like some things they

1:26:12 would turn to AI for and some things they wouldn't irrespective

1:26:16 of whether or not AI is capable of doing it

1:26:19 but because of a preference that they want humanto human interaction

1:26:25 and so what we're seeing right now is yeah

1:26:29 the the thing that happens with every wave of automation which

1:26:32 is that there is a bunch of entry-level work that gets

1:26:35 automated away and there There are also new jobs created,

1:26:39 but the jobs that are created are one in one of two categories.

1:26:43 There are people that get even higher skilled jobs and what he

1:26:47 was saying like we pay people more for like the handcrafted code now

1:26:52 and there's also the people who get way worse

1:26:54 jobs and so there was this amazing article in New

1:26:58 York magazine that was talking about how a lot

1:27:01 of people are getting laid off and then they

1:27:05 end up working in data annotation which is

1:27:08 the labor that I've been referring to throughout this conversation

1:27:11 that companies need in order to teach their models

1:27:14 the next thing that the companies are trying to automate.

1:27:17 And so like a marketer gets laid off and then they

1:27:20 go and work for a data annotation firm to train the models

1:27:25 on the very job that they were just laid off in which

1:27:29 will then perpetuate more layoffs if that model then develops that skill.

1:27:35 And the article was talking about how this has become a huge catchall

1:27:43 for a lot of people that are

1:27:44 struggling with finding job opportunities right now,

1:27:47 including like awardwinning directors in Hollywood that are actually secretly

1:27:52 doing this data annotation work to put food on the table.

1:27:55 And so when they talk about there's going to be mass unemployment and then

1:28:01 there's going to be some new jobs created that we can't even imagine,

1:28:04 I think a lot of these narratives rarely talk about like first of all,

1:28:08 why are some jobs going away?

1:28:10 It's not just because of the model capabilities,

1:28:11 it's also because of executive choices and because of the rhetoric

1:28:14 that they use if they want to just downsize.

1:28:16 Um, but the other thing that is rarely talked about is the jobs,

1:28:21 a lot of the jobs that are created are way worse than the jobs that were there

1:28:27 and it breaks the career ladder.

1:28:29 So, it's the entry level and the mid tier jobs that get gouged out.

1:28:33 It's higher order jobs and then way more lower order jobs that get created.

1:28:40 And so, how do people continue to progress in their careers?

1:28:44 There's no more rungs on the ladder.

1:28:46 I actually don't know the answer to this question.

1:28:47 And I've been furiously trying to find

1:28:49 a good answer to this question because I can, you know, everything is theory.

1:28:53 And for my audience, I would say most of my audience don't run businesses.

1:28:57 A lot of them do, a lot of them aspire to, but they don't run businesses.

1:28:59 So, they're kind of, they're also in the land of theory.

1:29:02 They're hearing lots of different things.

1:29:03 Jack Dorsey does his tweet saying he's halfing his headcount because of AI.

1:29:06 They don't know what's true.

1:29:07 They don't know the sort of internal economics at Jack's company

1:29:10 and did he bloat the company during the pandemic and he's just

1:29:12 using this as an excuse to make this share price spike seven

1:29:15 points because his investors now think they're an AI company or whatever.

1:29:19 Mh.

1:29:18 It's hard to pass through.

1:29:19 So eventually I go, okay, what am I doing?

1:29:22 I have hundred hundreds of team members,

1:29:24 probably 70 companies I invest in, maybe five

1:29:26 or six that I'm like the lead shareholder in.

1:29:28 What am I actually doing on a day-to-day basis right now?

1:29:30 I am I'm also I also consider myself to be head of recruitment

1:29:34 but in the last month in particular

1:29:36 I have met extremely capable candidates in terms

1:29:38 of cultural alignment hard work those kinds of things but I've had to take

1:29:42 a great deal of pause because when I run the experiment of can I

1:29:45 get an AI agent to do that exact same thing the answer is increasingly yes

1:29:50 especially in a world of open clause and so what I'm curious like

1:29:56 now you confront this decision where you're

1:29:58 seeing in this short-term period you could

1:30:02 just choose the AI agent and in the long-term period there is no career ladder.

1:30:10 So, so who are you promoting into these senior roles?

1:30:13 Like what how do you resolve it for your own company?

1:30:16 Yeah, it's a good question.

1:30:17 So, there's kind of two ways I'm thinking about it.

1:30:18 I think really deep expertise is very very valuable

1:30:22 because if you're now the orchestrator of potentially AI agents,

1:30:25 it's really about um having a deep understanding of the right question

1:30:28 to ask and and that's someone who has deep expertise on something.

1:30:31 So I need my CFO because if she's going to be orchestrating our team

1:30:35 of agents that might be doing financial analysis or whatever else,

1:30:38 she needs to understand what to tell them to do in our company.

1:30:44 Mhm.

1:30:43 And in turn financial analysts can't do that.

1:30:45 They need this the 50 odd years of experience that you know CLA has.

1:30:49 On the other end, I need Cass.

1:30:51 Cass is 25.

1:30:53 Cass knows everything about AI agents.

1:30:55 He's a young Japanese kid who's highly highly curious.

1:30:58 You know, on the weekend, he's building AI agents to solve problems in my life.

1:31:02 I need those two kinds of thinking,

1:31:04 which is highly proficient agent maxing young kids

1:31:07 or they don't necessarily need to be young,

1:31:08 but like really lean in high curiosity.

1:31:10 That's creating a force multiplier in my business.

1:31:12 And then I need deep expertise.

1:31:14 Now the everything else outside of there is another one I've thought

1:31:18 of another group is like people with extremely great IRL people skills

1:31:23 because we do meet people in real life.

1:31:26 We greet you when you arrive here.

1:31:27 We greet we when we go for lunch with big clients

1:31:29 that we have whether it's Apple or LinkedIn or whoever it might be.

1:31:32 We, you know, we need to smoosh.

1:31:36 Mhm.

1:31:35 And we have teams who, you know, are in person in the office.

1:31:38 So, we we do a lot of stuff IRL

1:31:40 and increasingly we're building communities even for this show.

1:31:42 We're doing community events all around the world.

1:31:43 So, we need people that are good at that as well.

1:31:46 IRL, bringing people together in real life and organizing stuff.

1:31:49 Those are the three groups of people that I'm like,

1:31:51 you know, irreplaceable right now.

1:31:53 And if you were to to all of the all the roles that could be done by AI agents,

1:32:00 if we were to replace them with AI agents,

1:32:01 do you think you would still have these three roles pools of people to hire

1:32:06 and promote into the three critical things that you need in the long term?

1:32:10 If things carry on at the the current rate of trajectory, yeah,

1:32:14 one could assert that even those roles would experience pressure.

1:32:18 If you just imagine like people think

1:32:19 of things either statically or linearly or exponentially.

1:32:22 Yeah,

1:32:22 you imagine an exponential rate of improvement, which is kind of what I've seen.

1:32:25 Even like a 10% compounding rate of improvement at some point,

1:32:32 at some point, at some point,

1:32:34 I think what remains is actually the IRL irreplaceably human stuff,

1:32:39 human to human, our Maslovian needs of being

1:32:41 in person like we are now aren't going to change.

1:32:43 We need connection.

1:32:44 Humans get very sick when they don't have other

1:32:46 human beings in their life and strong, deep relationships.

1:32:50 100% agree.

1:32:50 So that stuff is going to matter a whole lot.

1:32:52 I have this contrarian weird take that actually maybe this is

1:32:55 the first technology that's going to deliver on the promise

1:32:58 of making us human and connected because we're going to be

1:33:00 rendered useless of everything else other than what humans are good at.

1:33:03 Cuz all the other technology said, "Oh, we're going to make you more connected,

1:33:06 connecting the world." And they disconnected the world and isolated the world.

1:33:09 But maybe this is the one.

1:33:10 It's so intelligent now that it doesn't

1:33:12 need us to around in spreadsheets anymore.

1:33:14 Do you see that actually happening in real time right

1:33:18 now that it's making us more able to be in person,

1:33:23 connected with one another, having deeper social community engagements.

1:33:29 Yes.

1:33:30 Yes.

1:33:29 And I'll give you some data points.

1:33:32 Okay.

1:33:31 Data point number one,

1:33:32 the Financial Times released a report on social media usage.

1:33:36 And what they saw is 2022 was the peak and it's plateaued ever since.

1:33:40 The generation that's plateaued the fastest

1:33:42 and heading down is the younger generations.

1:33:45 The boomers are still off to the races, right?

1:33:47 So on Facebook and stuff.

1:33:48 And then you look at the way Gen Alfa are using social media.

1:33:52 They're not posting as much.

1:33:53 They call it uh posting zero.

1:33:55 They're scrolling sometimes,

1:33:56 but they're in dark social environments like WhatsApp and Snapchat and iMessage.

1:33:59 They're not like performing to the world.

1:34:01 They also value IRL experiences much more than any other generation.

1:34:04 They're like not getting smashed.

1:34:05 We're seeing every brand has a run club.

1:34:08 um I mean runs exploding around the world and we're

1:34:11 seeing this real sort of sort of almost like

1:34:14 innate realization that like technology let us down at some

1:34:18 fundamental level like dating apps let us down social

1:34:21 networking kind of has let us down and we're seeing

1:34:23 I think maybe a bifocation of society where a lot

1:34:26 of people are going this like I want to go back to what it is to be a human

1:34:29 and I I would imagine that in such

1:34:31 a world where intelligence is so sophisticated that we no

1:34:34 longer needed to sit at laptops and like I

1:34:37 think screen time is going to continue to fall.

1:34:38 I think you go into an office, you're not going to see people sat at laptops.

1:34:40 You're gonna see something completely different.

1:34:43 And I think maybe, you know, and then we talk about robots and Optimus robots.

1:34:48 Elon says there'll be 10 billion Optimus robots.

1:34:50 Elon has been wrong with timing before.

1:34:54 He's almost never been wrong on the big things completely.

1:34:58 He's just his timing is got a bad track record.

1:35:01 Um, so I think he's he's probably right.

1:35:03 You know, I think I've I've got some people on the way

1:35:05 from Boston Dynamics and these other big companies like Scale AI,

1:35:08 and they're actually bringing the robots here to show it,

1:35:10 like folding laundry, doing the dishes.

1:35:11 I'm not saying that's what I would want in my home,

1:35:13 but I think factory work is going to completely change.

1:35:15 I think a lot of manual labor is going to completely change,

1:35:17 and I think we're going to be forced to do what only we can do.

1:35:20 Um, Sebastian, who's the CEO of Cler, has actually just called me.

1:35:29 Hello, Sebastian.

1:35:30 You're right.

1:35:30 Hey, how are you?

1:35:31 I'm good.

1:35:31 How are you?

1:35:33 It's been a while.

1:35:34 It has been a while since you're on the show.

1:35:36 I was just saying we do need to get you back on.

1:35:38 I I just I just had a couple of simple questions cuz

1:35:40 you know I do a lot of interviews and um Clan has

1:35:43 always mentioned because I think the media has said that you like

1:35:46 double down on AI then you reversed because it didn't work out.

1:35:48 So I know I spoke to you a while ago and we exchanged a couple

1:35:51 of DMs about it but that was more than a it was almost a year ago now.

1:35:54 So I just wanted to get an update on Cler's

1:35:57 business AI agents and all of that if possible.

1:35:59 First and foremost, we were early on uh released um AI uh

1:36:04 to support our customer service which had that uh initial uh benefit

1:36:08 of uh more calls being dealt with by AI which customers liked

1:36:12 because those calls or chat messages were much much faster and more qualitative.

1:36:17 Then since then that has actually expanded slightly.

1:36:20 Um what we did however try to communicate as well is that we believed in a world

1:36:25 of where AI is cheap and available the value

1:36:29 of human interaction will be regarded as higher.

1:36:32 So the future of customer service VIP is

1:36:35 a human um we have then hence doubled down

1:36:38 on providing more of that but at the same

1:36:41 time the efficiency gains within the company has continued.

1:36:44 I mean we used to be about 6,000 people and and now we are less than 3,000 which

1:36:51 is 2 3 years since we stopped recruiting

1:36:54 and at same point in time our revenue has doubled right

1:36:56 so you can clearly see that AI has allowed

1:36:59 us to be do more with less people but we

1:37:02 have avoided layoffs and instead relied on natural attrition

1:37:07 when people kind of move on to other jobs.

1:37:09 I mean from my perspective we will continue

1:37:12 to be very you know not really recruit much.

1:37:15 I mean we recruit a little bit here and there but we expect that kind

1:37:19 of natural attrition of 10 15% per year to continue and to become fewer.

1:37:25 I think the big breakthrough was really

1:37:27 in November December last year where even the kind

1:37:30 of more most skeptical uh engineers who were

1:37:35 like very well-renowned and and appreciated like the founder

1:37:38 of Linux and stuff like that basically said

1:37:40 that coding has now been resolved and hence is

1:37:44 not you know uh you don't need to code

1:37:46 anymore and that was kind of a common sentiment.

1:37:48 So I think in in coding that's definitely an engineering work

1:37:51 that has been a tremendous shift in the last six months.

1:37:56 What do all these people go do Sebastian?

1:37:59 I am optimistic.

1:38:00 I mean I think obviously people will have a lot of opinions about this topic

1:38:03 but I still believe that we are going to move towards a richer society.

1:38:09 Now in the short term there could be more worry about

1:38:13 what happens if people don't get a job and and so forth.

1:38:15 But I think in the longer term,

1:38:17 I I am optimistic what it means for society and humanity.

1:38:21 Thank you so much, Seb.

1:38:22 I'll chat to you soon.

1:38:23 Thank you for taking the time.

1:38:24 I appreciate you, mate.

1:38:25 Thanks.

1:38:25 All right.

1:38:26 All right.

1:38:26 Byebye.

1:38:27 Byebye.

1:38:29 You know the little traditional SIM card that goes inside of our phones.

1:38:31 They haven't changed at all since they were invented in the '90s.

1:38:35 You have this physical piece of plastic

1:38:37 that means you're locked into one carrier, one network,

1:38:40 and the second you cross a border,

1:38:42 that carrier can start charging you whatever they want.

1:38:45 But there are alternatives and today's sponsor SY is one of them.

1:38:48 It's an eSIM app that gives you a safe

1:38:51 and secure data connection in over 200 destinations.

1:38:54 All of their ESIMs have built-in cyber security which is

1:38:57 great if you're traveling for work and looking at confidential material.

1:39:00 I've been using SY whenever I travel because the connection is

1:39:03 always reliable and it saves me a ton of roaming fees.

1:39:06 It also means I don't have to deal with all

1:39:08 of the faf that surrounds sorting out a SIM everywhere I go.

1:39:11 If you want to give it a try,

1:39:12 download the sale app from the app store now and scan the QR code on screen.

1:39:16 And if you want 15% off your first purchase,

1:39:19 use my code D O A when you get to check out.

1:39:24 That's D O A for 15% off.

1:39:27 Keep that to yourself.

1:39:29 This is something that I've made for you.

1:39:31 I've realized that the Dio audience are strivals that we want to accomplish.

1:39:37 And one of the things I've learned is that when you aim at the big big big goal,

1:39:42 it can feel incredibly psychologically uncomfortable because it's kind of like

1:39:47 being stood at the foot of Mount Everest and looking upwards.

1:39:50 The way to accomplish your goals is by breaking them down into tiny small steps.

1:39:55 And we call this in our team the 1%.

1:39:57 And actually this philosophy is highly responsible for much of our success here.

1:40:01 So, what we've done so that you at home

1:40:03 can accomplish any big goal that you have is we've made these 1% diaries and we

1:40:08 released these last year and they all sold out.

1:40:11 So, I asked my team over and over again to bring the diaries back,

1:40:14 but also to introduce some new colors

1:40:15 and to make some minor tweaks to the diary.

1:40:17 So, now we have a better range for you.

1:40:22 So, if you have a big goal in mind

1:40:23 and you need a framework and a process and some motivation,

1:40:27 then I highly recommend you get one

1:40:28 of these diaries before they all sell out once again.

1:40:31 And you can get yours at the diary.com.

1:40:34 And if you want the link, the link is in the description below.

1:40:38 Any thoughts?

1:40:39 Well, I actually had thoughts on something that you said before he called,

1:40:44 which is you were saying that the Jenzers like

1:40:48 there's this trend that they're actually disconnecting from technology.

1:40:50 So, they're becoming more in person.

1:40:52 And then there's this other class of workers

1:40:54 that are actually leaning into the technology,

1:40:55 but then becoming more human because they're leaning into the technology

1:41:00 because they're realizing that they should actually just be spending

1:41:02 more time doing inerson interactions rather than staring at a spreadsheet.

1:41:07 And so they're no longer doing the typing, whatever.

1:41:09 I really want to go back to this New York Magazine piece that just came out

1:41:13 because what you're describing is true for a very specific category of people,

1:41:18 which is often like the business owners

1:41:20 and leadership within companies that actually can make

1:41:23 these decisions on how they spend their time

1:41:26 and what they ultimately do with their time.

1:41:29 But what the piece talks about is the working class like people

1:41:35 like people who are not business owners that are then having to experience

1:41:41 being laid off and then working for the data annotation industry which

1:41:46 is now one of the top jobs on LinkedIn by the way.

1:41:49 Um the yeah so LinkedIn had a report that showed the top 10 jobs

1:41:54 with the highest growth in the last year and data annotation is on that list.

1:42:00 And for anyone that doesn't know what data annotation is.

1:42:02 Yeah.

1:42:02 So data annotation is the process of teaching these chat bots

1:42:08 or or any AI system to do what they ultimately are able to do.

1:42:13 So the fact that chat GBT can chat

1:42:15 is because there were tens of thousands or hundreds

1:42:17 of thousands of people that were literally typing

1:42:20 into a large language model and showing it.

1:42:23 This is how you're supposed to then respond

1:42:25 when a user types in a prompt like this.

1:42:28 Before they did that work, chatgbt didn't exist.

1:42:32 Like it just it would just you would prompt the model and the model

1:42:35 would generate some text that was not in dialogue with the person.

1:42:39 It would kind of generate something that was adjacently related.

1:42:42 Is this what they call reinforcement learning where

1:42:44 you kind of you give it like a

1:42:45 it's a part of the process of reinforcement learning.

1:42:47 So you do data annotation which is literally um showing lots

1:42:51 of different um you know examples of things that you want

1:42:55 the model to know and then reinforcement learning is getting the model

1:42:58 to then train on those examples iteratively in a way that then

1:43:02 gives the model some of those capabilities.

1:43:05 And what the New York Magazine piece highlighted is many many of the people

1:43:11 that are getting laid off now or or or are struggling to find work.

1:43:14 And these are highly educated people.

1:43:16 They're college graduates, PhD graduates, law degree graduates, doctors,

1:43:22 um and again like award-winning directors

1:43:25 that are that are then struggling to find

1:43:28 employment in the economy because the economy

1:43:30 has been very much restructured by AI.

1:43:33 they are then finding themselves being serving this industry

1:43:37 and the industry is designed in a way that is

1:43:40 extremely inhumane because what the companies the companies

1:43:45 that use these data annotation services like there's these third

1:43:48 party providers that are data annotation firms an open

1:43:52 AI a gro um a Google they will hire

1:43:56 these firms to then find the workers to perform

1:43:59 the data annotation tasks that they need for these These firms,

1:44:03 these third party firms,

1:44:04 they are incentivized to pit workers against each other because they want

1:44:10 this data annotation to happen at speed and as cheaply as possible so

1:44:13 that they can also compete with one another in this middle layer

1:44:17 to get the the the bid the the contract from the the client.

1:44:22 And so all of these workers that were interviewed for this New York Magazine

1:44:27 story talk about how they actually no longer have an ability to be human

1:44:32 because they are waiting at their laptop to be pinged on Slack for when

1:44:37 a project is going to open up

1:44:39 for data annotation because they've tried job hunting.

1:44:42 They literally can't find anything else.

1:44:43 This is the thing that's going to help

1:44:45 them put food on the table for their kids.

1:44:46 And there was this one woman who said like,

1:44:49 "I have so much anxiety about when the project is going to come,

1:44:53 when it's going to leave that when the project came,

1:44:56 it was right when my kid was coming off of off

1:44:59 of school." And I just started tasking furiously because I

1:45:03 don't know what's going to go and I need to earn

1:45:04 as much money as possible in this window of opportunity.

1:45:07 So then my when my kid came home and tried to talk to me,

1:45:10 I screamed at my child for for distracting me.

1:45:15 And then she was like, "I've become a monster and I'm not even allowed

1:45:20 to go to the bathroom or take care of my kids,

1:45:23 let alone myself, because this industry that is absorbing more

1:45:28 and more of the workers that are being laid off,

1:45:32 is mechanizing my life, atomizing my work, devaluing my expertise,

1:45:39 and then harvesting it for the perpetuation of this machine that all

1:45:44 of these AI executives are saying is

1:45:46 then going to come for everyone else's jobs.

1:45:49 And so what you were saying about these this class of workers,

1:45:54 the business owners that get to become more human because there are all

1:45:59 of these AI models now doing the tasks that they don't have to do anymore.

1:46:03 It is at the cost of the vast majority

1:46:06 of people who are not business owners that are struggling

1:46:09 to find work getting absorbed into the work of then

1:46:13 providing these technologies that the business owners can use

1:46:18 and instead of becoming more human they feel like their humanity has

1:46:22 been squeezed and diminished and they have no ability to have control,

1:46:29 agency and dignity in their lives anymore.

1:46:31 I think this is a big I think this is a big question that kind of pertains

1:46:34 to this graph here which is you know

1:46:36 all of these people if we believe anthropics prediction

1:46:40 of who will be disrupted these people

1:46:42 in these industries like arts and media legal um life

1:46:46 and social sciences architecture and engineering computer and maths

1:46:50 business and finance and management and also office and admin.

1:46:55 These people if we believe this would have

1:46:57 to retrain at something else and unlike the industrial

1:46:59 revolution where you might get 10 20 years

1:47:01 to retrain because factories take a long time to build.

1:47:04 The distribution layer that AI sits on top of is the open internet.

1:47:07 So this is why chat can go and get hundreds of millions of users

1:47:11 in no time at all and become the fastest growing company of all time.

1:47:14 Um one of my fears is that this disruption

1:47:16 takes place at a speed where we can't transition.

1:47:21 And that was you know that I think you you you said that sentence

1:47:25 in the passive voice the transition would happen

1:47:28 at a speed but who is driving that speed?

1:47:33 Um it's the companies and their race with one another.

1:47:37 Yeah.

1:47:37 And so they are driving the transition to happen at a speed at which it would

1:47:42 be really hard to take care of all of the people that would be bulldozed over by

1:47:49 this is one of the crazy questions that no one can answer

1:47:51 for me when I sit with these people that are AI CEOs.

1:47:53 So I go, "So what happens to the people if this is if

1:47:55 you agree that this is going to happen at super speed?" You know,

1:47:58 I spoke to that CEO of Uber, Dar,

1:48:00 who said very similar things to what you're saying is,

1:48:02 you know, there'll be data labeling jobs, for example, for the drivers.

1:48:05 But um they can't all become data labelers.

1:48:08 And there's a question around meaning and purpose and fulfillment.

1:48:10 And that comes from losing your meaning in life.

1:48:14 I s also sit here with so many people who talk

1:48:16 about how their father lost their job in Iran or some some

1:48:20 other country and came to the United States and had to be

1:48:23 a a toilet cleaner on particular case was a doctor in Iran

1:48:27 but came to the US and was a toilet cleaner and had

1:48:29 to deal with the sense of shame that that particular person felt

1:48:32 and the lack of dignity that that caused and how that made

1:48:35 that person's self-esteem feel and the depression

1:48:36 alcoholism that transpired from that.

1:48:39 um if this happens at a large scale across society,

1:48:43 there's going to be a ton of consequences like that.

1:48:45 I mean, this is this is like the core themes of my work.

1:48:48 And the reason why I'm critical

1:48:49 of these companies is that they are creating technologies

1:48:52 in a way that creates the halves and have nots in an extreme form that we have.

1:48:59 It's it's exacerbating the inequality that we already see in the world.

1:49:03 Like the people who have things will have way more riches.

1:49:07 they'll have way more free time.

1:49:09 They'll be allowed to be more human.

1:49:11 But the people who don't have things are even being squeezed even more.

1:49:17 And it's not just from a work perspective.

1:49:21 I mean, I talk in my book also about the environmental and public health

1:49:25 crisis that these companies have created where

1:49:29 they are building these colossal supercomput facilities.

1:49:34 there and and in in comm community like communities all around

1:49:38 the world and they specifically pick some of the most vulnerable communities.

1:49:42 We're sitting in Texas right now.

1:49:44 Open AAI's largest one of its largest

1:49:47 data center projects is being built in Abalene,

1:49:49 Texas as part of the Stargate initiative which was an effort announced

1:49:53 at the beginning of Trump's second administration

1:49:55 to spend $500 billion on AI computing infrastructure.

1:50:00 This facility consumes will when it's finished will consume more

1:50:06 than a gigawatt of power which is over 20% over 20%.

1:50:12 So this is actually a little bit inaccurate now.

1:50:14 Um this was something that circulated online

1:50:16 for a while but there's updated numbers

1:50:19 just for someone that can't see cuz they're listening on Spotify or something.

1:50:22 It's a picture of the size of this facility.

1:50:25 So this is not the Abene Texas one.

1:50:28 This is a meta facility.

1:50:29 Yeah.

1:50:29 So, let's first talk about opening eyes facility in Texas.

1:50:32 That one would be the size of Central Park and it would run a million

1:50:37 computer chips and it would require the power of more than 20% of New York City.

1:50:45 Do you know one of the things which I found confusing,

1:50:47 so I'd like to like alleviate the dissonance is I thought you

1:50:50 were saying earlier that you didn't think the job disruption promises were real.

1:50:55 No, what I was saying is that when we

1:50:59 talk about what these executives predict about the future,

1:51:04 we need to understand that they are ultimately trying to influence the public

1:51:09 in a way that allows them to continue maintaining control over the technology.

1:51:13 But objectively, do you think that the job disruption that they talk about where

1:51:16 Yeah.

1:51:16 Yeah.

1:51:16 I mean I I mentioned real well I I don't want to comment specifically on like

1:51:21 this chart but it's like we've already seen in job reports that there is

1:51:25 a restructuring of the economy happening right now.

1:51:28 Yeah.

1:51:28 But but going back to like the data center.

1:51:30 So this supercomputer facility it's a meta supercomputer facility

1:51:34 is being built in Louisiana

1:51:37 and it would be four times the size of the Abene Texas

1:51:40 one and use half of the average power demand of New York City.

1:51:44 So it's one the size of Manhattan.

1:51:46 This makes it seem like almost all of Manhattan,

1:51:48 but it's it would be 1/5 the size of Manhattan.

1:51:50 When these facilities go into these communities, what happens?

1:51:55 Power utility increases, grid reliability decreases.

1:51:59 The facilities also need fresh water to generate the power

1:52:04 for powering them as well as fresh water to cool.

1:52:06 And there have been lots of documented stories of communities

1:52:10 that are already really constrained in their freshwater resource.

1:52:12 they're under a drought when a facility comes in and then there are

1:52:16 people the community is actually like

1:52:17 competing with this facility for fresh water.

1:52:20 I talk about one of those communities

1:52:21 in my book and also sometimes these facilities instead

1:52:25 of connecting to the grid they instead a a power plant pops up next to it.

1:52:31 So in Memphis Tennessee where Musk built Colossus the supercomputer for training

1:52:36 Grock he used 35 methane gas turbines to power the facility.

1:52:42 This is a working-class community, a black and brown community,

1:52:45 a rural community that was not even told

1:52:48 that they would be the hosts of this facility.

1:52:51 And they discovered it because they literally smelled what seemed

1:52:56 like a gas leak in all of their living rooms.

1:52:59 And that's when they discovered that these methane gas

1:53:02 turbines were taking away their right to clean air.

1:53:07 And this is a community that's already

1:53:09 been facing a history of environmental racism.

1:53:12 They had already had lots of struggles to access their right to clean air.

1:53:17 And now there's this huge supercomput that's landed in their midst

1:53:23 that is pumping thousands of tons of toxins into their air,

1:53:27 exacerbating the asthmatic symptoms of the children,

1:53:31 exacerbating the respiratory illnesses of other people.

1:53:35 that it's it's one of the communities that has

1:53:36 the highest rates of um lung cancer and so

1:53:42 and that supercomputers taking their jobs

1:53:45 and then they also have supercomputers taking their jobs.

1:53:47 So, so this is what I mean is like the halves

1:53:49 and have nots are fundamentally being pulled apart even further.

1:53:55 Like if you in this version of Silicon Valley's future

1:54:00 are in the misfortunate category of being a have not,

1:54:05 we are talking about you now getting a job that is way

1:54:09 worse than what you had because you might be doing data annotation

1:54:14 and you might be treated as a machine rather than as a human to extract value

1:54:18 the value of your labor for perpetuating

1:54:20 this labor automating machine that these people are building.

1:54:25 You might be competing with these facilities for freshwater resources.

1:54:29 They're also polluting your air.

1:54:31 Your bills have increased.

1:54:33 So, the affordability crisis is getting worse.

1:54:37 Like, how is that making people able to be more human?

1:54:41 What do we do about it?

1:54:45 Yes.

1:54:45 Okay.

1:54:46 So, one of the analogies that I always

1:54:47 use is AI is like the word transportation.

1:54:51 Transportation can literally refer to everything from a bicycle to a rocket.

1:54:55 And we have nuanced conversations about transportation where we always say

1:55:00 we need to transition our transportation towards more uh sustainable options.

1:55:06 We need a transition towards you know public transport, electric vehicles.

1:55:10 And we don't we don't ever say everyone should get a rocket

1:55:14 to do every to serve all of their transportation needs, right?

1:55:17 Like we're in Austin.

1:55:19 If you use a rocket to fly from Dallas to Austin,

1:55:21 like that would just make not no sense.

1:55:24 It's just a disproportionate use of resources to get

1:55:27 the benefit of getting from point A to point B.

1:55:31 This how we should think about AI.

1:55:33 So all of the models that we've been talking about,

1:55:35 I like to think of them as the rockets of AI.

1:55:39 They use an extraordinary amount of resources and they provide

1:55:41 benefit some dramatic benefit to some people but they're also

1:55:46 exacting an extraordinary cost on a large swath of people

1:55:49 because of the like the costs of developing this technology.

1:55:57 Why don't we build more bicycles of AI?

1:56:00 This is things like deep minds alpha fold which is a system

1:56:04 that predicts how proteins will fold based on amino acid sequences.

1:56:08 It's really important for accelerating drug discovery for understanding human

1:56:15 disease and it won the Nobel Prize in chemistry in 2024.

1:56:18 And the reason why it's a bicycle of AI

1:56:20 is because you're using small curated data sets.

1:56:24 you're just you just have data

1:56:26 that has amino acid sequences and protein folding.

1:56:30 So that means you need significantly

1:56:33 less computational resources to develop the system,

1:56:36 which means significantly less energy,

1:56:38 which means less emissions, so on and so forth.

1:56:40 And you're providing enormous benefit to people.

1:56:43 It feels like the horse has left the stable

1:56:47 in this regard because they've already taken people's IP,

1:56:50 they've taken media, they they train on this podcast.

1:56:53 We know they do because it it shows that they do.

1:56:55 Um I think there's a button actually in the back end of YouTube now that allows

1:56:58 you just to click it and it says we will train on your YouTube channel.

1:57:02 Um so the horses kind of left.

1:57:04 Here's the thing.

1:57:05 If the horse truly had left the stables,

1:57:07 they wouldn't have to train on anything anymore.

1:57:09 Why is it that their appetite for data has actually expanded?

1:57:14 It's because in order to build the next generations of their technologies,

1:57:17 in order to have the technologies continue to be relevant and continue

1:57:22 to update with the pace of new knowledge creation and society's evolvement,

1:57:28 they need to train again and again and again and again.

1:57:32 And why are they employing actually more

1:57:34 and more and more data annotation workers over time?

1:57:37 It's because they need more and more of that work over time.

1:57:41 I mean, I've been reporting on data annotation work for over 7 years now,

1:57:46 and it's not gone down.

1:57:48 It's gone it's increased.

1:57:50 Do you think there's any chance of it going down?

1:57:53 Do you think there's any chance of this sort

1:57:54 of brute force scaling approach where you take data,

1:57:57 you take computational power,

1:58:00 energy, and you, you know, you have um the data labelers and, you know,

1:58:05 building out more and more parameters for the models.

1:58:07 Do you think there's any chance it's going to stop or go

1:58:10 in a different direction other than the one it's going in now?

1:58:12 I would love to reframe the question and say

1:58:15 what should we be doing in this moment where

1:58:17 it's not going down where we do recognize

1:58:20 that actually these companies in this moment need continued resources,

1:58:24 inputs and labor to perpetuate what they are doing.

1:58:28 Yeah.

1:58:28 because this sounds like stop and I just feel like stop is like a HUD.

1:58:33 It feels like I just think you know

1:58:34 with the government in place they're supporting these companies like crazy.

1:58:37 Globally this is happening.

1:58:39 So I'm like stop doesn't feel I always say we need to break up the empire and we

1:58:43 need to develop alternatives and we are already seeing a flourishing

1:58:48 of incredible grassroots movements that are

1:58:51 applying an enormous amount of pressure

1:58:53 to the way that the empire is trying to unfold its agenda.

1:58:58 80% of Americans in the most recent poll

1:59:00 think that the AI industry need to be regulated.

1:59:04 Yeah.

1:59:04 When was the last time that 80% of Americans were on the same side of an issue?

1:59:07 No.

1:59:07 Yeah.

1:59:08 When I have these conversations on the podcast, the comment section are clear.

1:59:11 Yeah.

1:59:11 There's no there's no disagreement.

1:59:12 There's no one in there going, "Oh, no.

1:59:13 I think they should crack on." Yeah.

1:59:15 Dozens dozens of protests against data centers have

1:59:18 broken out all around this country and the US, all around the world.

1:59:22 So, what do we do about it?

1:59:23 So, these are thing people that are doing something about it.

1:59:27 They are actually reasserting

1:59:29 their agency and exercising democratic contestation

1:59:33 against the ways that the empires are going about their business.

1:59:36 What goal should we be aiming at?

1:59:38 So, if I said to my audience, Janet at home,

1:59:40 because this is kind of what I see in the comments, it's hopelessness.

1:59:42 It's like, what can I do?

1:59:43 I'm just a Yeah.

1:59:44 Well, well, well, the goal is not that we completely get rid of this technology.

1:59:49 The goal is that these companies need to stop being empires.

1:59:52 And the way I define like a typical

1:59:53 business versus an empire is that the empires

1:59:56 are predicated on this idea that they do

1:59:58 not have to provide a fair exchange of value

2:00:01 with the workers who work for them or the people who use them or all

2:00:04 of the other people that are involved in like

2:00:06 the supply chain of producing and deploying these technologies.

2:00:09 They can extract and exploit and extract and exploit

2:00:11 and get more value than what they offer.

2:00:14 Whereas typical businesses, there's a fair exchange.

2:00:17 you you buy a service,

2:00:18 you feel like you got the same amount of value as the service that you provided.

2:00:22 But like for these data annotation workers, for example,

2:00:24 they do not feel in any way that they're being

2:00:26 paid the same value that they provide to these companies.

2:00:29 So that's like for me the north star is like we should be

2:00:33 pushing back and holding accountable these companies

2:00:37 when they operate in an imperial way.

2:00:40 And that's what we've seen with all of these people that are now

2:00:43 literally protesting in the streets against

2:00:44 data centers and having an enormous effect,

2:00:47 by the way, actually stalling data center projects and also

2:00:51 completely banning data centers from being developed in their localities.

2:00:54 We're seeing that with artisan writers that are suing

2:00:57 these companies for intellectual property infringement and creating a huge

2:01:01 public conversation about what is it that we actually

2:01:05 how do we actually want to protect our intellectual property?

2:01:08 It's like I three weeks ago I met Megan Garcia who is the mother of Sul Settzer

2:01:14 III who is the 14-year-old who died by suicide

2:01:19 after being sexually groomed by a characterized chatbot.

2:01:23 And she when that happened I mean obviously was

2:01:29 incredibly devastated by what had happened to her son.

2:01:33 She also decided to do something about it.

2:01:35 She sued the companies and that lawsuit then sparked many other parents

2:01:40 and families who were actually experiencing similar

2:01:42 things to sue these companies as well.

2:01:45 That has created an enormous public conversation about what

2:01:50 these companies are actually doing when they exploit and they extract.

2:01:54 What is the cost to the lives of people around the world including children?

2:02:00 So, what do you think my audience should do if

2:02:02 they if they agree with everything written in your book,

2:02:04 Age Empire of AI, Dreams and Nightmares, and Sam Mortman's Open AI?

2:02:09 If they agree with everything said here,

2:02:11 if they agree with everything we've discussed today,

2:02:13 they're concerned about their kids,

2:02:14 they they don't want everyone to become data labelers,

2:02:17 they don't think that's a, you know,

2:02:18 particularly great solution, what what can they actually go and do?

2:02:22 When I was writing the book, the only discourse that was happening was

2:02:25 this is the best thing since sliced bread.

2:02:28 Mhm.

2:02:28 because of all of the actions of these people like saying when they're

2:02:32 comp they're they're not happy with the things that these companies are doing.

2:02:37 We now have 80% of Americans that want to regulate this industry.

2:02:40 And so I would say to people,

2:02:42 think about all of the ways that your life intersects

2:02:45 with the resources and the that the AI industry needs

2:02:49 to perpetuate what they do and also the spaces that they

2:02:52 would need to deploy these technologies to continue having broad-based adoption

2:02:58 in their work.

2:03:00 So you're a data donor to these companies.

2:03:04 You could withhold that data.

2:03:06 And that's what those artists and writers are are doing.

2:03:08 like they're suing these companies to withhold to try

2:03:10 and create mechanisms by which that data would then be withheld.

2:03:14 You probably have a data center popping up around you.

2:03:17 If you're at a school environment or a company environment,

2:03:21 you're probably having a discussion in those environments right

2:03:23 now about what should the AI adoption policy be?

2:03:27 And these companies they like I was talking with some open air

2:03:31 employees just the other day and they were telling me that it's

2:03:35 understood internally that the revenue targets for the company are extraordinary

2:03:42 and they need things to go flawlessly for it to all work out.

2:03:48 And so they would need every single person

2:03:51 to adopt this, every single space to adopt this.

2:03:54 They would need to be able to build their data

2:03:56 centers at the speed that they're trying to build them.

2:03:59 And so what I would say to everyone of your viewers is let's

2:04:02 not make it go flawlessly if we don't agree with what they are doing.

2:04:06 Ah, okay.

2:04:06 I got you.

2:04:08 And then let's build alternatives.

2:04:09 Because the thing is what I'm saying

2:04:13 is not that these technologies don't have utility.

2:04:16 It's that specifically the political economy that has emerged

2:04:19 to support the production of these technologies right now

2:04:23 is exacting a lot of harm on people.

2:04:25 But we have research that shows that the very same capabilities could

2:04:30 be developed with much more efficient

2:04:33 methods with much less resource consumption.

2:04:36 And we have a lot of different other AI systems at our disposal that are

2:04:40 like the bicycles of AI that we also

2:04:42 know provide extraordinary benefit at very little cost.

2:04:46 So let's break up the empire and let's forge new

2:04:49 paths of AI development that are broadly beneficial to everyone.

2:04:53 It's strange.

2:04:54 I'm quite I think I'm I'm I've trained

2:04:57 myself to deal with dichotoies in my head.

2:05:00 And this for me is such is a dichotomy where I as a CEO and as a founder,

2:05:04 as an entrepreneur and someone that loves technology, I think it's incredible.

2:05:08 It's absolutely incredible AI.

2:05:09 It's just so amazing and incredible the things it's enabled me to do and create.

2:05:13 Yeah.

2:05:13 Because it's designed to enable people like you.

2:05:16 And my car driving in the morning and being safer.

2:05:20 Incredible.

2:05:21 Um I think you know the billion odd people

2:05:24 that use AI tools or chat or whatever it might be,

2:05:26 they'd probably say that it's added value to their life.

2:05:29 But and this is the part that people find confusing that you can and I like I

2:05:32 invest in companies that are you know heavily using

2:05:34 AI but and the big butt is is it possible to think that is true and also

2:05:39 think that there are significant unintended consequences which technology

2:05:44 in the history of technology should have taught us

2:05:45 to take a moment to pause to talk about because

2:05:48 I think this is absolutely like you can have both of these things in your head

2:05:53 and what I'm saying is that this tension doesn't have to be a tension

2:05:58 because we could actually preserve the utility

2:06:01 and benefits of these technologies but actually develop

2:06:04 and design them in a different way

2:06:06 that doesn't have all of these unintended consequences.

2:06:09 Yes.

2:06:09 And I think there needs to be a big social conversation which is why

2:06:11 I have so many conversations about AI in the show like there needs to be

2:06:14 a big social conse uh conversation about

2:06:16 being intentional about the social impact um

2:06:20 the social and environmental impact and that conversation

2:06:22 is not being had in the in government.

2:06:25 From what I can see,

2:06:26 the conversation takes place in the industry and actually trying

2:06:29 to pull it out of the industry and and open

2:06:31 people's minds to it is hopefully what we've been doing

2:06:34 over the last couple of months with this subject because

2:06:35 I think it's actually been it it has

2:06:38 been been happening everywhere outside of the industry

2:06:41 and for local governments and state level governments

2:06:44 there have been huge conversations about this everywhere.

2:06:47 Like I've been on book tour, I've been to dozens of cities around the world.

2:06:51 People are having these crucial conversations everywhere.

2:06:56 I have not gone to a single city.

2:06:57 Yes.

2:06:57 Everywhere.

2:06:58 Even here in South by.

2:06:59 Yeah.

2:06:59 I haven't gone to a single city where the room is not packed and people are not

2:07:03 wrestling with the same exact questions as every other

2:07:06 person in every other room that I've been in.

2:07:08 Speaking of packed rooms,

2:07:08 I know you've got to go cuz you've got you've got to talk today.

2:07:11 So, I'm going to we've got a last

2:07:13 question which is the closing tradition on this podcast.

2:07:15 How would your advice to a friend with a terminal

2:07:17 diagnosis differ from what you would do yourself?

2:07:23 That's a great question.

2:07:24 Differ from what you would do yourself?

2:07:26 Oh my god.

2:07:26 I have I I would tell them like enjoy like live life for yourself.

2:07:33 Um you wouldn't do it and take it easy.

2:07:35 And yeah, I I I am not taking it easy.

2:07:39 Well, I think it's a good thing you're not taking

2:07:40 it easy because you're leading a conversation which is incredibly important.

2:07:43 And I think that's the thing.

2:07:45 I think the conversation is the important thing.

2:07:47 And so, you know, because of algorithms and echo chambers,

2:07:50 it's so rare to have a conversation these days, especially a long form one.

2:07:54 I agree.

2:07:55 Like this.

2:07:55 So, I think they're so important.

2:07:56 And your book is for anyone that's curious about

2:08:00 I think a lot of people would have learned a lot of stuff today cuz I

2:08:02 sit here with and interview AI people all

2:08:04 the time and I've learned so much today.

2:08:06 From reading your book and the extensive

2:08:08 objective perspective that your book takes,

2:08:11 you you're able to unravel all of these stories

2:08:13 that we sometimes see in tweets and we don't know if

2:08:15 they're true or not because you've gone and met the people

2:08:17 and you've done your research and you're incredibly intelligent person,

2:08:20 extremely intelligent person who clearly has

2:08:23 humanity's interests as your north star

2:08:26 and that shows up in everything you do and everything you say.

2:08:28 So please continue to fight in the way

2:08:30 that you are um because it's an incredibly important one.

2:08:33 people like you that are, I think, galvanizing the world to take the collective

2:08:39 action that we're starting to see everywhere.

2:08:43 Yeah.

2:08:43 Empire of AI: Dreams and Nightmares in Sam Alman's Open AI by Karen How.

2:08:47 I'll link it below for anyone that wants to read this book.

2:08:49 I highly recommend you do.

2:08:50 It's a New York Times bestseller for good reason.

2:08:52 Karen, thank you.

2:08:53 Thank you so much, Stephen.

2:08:55 YouTube have this new crazy algorithm where they know exactly what video you

2:08:58 would like to watch next based on AI and all of your viewing behavior.

2:09:02 And the algorithm says that this video is the perfect video for you.

2:09:07 It's different for everybody looking right now.