WEBVTT

1
00:00:00.000 --> 00:00:04.480
So much of what's happening today in the AI industry is extremely inhumane.

2
00:00:04.480 --> 00:00:05.840
But this is me playing devil's advocate.

3
00:00:05.840 --> 00:00:10.880
And logically, it could be the case that the civilization that accelerate the research with AI

4
00:00:10.880 --> 00:00:13.120
is going to be the superior civilization.

5
00:00:13.120 --> 00:00:13.840
No, it's not.

6
00:00:13.840 --> 00:00:15.280
This is a prediction that you're making, right?

7
00:00:15.280 --> 00:00:18.240
Animals making, Zuckerberg's making, or women's making.

8
00:00:18.240 --> 00:00:20.000
And do you know what the common future of all of this?

9
00:00:20.000 --> 00:00:22.320
They profit enormously off of this myth.

10
00:00:22.320 --> 00:00:28.160
You know, I have all these internal documents showing that they're purposely trying to create that feeling within the public

11
00:00:28.160 --> 00:00:30.960
so that they can extract and exploit and extract and exploit.

12
00:00:31.760 --> 00:00:32.800
So what do we do about it?

13
00:00:32.800 --> 00:00:34.400
We need to break up the empires of AI.

14
00:00:35.120 --> 00:00:38.160
You know, I've been covering the tech industry for over eight years.

15
00:00:38.160 --> 00:00:42.640
Interviewed over 250 people, including former or current opening eye employees and executives.

16
00:00:42.640 --> 00:00:47.440
And I can tell you that there are many parallels between the empires of AI and the empires of old, right?

17
00:00:47.440 --> 00:00:50.400
Like lay claim to the intellectual property of artist writers and creators

18
00:00:50.400 --> 00:00:52.000
in the pursuit of training these models.

19
00:00:52.000 --> 00:00:56.000
Second, they exploit an extraordinary amount of labor, which breaks the career ladder

20
00:00:56.000 --> 00:00:59.680
because someone gets laid off and then they work to train the models

21
00:00:59.680 --> 00:01:02.240
on the very job that they were just laid off in,

22
00:01:02.240 --> 00:01:05.680
which will then perpetuate more layoffs if that model then develops that skill.

23
00:01:05.680 --> 00:01:09.760
And when they talk about that there's going to be some new jobs created that we can't even imagine,

24
00:01:09.760 --> 00:01:14.400
a lot of the jobs that are created are way worse than the jobs that were there.

25
00:01:14.400 --> 00:01:18.800
And then there's the environmental and public health crisis that these companies have created.

26
00:01:18.800 --> 00:01:21.760
And how they're able to also spend hundreds of millions

27
00:01:22.000 --> 00:01:25.040
to try and kill every possible piece of legislation that gets in their way

28
00:01:25.040 --> 00:01:29.600
and will censor researchers that are inconvenient to the empires agenda.

29
00:01:29.600 --> 00:01:33.680
But what I'm saying is not that these technologies don't have utility.

30
00:01:33.680 --> 00:01:36.240
It's that the production of these technologies right now

31
00:01:36.240 --> 00:01:38.240
is exacting a lot of harm on people.

32
00:01:38.240 --> 00:01:44.080
But we have research that shows that the very same capabilities could be developed

33
00:01:44.080 --> 00:01:48.000
in a different way that doesn't have all of these unintended consequences.

34
00:01:48.000 --> 00:01:49.360
So let's talk about all of that.

35
00:01:52.720 --> 00:01:55.440
Guys, I've got a favor to ask before this episode begins.

36
00:01:55.440 --> 00:02:00.240
The algorithm, if you follow a show, will deliver you the best episodes from that show

37
00:02:00.240 --> 00:02:01.760
very prominently in your feed.

38
00:02:01.760 --> 00:02:03.840
So when we have our best episodes on this show,

39
00:02:03.840 --> 00:02:06.000
the most shared episodes, the most rated episodes,

40
00:02:06.000 --> 00:02:07.120
I would love you to know.

41
00:02:07.120 --> 00:02:10.160
And the simple way for you to know that is to hit that follow button.

42
00:02:10.160 --> 00:02:14.720
But also, it's the simple, easy, free thing that you can do to help us make this show better.

43
00:02:14.720 --> 00:02:18.960
I would be hugely grateful if you could take a minute on the app you're listening to this on right now

44
00:02:18.960 --> 00:02:20.160
and hit that follow button.

45
00:02:20.160 --> 00:02:22.960
Thank you so, so, so much.

46
00:02:27.600 --> 00:02:28.800
Karen, how?

47
00:02:29.680 --> 00:02:33.040
You've written this book in front of me here called Empire of AI,

48
00:02:33.040 --> 00:02:36.080
Dreams and Nightmares in Sam Altman's Open AI.

49
00:02:36.080 --> 00:02:40.960
I guess my first question is what is the research and the journey you went on

50
00:02:40.960 --> 00:02:44.320
in order to write this book we're going to talk about and the subjects within it today?

51
00:02:44.320 --> 00:02:46.880
I took a strange route into journalism.

52
00:02:46.880 --> 00:02:48.640
I studied mechanical engineering at MIT.

53
00:02:49.280 --> 00:02:52.080
And so when I graduated, I moved to San Francisco.

54
00:02:52.080 --> 00:02:53.280
I joined a tech startup.

55
00:02:53.280 --> 00:02:54.720
I became part of Silicon Valley.

56
00:02:55.520 --> 00:02:59.360
And I basically received an education in what Silicon Valley is about.

57
00:02:59.360 --> 00:03:03.440
Because a few months into joining a very mission driven startup that was focused on building

58
00:03:03.440 --> 00:03:07.120
technologies that would help facilitate the fight against climate change,

59
00:03:07.680 --> 00:03:11.120
the board fire the CEO because the company was not profitable.

60
00:03:11.120 --> 00:03:18.160
And this was in hindsight a very pivotal moment for me because I thought if this hub is ultimately

61
00:03:18.800 --> 00:03:24.480
geared towards building profitable technologies and many of the problems in the world

62
00:03:24.480 --> 00:03:28.240
that I think need solved are not profitable problems like climate change,

63
00:03:29.120 --> 00:03:31.600
then what are we actually doing here?

64
00:03:31.600 --> 00:03:35.440
Like, how did we get to a point where innovation is not actually

65
00:03:36.080 --> 00:03:40.160
necessarily working in the public benefit and sometimes even undermining the public benefit

66
00:03:40.160 --> 00:03:41.440
in pursuit of profit?

67
00:03:41.440 --> 00:03:47.600
In that moment, I had a bit of a crisis where I thought, well, I just spent four years

68
00:03:48.480 --> 00:03:55.200
trying to set myself up for this career that I now don't think I am cut out for.

69
00:03:55.200 --> 00:04:00.000
And I thought, well, I might as well just try something totally different.

70
00:04:00.000 --> 00:04:00.960
I've always liked writing.

71
00:04:01.600 --> 00:04:08.080
And that's how after two years, I landed at a role at MIT Technology Review covering AI

72
00:04:08.080 --> 00:04:14.080
full-time. And that gave me a space to then explore all of these questions of who gets to decide

73
00:04:14.080 --> 00:04:18.960
what technologies we built, how does money and ideology also drive the production of those

74
00:04:18.960 --> 00:04:25.040
technologies, and how do we ultimately make sure that we actually reimagine the innovation

75
00:04:25.040 --> 00:04:29.840
ecosystem to work for a broad base of people all around the world?

76
00:04:30.560 --> 00:04:36.320
And so that is kind of how I then set off on this journey of ultimately writing a book.

77
00:04:36.400 --> 00:04:42.560
I didn't realize that I was working towards writing a book, but starting in 2018 when I took

78
00:04:42.560 --> 00:04:48.720
that job was essentially the moment in which I began researching the story that I documented it.

79
00:04:49.280 --> 00:04:52.880
Very timely time to start working in artificial intelligence. For anyone that doesn't know,

80
00:04:52.880 --> 00:04:58.960
this is pre-open AI, chat, GPT launch moment that shook the world.

81
00:04:59.680 --> 00:05:02.720
But in writing this book, you interviewed a lot of people and went to a lot of places. Can you

82
00:05:02.720 --> 00:05:06.480
give me a flavor of how many people you've interviewed, where it's taken you around the world,

83
00:05:06.480 --> 00:05:12.080
et cetera? I interviewed over 250 people, so over 300 interviews, over 90 of those people

84
00:05:12.880 --> 00:05:19.120
were former or current open AI employees and executives. So the book covers the inside story

85
00:05:19.120 --> 00:05:25.520
of open AI's first decade and how it ultimately got to where it is today. But I didn't want to

86
00:05:25.520 --> 00:05:31.920
write a corporate book. I felt very strongly that in order to help people understand the impact

87
00:05:31.920 --> 00:05:36.960
of the AI industry, we would also have to travel well beyond Silicon Valley. These companies tell

88
00:05:36.960 --> 00:05:43.040
us that AI is going to benefit everyone and that's their mission. But you really start to see

89
00:05:43.040 --> 00:05:48.720
that rhetoric break down when you go to the places that look nothing like Silicon Valley,

90
00:05:48.720 --> 00:05:53.040
that speak nothing like Silicon Valley, and that have a history and culture that are fundamentally

91
00:05:53.040 --> 00:05:59.840
different as well. And that's where you start to really understand the true reality of how this

92
00:05:59.840 --> 00:06:06.640
industry is unfolding around us. Karen, I often try and steer conversations. But in this situation,

93
00:06:06.640 --> 00:06:11.920
I feel like it's probably my responsibility to follow. So with that in mind, I'm going to ask you,

94
00:06:11.920 --> 00:06:15.600
where does this journey begin and where should we be starting if we're talking about the subjects

95
00:06:15.600 --> 00:06:21.600
of Empire of AI, AI generally, artificial intelligence, and also I would say one thing I'm really

96
00:06:21.600 --> 00:06:26.640
keen to do in this conversation, which I often see in conversations is left out, is let's assume

97
00:06:26.720 --> 00:06:31.360
that our viewers know nothing about AI. So they don't know what scaling laws are or GPUs or

98
00:06:31.360 --> 00:06:36.720
compute or whatever. And let's try and keep this as simple as we possibly can in terms of language

99
00:06:36.720 --> 00:06:41.200
or explain all the complicated language so that we can bring as much people with us as we

100
00:06:41.200 --> 00:06:48.480
possibly can. Where should we start? I think we should start with when AI started as a field.

101
00:06:48.480 --> 00:06:54.880
So this was back in 1956 and there were a group of scientists that gathered at Dartmouth University

102
00:06:55.680 --> 00:07:01.600
to start a new discipline, a scientific discipline to try and chase an ambition. And specifically,

103
00:07:01.600 --> 00:07:06.080
an assistant professor at Dartmouth University, John McCarthy, decided to name this discipline

104
00:07:06.080 --> 00:07:12.000
artificial intelligence. This was not the first name that he tried the previous year. He tried

105
00:07:12.000 --> 00:07:17.440
to name it Atomata studies. And the reason why some of his colleagues were concerned about this name

106
00:07:17.440 --> 00:07:23.440
was because it pegged the idea of this discipline to recreating human intelligence.

107
00:07:24.240 --> 00:07:31.920
And back then, as is true today, we have no scientific consensus around what human intelligence is.

108
00:07:31.920 --> 00:07:39.120
There's no definition from psychology, biology, neurology. And in fact, every attempt in history

109
00:07:39.120 --> 00:07:46.720
to quantify and rank human intelligence has been driven by nefarious motives. It's been driven

110
00:07:46.720 --> 00:07:53.760
by a desire to prove scientifically that certain groups of people are inferior to other groups of

111
00:07:53.760 --> 00:08:00.080
people. There are no goal posts for this field. And there are no goal posts for the industry when

112
00:08:00.080 --> 00:08:06.400
they say that they are ultimately trying to recreate AI systems that would be as smart as humans.

113
00:08:06.400 --> 00:08:12.240
How do we even define what that means? And when are we going to get there if we don't know how to

114
00:08:12.240 --> 00:08:19.200
define the destination? And what that effectively means is that these companies can just use the

115
00:08:19.200 --> 00:08:24.080
term artificial general intelligence, which is now the term to refer to this ambitious

116
00:08:25.120 --> 00:08:30.560
goal to recreate human intelligence. They can use it however they want to. And they can define

117
00:08:30.560 --> 00:08:35.440
and redefine it based on what is convenient for them. So in open AI's history, it has defined

118
00:08:35.440 --> 00:08:40.800
and redefined it many times. When Sam Altman is talking with Congress, AGI is a system that's

119
00:08:40.800 --> 00:08:46.800
going to cure cancer, solve climate change, cure poverty. When he's talking with consumers,

120
00:08:46.800 --> 00:08:51.760
that he's trying to sell his products to, it's the most amazing digital assistant that you're

121
00:08:51.760 --> 00:08:57.360
ever going to have. When he was talking with Microsoft, you know, in the deal that Open AI

122
00:08:57.360 --> 00:09:03.360
and Microsoft struck where Microsoft invested in the company, it was defined as a system that will

123
00:09:03.360 --> 00:09:09.120
generate $100 billion of revenue. And on Open AI's own website, they define it as highly

124
00:09:09.120 --> 00:09:16.560
autonomous systems that outperform humans in most economically valuable work. This is like not a

125
00:09:16.560 --> 00:09:22.800
coherent vision of one technology. These are very different definitions that are spoken out loud

126
00:09:22.800 --> 00:09:30.160
to the audience that needs to be mobilized to ward off regulation or get more consumer buy-in

127
00:09:30.160 --> 00:09:38.400
into the industry's quest or to get more capital, more resources for continuing on this journey

128
00:09:38.400 --> 00:09:45.120
with ambiguous definitions. I mean, speaking about different definitions through time. In 2015,

129
00:09:45.120 --> 00:09:50.640
in a blog post that Sam Altman wrote before Open AI was officially announced, he explicitly outlined

130
00:09:50.640 --> 00:09:56.160
the existential risk by saying, development of superhuman machine intelligence is probably the

131
00:09:56.160 --> 00:10:00.400
greatest threat to the continued existence of humanity. There are other threats that I think are

132
00:10:00.400 --> 00:10:06.800
more certain to happen, for example, an engineered virus, but AIs probably the most likely way to

133
00:10:06.800 --> 00:10:13.760
destroy everything. In general, when Altman is writing for the public or speaking for the public,

134
00:10:13.760 --> 00:10:19.520
he does not just have the public as the audience in mind. There are other people that he is trying

135
00:10:19.520 --> 00:10:26.560
to motivate or mobilize when he says these things. And in that particular moment, Altman was trying

136
00:10:26.560 --> 00:10:34.160
to convince Elon Musk to join him on co-founding Open AI. And Musk, in particular, was spending all

137
00:10:34.160 --> 00:10:41.120
of his time sounding the alarm on what he saw as a huge existential threat that AI could pose.

138
00:10:41.760 --> 00:10:47.040
And so, in that blog post, if you look at the language that Altman uses, side-by-side with the

139
00:10:47.040 --> 00:10:52.080
language that Musk was using at the time, it mirrors all the things that Musk was saying.

140
00:10:52.080 --> 00:10:57.440
It's identical. Ten years ago, Musk was going on podcast, saying, tweeting, whatever that,

141
00:10:57.440 --> 00:10:59.920
the greatest existential risk to humanity was AI.

142
00:11:00.000 --> 00:11:06.000
Yeah. And so, you know, like his parenthetical, there are other things that might actually be

143
00:11:06.000 --> 00:11:11.680
more likely to happen like engineered viruses. It's because up until then, Altman had been talking

144
00:11:11.680 --> 00:11:18.080
just about engineered viruses. And so now that he needs a pivot to speak to an audience of one,

145
00:11:18.080 --> 00:11:24.560
to Musk, he needs to kind of resolve the contradiction between what he's now elevating as his new

146
00:11:24.560 --> 00:11:30.480
central fear to be the same as Musk's new central fear with what he had previously been saying.

147
00:11:30.480 --> 00:11:34.720
So that's why he's like, I think this is now even though before I said this.

148
00:11:35.920 --> 00:11:42.080
And are you saying that Sam Altman manipulated Musk? Because Elon did end up donating a huge

149
00:11:42.080 --> 00:11:48.000
amount of money to Open AI and co-founding it, I believe, with Sam Altman. Elon most did end up

150
00:11:48.000 --> 00:11:53.600
co-founding it with Altman. And certainly from Musk's perspective, he does feel manipulated,

151
00:11:53.600 --> 00:12:02.560
because he feels like Altman was engineering his language in a way that would make Musk trust

152
00:12:02.560 --> 00:12:12.240
him as a partner in this endeavor. And of course, then Musk is leaves. And through some of the documents

153
00:12:12.240 --> 00:12:18.880
that came out during the lawsuit that Musk and Altman are engaged in now, it has become clear that

154
00:12:18.880 --> 00:12:24.640
there was a degree to which Musk was actually muscled out a little bit. And so that's why

155
00:12:24.640 --> 00:12:31.920
he's left with this very intense personal vendetta against Altman saying that somehow Altman tricked

156
00:12:31.920 --> 00:12:37.920
him into being part of this. So in 2015, Sam Altman is writing these blood posts saying this is

157
00:12:37.920 --> 00:12:43.120
one of the greatest existential threats at the same time. In 2015, Musk is doing some very famous

158
00:12:43.120 --> 00:12:48.480
speeches at the time at MIT. He said that AI was the biggest existential threat and compared

159
00:12:48.480 --> 00:12:53.600
developing AI to summoning the demon. And what you're saying here is you're saying that Sam Altman

160
00:12:53.600 --> 00:12:58.880
was just mirroring the language that Elon was using to get Elon involved in Open AI. And later,

161
00:12:58.880 --> 00:13:04.480
it appears, and again, there's a legal case taking place now, that Sam might have muscled Elon

162
00:13:04.480 --> 00:13:08.800
out in some capacity. Yeah. So we know from the lawsuit and the documents that have come out in

163
00:13:08.800 --> 00:13:15.280
the lawsuit that Ilya Satskever, who was the chief scientist of Open AI at the time and Greg

164
00:13:15.280 --> 00:13:21.840
Brockman, chief technology officer at the time, when they were deciding whether or not to maintain

165
00:13:21.840 --> 00:13:26.000
Open AI as a non-profit because it was originally found as a non-profit, they decided, okay,

166
00:13:26.000 --> 00:13:30.480
we need to create a for-profit entity. But the question was who should be the CEO of this for-profit

167
00:13:30.480 --> 00:13:34.960
entity? Should it be Musk or should it be Altman because they were the two co-chairmen of the

168
00:13:34.960 --> 00:13:44.240
non-profit? And in the emails, it became clear that Ilya and Greg first chose Musk to be the CEO.

169
00:13:45.280 --> 00:13:52.560
But through my reporting, I discovered that Altman then appealed personally to Greg Brockman,

170
00:13:52.560 --> 00:13:57.360
who was a friend of his that they'd known each other for many years through the Silicon Valley

171
00:13:58.080 --> 00:14:05.680
and said, don't you think that it would be a little bit dangerous to have Musk be the CEO of this

172
00:14:05.680 --> 00:14:11.840
company, this new for-profit entity? Because, you know, he's a famous guy. He has a lot of pressures

173
00:14:11.840 --> 00:14:18.080
in the world. He could be threatened, he could act erratically, he could be unpredictable,

174
00:14:18.080 --> 00:14:23.200
and do we really want a technology that could be super powerful in the future to end up in the

175
00:14:23.200 --> 00:14:30.560
hands of this man? And that convinced Greg and Greg then convinced Ilya, you know, I think

176
00:14:31.120 --> 00:14:37.600
there's a point here, do we really want to give this much power to Musk? And that is why Musk

177
00:14:37.600 --> 00:14:43.280
then leaves because then the two switch their allegiances, they say, actually we want Altman

178
00:14:43.280 --> 00:14:48.960
to be the CEO and then Musk is like, if I'm not CEO, I'm out. So it sounds like Sam again managed

179
00:14:49.040 --> 00:14:55.280
persuade someone to do something. I guess this begs the question, what do you think of Sam Altman?

180
00:14:57.280 --> 00:15:03.360
I think he's a very controversial figure. You did an interesting pause. It's a pause where

181
00:15:03.360 --> 00:15:12.320
someone tries to select their words. Well, this is what's so interesting about those interviews is

182
00:15:12.400 --> 00:15:20.400
people are extremely polarized on Altman. No one has in between feelings about him. Either they

183
00:15:20.400 --> 00:15:25.360
think he's the greatest tech leader of this generation akin to the chief jobs of the modern era,

184
00:15:25.920 --> 00:15:33.760
or they think that he's really manipulative and an abuser and a liar. And what I realized,

185
00:15:33.760 --> 00:15:40.160
because I interviewed so many people, is it really comes down to what that person's vision of

186
00:15:40.160 --> 00:15:47.440
the future is and what their goals are. So if you align with Altman's vision of the future,

187
00:15:47.440 --> 00:15:52.080
you're going to think he's the greatest asset ever to have on your side, because this man is

188
00:15:52.080 --> 00:15:56.320
really persuasive. He's incredible at telling stories. He's incredible at mobilizing capital,

189
00:15:56.320 --> 00:16:02.240
at recruiting talent, at getting all the inputs that you need to then make that future happen.

190
00:16:02.240 --> 00:16:09.040
But if you don't agree with his vision of the future, then you begin to feel like you're being

191
00:16:09.040 --> 00:16:15.360
manipulated by him to support his vision, even if you fundamentally don't agree with it. And

192
00:16:15.360 --> 00:16:22.320
this is the story, especially of Dario Amade, CEO of Anthropic, who was originally an executive at

193
00:16:22.320 --> 00:16:27.440
OpenAI. So for people that don't know, Dario now runs Anthropic, which is the maker of Claude.

194
00:16:27.440 --> 00:16:32.240
A lot of people probably are more familiar with Claude. Yeah. And it's one of the biggest competitors

195
00:16:32.240 --> 00:16:42.800
to OpenAI. And Amade at the time when he was an executive at OpenAI, he thought that Altman was

196
00:16:42.800 --> 00:16:49.200
on the same page with him. And then over time began to feel that Altman was actually on exactly

197
00:16:49.200 --> 00:16:57.920
the opposite page of him and felt that Altman had used Amade's intelligence, capability, skills,

198
00:16:58.640 --> 00:17:04.800
to build things and bring about a vision of the future that he actually fundamentally didn't

199
00:17:04.800 --> 00:17:10.400
agree with. And so that's why people end up with this bad taste in their mouths. And so I've

200
00:17:10.400 --> 00:17:16.000
been covering the tech industry for over eight years and covered many companies. I've covered

201
00:17:16.000 --> 00:17:24.240
Meta, Google, Microsoft, in addition to OpenAI. And OpenAI and Altman is the only figure that I've

202
00:17:24.240 --> 00:17:32.480
seen this degree of polarization with where people cannot decide whether he's the greatest or the

203
00:17:32.480 --> 00:17:38.960
worst. You mentioned Dario there. And what I found really interesting is to look at how people's

204
00:17:38.960 --> 00:17:43.760
quotes evolve over time with their incentives. So I was looking at all of the things they've

205
00:17:43.760 --> 00:17:48.320
said on the record on podcasts in their blog post to see how it's evolved over time. And Dario,

206
00:17:48.320 --> 00:17:53.600
who was the former VP of Research at OpenAI, and has now moved on to Anthropic, who are taking a

207
00:17:53.600 --> 00:17:58.720
slightly different approach to developing AI, said back in 2017 while he was still open AI,

208
00:17:59.360 --> 00:18:04.800
that this is a quote, I think at the extreme end is the Nick Bosterium style of fear that an

209
00:18:04.800 --> 00:18:10.880
AGI could destroy humanity. I can't see any reason in principle why that couldn't happen.

210
00:18:10.880 --> 00:18:15.920
My chance that something goes really quite catastrophically wrong on the scale of human civilization

211
00:18:16.720 --> 00:18:24.480
might be somewhere between 10% and 25%. And also you mentioned Ilya, who was a co-founder of OpenAI,

212
00:18:24.480 --> 00:18:28.960
and then left. I guess the first question I'd ask is, why did Ilya leave?

213
00:18:29.920 --> 00:18:37.280
That's a great question. So he was instrumental in trying to get Sam Altman fired. And he's

214
00:18:37.280 --> 00:18:43.280
another one of the people who over time began to feel like he was being manipulated by Altman

215
00:18:43.280 --> 00:18:49.440
towards contributing something that he didn't believe in. Because I interviewed a lot of people.

216
00:18:49.440 --> 00:18:59.120
Ilya in particular had two pillars that he cared about deeply. One is making sure we get to

217
00:18:59.120 --> 00:19:06.640
so-called AGI, and the other is making sure that we get to it safely. And he felt that Altman

218
00:19:06.640 --> 00:19:12.880
was actively undermining both things. He felt that Altman was creating a very chaotic environment

219
00:19:12.880 --> 00:19:17.760
within the company where he was pitting teams against each other, where he was telling different

220
00:19:17.760 --> 00:19:25.120
things to different people. Have you ever spoken to him? So I interviewed him in 2019 for a profile

221
00:19:25.120 --> 00:19:31.600
that I did of OpenAI for MIT Technology Review. And back in 2019, he has a quote where he says,

222
00:19:31.600 --> 00:19:35.840
the future is going to be good for AIs regardless. It would be nice if it was also good for humans as well.

223
00:19:36.640 --> 00:19:40.640
It's not that it's going to actively hate humans or want to harm them, but it's just going to be

224
00:19:40.640 --> 00:19:45.040
so powerful. And I think a good analogy would be the way that humans treat animals. It's not that we

225
00:19:45.040 --> 00:19:49.760
hate animals. I think humans love animals, and I have a lot of affection for them. But when the time

226
00:19:49.760 --> 00:19:54.400
comes to build a highway between two cities, we are not asking the animals for permission.

227
00:19:54.400 --> 00:19:59.280
We just do it because it's important to us. And I think by default, that's the kind of relationship

228
00:19:59.280 --> 00:20:05.920
that's going to be between us and AI, which are truly autonomous and operating on their own

229
00:20:05.920 --> 00:20:08.800
behalf. And that was in 2019, the year that you interviewed him.

230
00:20:09.360 --> 00:20:13.760
One of the things that I feel like we should take a step back to examine is going back to this

231
00:20:13.760 --> 00:20:19.280
idea of what even is artificial intelligence and what do we mean by intelligence. And a huge part

232
00:20:19.280 --> 00:20:25.040
of the views of the different people and the quotes that you're reading derives from a specific

233
00:20:25.040 --> 00:20:31.200
belief that they each have in this question of what is intelligence, what constitutes intelligence.

234
00:20:32.160 --> 00:20:40.960
For Ilya, he has throughout his research career felt that ultimately our brains are

235
00:20:40.960 --> 00:20:48.240
giant statistical models. This is not something that we actually know, but this is his own hypothesis,

236
00:20:48.240 --> 00:20:54.480
also the hypothesis of his mentor, Jeffrey Hinton, who also was on this podcast. This is why they

237
00:20:54.480 --> 00:21:00.000
have such a strong conviction in the idea of building AI systems that are statistical models,

238
00:21:00.000 --> 00:21:06.800
and that this particular approach is going to lead to intelligence systems as we are intelligent.

239
00:21:06.800 --> 00:21:13.200
It's a hypothesis that they have. It's not one that has been proven by science. And some people

240
00:21:13.200 --> 00:21:19.440
vehemently disagree with them on this particular thing. But if you step into their shoes and take

241
00:21:19.440 --> 00:21:25.520
on that hypothesis and assume that it's true that our brains are in fact statistical engines,

242
00:21:26.080 --> 00:21:31.840
and that these systems that they're building are also statistical engines that they're making

243
00:21:31.840 --> 00:21:36.960
bigger and bigger and bigger until they become the size of the human brain, that's why they say

244
00:21:37.920 --> 00:21:43.680
that making this comparison where the system will become equal to human intelligence and then

245
00:21:43.680 --> 00:21:50.960
maybe exceed human intelligence is relevant in their framework. And Ilya gave a talk at one point

246
00:21:51.440 --> 00:21:56.160
at this really prominent AI research conference that happens every year called neural

247
00:21:56.160 --> 00:22:03.360
information processing systems. It's mouthful. But he gave this keynote where he shows this chart

248
00:22:03.360 --> 00:22:11.600
of the size of brains and the intelligence of his species. And it's roughly linear, the bigger

249
00:22:11.600 --> 00:22:17.360
the size of the brain, the more intelligent the species. And so for him, he thinks he's

250
00:22:18.080 --> 00:22:24.640
building a digital brain because he thinks brains are just statistical engines. So from that logic,

251
00:22:24.640 --> 00:22:29.280
it's like, okay, if we then build a bigger statistical engine than the human brain,

252
00:22:30.240 --> 00:22:36.400
then based on this chart, it will be more intelligent and then we will be subjected to the same

253
00:22:36.400 --> 00:22:41.680
treatment that we've subjected animals. But it's really important to understand that these are

254
00:22:41.680 --> 00:22:47.920
scientific hypotheses of specific individuals within the AI research community. And there's a lot

255
00:22:47.920 --> 00:22:53.760
a lot of debate about whether this is in fact the case. And some of the biggest critics say

256
00:22:54.320 --> 00:22:58.160
it's very reductive to think of our brains as simply just statistical engines.

257
00:22:58.160 --> 00:23:06.480
Why does it matter to know the mechanism? Is it not just important to know the outcome? Which is

258
00:23:07.040 --> 00:23:11.920
that it's going to be able to do make a video for me or agents are going to be able to do the work

259
00:23:11.920 --> 00:23:17.280
that I do. Does it really, really matter for us to know the mechanism behind it? Yes and no.

260
00:23:18.000 --> 00:23:27.200
So it matters because these companies, they are driving their future actions based on this hypothesis.

261
00:23:27.920 --> 00:23:34.960
So they have decided we think that this hypothesis is true. We should just continue building

262
00:23:34.960 --> 00:23:39.280
larger and larger statistical models in the pursuit of artificial general intelligence.

263
00:23:40.000 --> 00:23:45.040
And that's then having global consequences. In order to continue doing that,

264
00:23:45.040 --> 00:23:49.600
they're hoovering up more and more data. They're building more and more data centers.

265
00:23:49.600 --> 00:23:56.800
They are having exploiting more and more labor in order to continue on this path. Here's a question

266
00:23:56.800 --> 00:24:02.560
that I think is important to ask is why are we trying to build AI systems that are duplicative of

267
00:24:02.560 --> 00:24:08.000
humans? We're kind of having this conversation right now where we've just taken the premise of

268
00:24:08.000 --> 00:24:14.160
this industry as a good thing. Like they said that we should be building AI, so we say that we

269
00:24:14.160 --> 00:24:20.880
should be building AI. But I would like to ask, why are we doing that? Why is it that we are building

270
00:24:20.880 --> 00:24:27.760
a technology that is ultimately designed to replace and automate people away? That is not the

271
00:24:27.760 --> 00:24:34.400
enterprise of technology. Like we should be building technology and the purpose of technology

272
00:24:34.400 --> 00:24:41.440
throughout history has been to improve human flourishing, not to replace people. And so

273
00:24:42.320 --> 00:24:48.320
this is like a critical part of my critique of these companies and the scientists that have just

274
00:24:48.320 --> 00:24:53.360
adopted this goal and have relentlessly pursued it and have had enormous capital and

275
00:24:53.360 --> 00:24:59.920
enormous resources to pursue it. Is this the right goal? Why are we doing this? Why can't we just

276
00:24:59.920 --> 00:25:06.720
build AI systems that do things like accelerate drug discovery and improve people's health care

277
00:25:06.720 --> 00:25:11.760
outcomes, which are systems that have nothing to do with the statistical engines that they're trying

278
00:25:11.760 --> 00:25:16.320
to build to duplicate the human brain? So why are they doing it? I mean, you've interviewed all these

279
00:25:16.320 --> 00:25:21.800
people. I think it's 300 people in total, 80 or 90 of them from OpenAI, the maker of chat

280
00:25:22.520 --> 00:25:27.560
why do you think they're doing it? I think it's because they're driven by an imperial agenda

281
00:25:27.560 --> 00:25:31.720
and that is why I call these companies empires of AI. What do you mean by an imperial agenda?

282
00:25:32.440 --> 00:25:39.640
What does that term mean? Empire is the only metaphor that I've ever found to fully encapsulate

283
00:25:39.640 --> 00:25:46.520
all of the dimensions of what these companies do and the scale that they operate and what motivates

284
00:25:46.520 --> 00:25:51.720
them to do what they do. And there are many parallels that you see between what I call the

285
00:25:51.720 --> 00:25:57.000
empires of AI and the empires of old. They lay claim to resources that are not their own and

286
00:25:57.000 --> 00:26:01.160
the pursuit of training these models. That's the data of individuals, the intellectual property

287
00:26:01.160 --> 00:26:06.520
of artist writers and creators. They're land grabbing in order to build these super computer

288
00:26:06.520 --> 00:26:10.600
facilities for training the next generation models. Second, they exploit an extraordinary amount

289
00:26:10.600 --> 00:26:17.560
of labor. They contract hundreds of thousands of workers all around the world, including in the US

290
00:26:17.560 --> 00:26:24.440
to ultimately make these technologies. We can talk about that more. And they also design their

291
00:26:24.440 --> 00:26:31.080
tools to be labor automating so that when the technologies are deployed, it also affects labor

292
00:26:31.080 --> 00:26:37.320
rights because it erodes away labor rights. And this is a political choice that they have. Third,

293
00:26:37.320 --> 00:26:41.720
they monopolize knowledge production. So they project this idea that they're the only ones that

294
00:26:41.720 --> 00:26:45.800
really understand how the technology works. And so if the public doesn't like it, it's because

295
00:26:45.800 --> 00:26:50.040
they don't actually know enough about this technology. They do this to the public. They do this

296
00:26:50.040 --> 00:26:55.480
to policymakers. And they've also captured the majority of the scientists that are working on

297
00:26:55.480 --> 00:26:59.800
understanding the limitations and capabilities of AI. You think they're gaslighting the public?

298
00:27:00.440 --> 00:27:06.520
Anyway, they are. Yeah. So if most of the climate scientists in the world were bankrolled

299
00:27:06.520 --> 00:27:11.640
by fossil fuel companies, do you think we would get an accurate picture of the climate crisis?

300
00:27:11.640 --> 00:27:17.240
No. And in the same way, they employ and bankrolled the AI industry,

301
00:27:17.240 --> 00:27:22.520
employees and bankrolls most of the AI researchers in the world. So they set the agenda on AI

302
00:27:22.520 --> 00:27:28.920
research in soft ways, simply by funneling money to their priorities so that only certain types

303
00:27:28.920 --> 00:27:36.200
of AI research are produced. But they also will censor researchers when they do not like what the

304
00:27:36.200 --> 00:27:42.760
researcher has found. And so I talk about the case of Dr. Timmy Geberu in my book, who was the ethical

305
00:27:42.760 --> 00:27:50.760
AI team co-lead at Google when she was literally hired to critique the types of AI systems that Google

306
00:27:50.760 --> 00:27:56.760
was building. She then co-wrote a critical research paper that was showing how large language

307
00:27:56.760 --> 00:28:03.960
models specifically were leading to certain types of harmful outcomes. And in an attempt to try

308
00:28:03.960 --> 00:28:10.360
and stop this research from being published, Google ended up firing Geberu and then fired her

309
00:28:10.360 --> 00:28:21.000
other co-lead Margaret Mitchell. And so they control and quash the research that is inconvenient to

310
00:28:21.000 --> 00:28:26.040
the empire's agenda. Did you have an example where this is happening to journalists as well?

311
00:28:26.040 --> 00:28:31.240
I don't know, asking questions of their team members. I think I was watching a video of yours where

312
00:28:31.960 --> 00:28:35.000
there was a young man that was saying he had someone show up at his door,

313
00:28:35.000 --> 00:28:39.960
knocked on his door and asked for information, emails, text messages. And this person was from one

314
00:28:39.960 --> 00:28:47.800
of the big AI companies. This was opening. I started subpoenaing some of its critics. Yeah. As part of a

315
00:28:47.800 --> 00:28:53.960
story, what appears to be a campaign of intimidation, but also what appeared to be a campaign of

316
00:28:53.960 --> 00:29:00.120
phishing for more information to figure out, to map out the network of critics further.

317
00:29:00.120 --> 00:29:07.640
But this was a man who runs a small watchdog nonprofit. And they had been doing a lot of work during

318
00:29:07.640 --> 00:29:14.200
that time to try and ask questions about open AI's attempt to convert from a nonprofit to a

319
00:29:14.200 --> 00:29:19.240
for-profit. Ultimately, opening AI was successful in that conversion. But during the period where

320
00:29:20.200 --> 00:29:25.880
it was sort of existential for open AI to complete this conversion, there were a lot of civil

321
00:29:25.880 --> 00:29:33.560
society groups and watchdog groups like Midas who were trying to prevent the process from happening

322
00:29:33.560 --> 00:29:38.280
in the dead of night. They were trying to get more transparency. They were trying to have more

323
00:29:38.280 --> 00:29:45.240
public debate about this because it's unprecedented. And it was then that there was a knock on his

324
00:29:45.240 --> 00:29:53.240
door and he was served papers. What do the papers say? The papers asked him to reproduce every single

325
00:29:53.240 --> 00:29:58.520
piece of communication that he had had that might have involved Musk. So this was like the strange

326
00:29:58.520 --> 00:30:04.200
paranoia that opening I had that Musk was somehow funding these people to block the conversion.

327
00:30:04.280 --> 00:30:09.720
None of them were actually funded by Musk. So in this particular case, the request he simply was

328
00:30:09.720 --> 00:30:13.160
just answered, you know, I don't have any documents because this doesn't exist.

329
00:30:13.960 --> 00:30:17.640
So going back to this point of empires, you were saying that one of the factors of an empire

330
00:30:17.640 --> 00:30:23.400
is a land grab. And then the next one was was labor exploitation. Labor exploitation.

331
00:30:23.400 --> 00:30:30.520
The third one, controlling knowledge production. And one of the other ones that's really

332
00:30:30.520 --> 00:30:38.120
important to understand about the AI empires in particular is empires always have this narrative

333
00:30:38.120 --> 00:30:45.080
that they say to the public like, we're the good empire. And we need to be an empire in the first

334
00:30:45.080 --> 00:30:51.960
place because there are also bad empires in the world. And if you allow us to take all the resources

335
00:30:51.960 --> 00:30:58.360
and use all the labor, then we promise we will bring you progress and modernity for everyone.

336
00:30:59.080 --> 00:31:06.040
We will bring you to this utopian state akin to an AI heaven. But if the evil empire does it first,

337
00:31:06.040 --> 00:31:10.120
we will descend into a hell. An evil empire being in this case.

338
00:31:10.120 --> 00:31:16.760
In this case, most often it's China. But actually in the early days, open AI evoked Google

339
00:31:16.760 --> 00:31:22.680
as the evil empire. So all of their decisions were about we need to do it first because otherwise

340
00:31:22.680 --> 00:31:28.280
Google this evil corporation that's driven by profit us as a benevolent non-profit,

341
00:31:29.000 --> 00:31:34.840
like this is a this is an critical contest of who wins.

342
00:31:34.840 --> 00:31:40.920
Do you think the people building these AI companies believe that the outcome is going to be all good?

343
00:31:41.640 --> 00:31:45.240
Now, do you think they think that it's going to be it's going to serve everyone,

344
00:31:45.240 --> 00:31:48.040
it's going to be the age of abundance, everything's going to go well.

345
00:31:48.040 --> 00:31:55.400
What do you think they believe? So this is so funny is such a core part of the mythology that

346
00:31:55.400 --> 00:32:03.160
they create around the AI industry includes the belief that it could go very badly. It goes hand

347
00:32:03.160 --> 00:32:10.760
in hand like they need that part of the myth in order to then say and that's why we need to be in

348
00:32:10.760 --> 00:32:13.720
control of the technology because that's the only way that it's going to go really, really well.

349
00:32:14.360 --> 00:32:19.320
And Altman has said publicly, you know, the worst case lights out for everyone,

350
00:32:20.040 --> 00:32:26.040
but best case, we cure cancer, we solve climate change and there's abundance. And Dario Amade,

351
00:32:26.040 --> 00:32:32.920
same kind of rhetoric, who's like worst case, catastrophic or existential harm for humanity,

352
00:32:33.640 --> 00:32:39.480
best case mass human flourishing. So this is like two sides of the same coin, like they have to

353
00:32:39.480 --> 00:32:47.560
use both of these narratives in order to continue justifying an extremely anti-democratic approach

354
00:32:47.560 --> 00:32:52.840
to AI development where there should not be broad participation in developing this technology.

355
00:32:52.840 --> 00:32:55.880
They must be the ones controlling it at every step of the way.

356
00:32:57.080 --> 00:33:01.400
Sam Altman did a tweet saying, there are some books coming out about Open AI and me.

357
00:33:01.400 --> 00:33:05.800
We only participated in two of them, one by Keshe Haguey.

358
00:33:06.600 --> 00:33:10.520
Keeche Haguey focused on me and one by Ashley Vance on Open AI.

359
00:33:12.440 --> 00:33:16.760
He went on to say, no book will get everything right, especially when some people are so intent

360
00:33:16.760 --> 00:33:20.120
on twisting things, but these two authors are trying to.

361
00:33:21.960 --> 00:33:29.480
You quote retweeted that, tweet from Sam Altman and you said, the unnamed book Empire of AI is mine.

362
00:33:31.400 --> 00:33:35.080
Do you believe that tweet from Sam Altman was in reference to your book?

363
00:33:35.160 --> 00:33:37.960
100% because there's only three books coming out about him.

364
00:33:38.760 --> 00:33:43.080
And he caught wind that your book was coming out and he knew my book was coming out because I had

365
00:33:43.080 --> 00:33:47.240
contacted Open AI from the very beginning of my process and said, I'm working on a book now,

366
00:33:47.240 --> 00:33:53.160
will you participate in it? And actually initially they said yes, even though so my history with

367
00:33:53.160 --> 00:33:59.000
Open AI, I profiled the company for MIT Technology Review. I embedded within the office for three days

368
00:33:59.000 --> 00:34:06.600
in 2019. My profile comes out in 2020. The leadership are very unhappy. And in my book I actually

369
00:34:06.600 --> 00:34:13.480
quote an email that I received that Sam Altman sent to the company about my profile saying, yeah,

370
00:34:13.480 --> 00:34:26.120
this is not great. And from then on, the company's stance to me was we are not going to participate

371
00:34:26.120 --> 00:34:30.680
in anything that you do. We are not going to respond to anything, any questions that you receive.

372
00:34:30.680 --> 00:34:36.600
And this was, you know, this was things that they explicitly articulated. It wasn't like me

373
00:34:36.600 --> 00:34:43.720
inferring. So I had a colleague at MIT Technology Review that also covered AI. And at one point,

374
00:34:43.720 --> 00:34:47.320
Open AI sent him this press release being like, we would love for you to cover this story. And he was

375
00:34:47.320 --> 00:34:53.960
like, I'm really busy. Will you send it to Karen? And they were like, oh no, we have a history you

376
00:34:53.960 --> 00:35:02.280
understand. And so for three years, they refused to talk to me. But then I ended up at the

377
00:35:02.280 --> 00:35:08.760
Well Street Journal where if they felt a bit compelled because it was the journal to reopen the

378
00:35:08.760 --> 00:35:15.960
lines of communication. And so I started having, you know, a more dialogue with them. Every time I

379
00:35:15.960 --> 00:35:19.800
wrote a piece, I would always send them, here's my request for comment. I would always ask them,

380
00:35:19.800 --> 00:35:24.600
like, we'll use it for interviews. And we did get to a more productive relationship.

381
00:35:25.320 --> 00:35:29.320
And then I embarked on the book. So I left the journal to focus on the book full time.

382
00:35:29.960 --> 00:35:36.760
And I told them right away, I'm working on this book. I want to continue this productive conversation

383
00:35:36.760 --> 00:35:44.840
where I make sure I reflect Open AI's perspective in the book. And so they were like, we can arrange

384
00:35:44.840 --> 00:35:51.960
interviews for you. You can come back to the office. We'll set up some conversations. And then

385
00:35:51.960 --> 00:35:58.840
as we were going back and forth on this, the board fire Sam Altman. And that's when things started

386
00:35:58.840 --> 00:36:05.560
going kind of south because the company started becoming very sensitive to scrutiny. And so then

387
00:36:05.560 --> 00:36:09.240
they started pushing kicking the can down the road, down the road, down the road. And I kept saying,

388
00:36:09.240 --> 00:36:14.920
hey, when are we rescheduling this? What's going on? And then I get an email saying, we are not going

389
00:36:14.920 --> 00:36:18.200
to participate at all. You are not coming to the office. You're not doing interviews. And I had

390
00:36:18.200 --> 00:36:23.720
actually already booked my tickets. So I was already going to fly to San Francisco to have

391
00:36:24.680 --> 00:36:32.840
the interviews. And so then I told them, I was like, that's fine. I will still engage in the

392
00:36:32.840 --> 00:36:37.960
process. Well, give you extensive requests for comment. As through my reporting, I'll keep

393
00:36:38.040 --> 00:36:43.080
you updated on all the things that I'm finding so that you can choose to still comment. I gave

394
00:36:43.080 --> 00:36:50.040
them 40 pages of requests for comment. And I gave them for a month to respond to all of that.

395
00:36:50.040 --> 00:36:54.760
So this was when the tweet came out was we were doing all this back and forth trying to

396
00:36:56.120 --> 00:37:05.720
and that's when Altman tweeted this. And they never responded to a single one of the 40 pages.

397
00:37:05.720 --> 00:37:10.280
San Montgomery does a lot of interviews. Yeah. You know, still a lot of interviews all the time. He's

398
00:37:10.280 --> 00:37:15.560
done every podcast. I've seen him on everything from Tucker Carlson. I think he's done Theo Von

399
00:37:15.560 --> 00:37:20.680
Joe Rogan. It podcasts all over the world. I wonder why he won't do mine.

400
00:37:25.400 --> 00:37:30.760
Well, maybe I don't know why I think I'm fair with everyone. I just ask questions I genuinely

401
00:37:30.760 --> 00:37:35.400
care about. I don't come in with huge preconceptions. I at least meet people for the first time.

402
00:37:35.480 --> 00:37:40.600
But I've heard through the grapevine that he doesn't want to do mine.

403
00:37:40.600 --> 00:37:46.520
I mean, going back to what you were saying earlier that with the way that opening

404
00:37:46.520 --> 00:37:51.320
eye and these companies control research, you asked, do they also do this with journalists?

405
00:37:52.360 --> 00:37:57.480
I mean, yes, the answer is yes. And apparently they also do it with anyone who has,

406
00:37:57.480 --> 00:38:03.400
you know, a broad mass communications platform. It's not just about the conversation that you're

407
00:38:03.400 --> 00:38:10.360
going to have with them. It's about who you also choose to platform. And there's this huge

408
00:38:10.360 --> 00:38:17.480
problem in technology journalism where companies know that a really big carrot that they can give

409
00:38:17.480 --> 00:38:25.720
to technology journalists is access. And they will withhold that access at the drop of a hat

410
00:38:25.720 --> 00:38:29.880
if they catch wind that you're speaking to someone that they didn't want you to speak to.

411
00:38:29.880 --> 00:38:34.120
This is so true. And I don't think the average person really truly understands this.

412
00:38:34.120 --> 00:38:39.320
Yeah. So this kind of sounds like theory as you say it. But I'm not going to name names here

413
00:38:39.320 --> 00:38:46.440
because I don't think it's important. But there is a particular person in AI who's team

414
00:38:46.440 --> 00:38:51.880
have basically dangled the carrot of them coming here for like 18 months. And I'm like, you don't

415
00:38:51.880 --> 00:38:55.640
have to thank the carrot. I'm going to speak to whoever I want to regardless of the carrot or not.

416
00:38:55.640 --> 00:38:59.480
And when this person comes if they want to come, I'll give them a fair shot. I'll ask them

417
00:38:59.480 --> 00:39:03.560
all genuinely curious questions about what they're doing there incentives. I won't

418
00:39:03.560 --> 00:39:07.400
gotcha them. I don't have a history of ever gotchering anybody. Even if I just think like even if

419
00:39:07.400 --> 00:39:11.640
I have a different opinion, I'll ask the question. Yeah. But they dangle carrots and they say, well,

420
00:39:11.640 --> 00:39:15.880
if you know he might he's thinking about it, let's think about a day. And what the strategy is

421
00:39:15.880 --> 00:39:20.440
and I don't think they think those people don't understand is if we just dangle it for long enough,

422
00:39:20.440 --> 00:39:25.320
then they will perform in the way that we want them to do. And they'll be,

423
00:39:25.640 --> 00:39:31.960
they'll be pleasant about us. They won't be critical. They won't give a, they won't

424
00:39:31.960 --> 00:39:37.240
platform our critics critics. And I think a lot of their game is just dangle the carrot forever.

425
00:39:37.240 --> 00:39:40.840
Yes. Yeah. That's like the optimal outcome is if we just dangle it. If we just tell them,

426
00:39:40.840 --> 00:39:44.600
yeah, no, we're just trying to look at the schedule. It just doesn't work. I think in the

427
00:39:44.600 --> 00:39:48.200
modern world, you just have to go there and give your opinion and allow the clash of ideas

428
00:39:48.200 --> 00:39:53.160
in the public forum. Let the viewers decide for themselves. Yeah. What they think. Yeah.

429
00:39:53.240 --> 00:39:59.080
But this is a, yeah, this is such a huge part of their machinery is the way that they use

430
00:39:59.080 --> 00:40:04.120
these tactics to massage the public image of these companies and make sure that information that

431
00:40:04.120 --> 00:40:10.040
they don't want out and even opinions that they don't want out there go out there. And so this is,

432
00:40:10.040 --> 00:40:17.560
this is, you know, I feel very lucky now that opening I shut the door early on me.

433
00:40:18.520 --> 00:40:22.840
At the time, I didn't feel lucky. I felt like I had screwed myself over. I was like,

434
00:40:22.840 --> 00:40:26.600
should I have been nicer to them and the profile so that I could maintain access?

435
00:40:26.600 --> 00:40:31.320
Which is a horrible question to ask as a journalist, right? Like you're supposed to report the truth

436
00:40:31.320 --> 00:40:35.960
and you're always supposed to report in the interest of the public. Like that is the point of

437
00:40:35.960 --> 00:40:41.400
journalism. And in that moment, I was like relatively junior in my career. I was like,

438
00:40:41.960 --> 00:40:48.120
did I misunderstand what journalism about is about like, should I have actually been playing

439
00:40:48.120 --> 00:40:55.400
the access game? But it was too late. I have the door shut to me. And so I had to build my career

440
00:40:55.400 --> 00:41:00.360
understanding that the door, the front door was never going to be open. Yeah. And that actually

441
00:41:00.360 --> 00:41:09.640
really strengthened my own ability to just tell like it is like that. Yeah. And just report what

442
00:41:09.640 --> 00:41:14.440
I see are the facts being presented to me irrespective of whether the company likes it or not.

443
00:41:14.440 --> 00:41:20.680
And most of the company really does not like it. But I can continue to do the work. They don't

444
00:41:20.680 --> 00:41:25.000
need to open the front door for me. I was still able to do more than 300 interviews.

445
00:41:25.880 --> 00:41:33.160
So Sam Altman gets kicked off the open AI executive team.

446
00:41:33.160 --> 00:41:42.680
Did you find out why that happened? Yeah. There's a scene by scene recounting. From who?

447
00:41:42.680 --> 00:41:47.960
I can't remember the exact number of sources. So I don't want to misquote myself. But it was

448
00:41:47.960 --> 00:41:52.760
around six or seven people that were directly involved or had spoken to people directly involved

449
00:41:52.840 --> 00:42:05.240
in the decision-making process. So Ilya Satskever, I seeing these serious concerns about the way

450
00:42:05.240 --> 00:42:13.960
that Altman's behavior is leading to bad research outcomes and poor decision-making at the company.

451
00:42:15.400 --> 00:42:20.520
He then approaches a board member, Helen Toner. Ilya for anyone that doesn't know,

452
00:42:21.480 --> 00:42:25.320
the co-founder we mentioned earlier, the co-founder of Open AI we mentioned earlier. Yes.

453
00:42:26.280 --> 00:42:34.440
And he kind of does a bit of a sounding board thing to Helen just because Ilya is freaking out.

454
00:42:34.440 --> 00:42:39.640
He's like, he's been like sitting on this, these concerns for a while. And he's like, if I tell

455
00:42:39.640 --> 00:42:50.120
this to someone, this could also be really bad for me if Altman finds out. And so he asks for

456
00:42:50.520 --> 00:42:57.960
a meeting with Toner. And in that first meeting, he's like, he barely says a thing.

457
00:42:57.960 --> 00:43:04.040
He's just like dancing around, trying to figure out, hey, is this someone that I can maybe trust

458
00:43:04.040 --> 00:43:09.080
to divulge more information? And Toner's role in responsibilities at Open AI were...

459
00:43:09.080 --> 00:43:13.640
She was a board member. Yeah. And specifically an independent board member. So

460
00:43:13.640 --> 00:43:19.080
Open AI, when it was a nonprofit, the board was split between people who had a stake,

461
00:43:19.080 --> 00:43:22.920
financial stake in the company, and then people who were fully independent. And this was meant

462
00:43:22.920 --> 00:43:28.520
to be a structure that would balance the decision making to be in the benefit of the public interest

463
00:43:28.520 --> 00:43:36.120
rather than to be in the benefit of the for-profit entity that Open AI then created. And Ilya, as a

464
00:43:37.720 --> 00:43:44.920
non-independent board member, was approaching Toner as an independent board member to try and see

465
00:43:44.920 --> 00:43:51.400
whether or not she was potentially seeing or hearing the same things that he was about the effect

466
00:43:51.400 --> 00:43:56.600
that Altman was having on the company. This then sets off a series of conversations first between

467
00:43:57.320 --> 00:44:05.240
Ilya and Helen and then between Mir Morati and some of the board members. So Mir Morati was

468
00:44:05.240 --> 00:44:10.120
at that point, the chief technology officer of Open AI, where these two senior leaders

469
00:44:10.120 --> 00:44:13.720
essentially through these conversations and through documentation that they're pulling together,

470
00:44:13.720 --> 00:44:18.600
like email Slack messages and so forth, they convey to the independent board members,

471
00:44:18.600 --> 00:44:27.080
three independent board members. We are very concerned about Altman's leadership. He is creating

472
00:44:27.880 --> 00:44:37.800
too much instability at the company. And he is the root of the problem. They were trying to

473
00:44:37.800 --> 00:44:43.480
say to these independent board members, the problem will not be fixed unless Altman is removed.

474
00:44:43.720 --> 00:44:49.480
Because of the way that he's pitting teams against each other and creating this environment where

475
00:44:49.480 --> 00:44:54.280
people are unable to trust each other anymore. And they're competing rather than collaborating on

476
00:44:54.280 --> 00:44:59.240
what's supposed to be this really, really important technology. When you say instability,

477
00:45:00.200 --> 00:45:04.280
that's quite a vague term. That could mean lots of things. Like instability could mean pushing

478
00:45:04.280 --> 00:45:10.120
people hard to work harder. What do you mean by instability? In specific times as you can

479
00:45:10.200 --> 00:45:16.040
possibly say them. When chat GPT came out in the world, Open AI was wholly unprepared.

480
00:45:17.000 --> 00:45:23.480
They didn't think that they were launching a gangbusters product. They thought they were releasing

481
00:45:23.480 --> 00:45:29.000
a research preview that would help them get the data flywheel going, collect a bunch of data

482
00:45:29.000 --> 00:45:34.360
from users that would then inform what they thought would be the gangbusters product,

483
00:45:34.360 --> 00:45:43.640
which was a chatbot using GPT 4. And chat GPT was using GPT 3.5. Because of that,

484
00:45:44.680 --> 00:45:52.200
there were servers crashing all the time because they had to scale their infrastructure faster than

485
00:45:52.200 --> 00:45:58.360
any company in history. And there were all of these outages. They were trying to also

486
00:45:58.360 --> 00:46:03.320
hire faster than any company in history to try and have more personnel there. And they were then

487
00:46:03.400 --> 00:46:07.160
sometimes hiring people that they were like, actually, we made a mistake. We shouldn't have hired

488
00:46:07.160 --> 00:46:12.040
you. So they were firing people left and right. And people were just disappearing off of slack.

489
00:46:12.040 --> 00:46:16.840
And that's how their colleagues would learn that they were no longer at the company. And so

490
00:46:16.840 --> 00:46:23.640
it was, yes, like many fast growing companies, a very chaotic environment, and a particularly

491
00:46:23.640 --> 00:46:31.160
chaotic environment because it was extra fast. Like they had to accelerate more than any other

492
00:46:31.160 --> 00:46:38.360
startup. And on top of that mirror morality and Elise, that's never felt that Altman was making

493
00:46:38.360 --> 00:46:44.920
it worse. Like he was not actually effectively ameliorating the circumstances of the chaos. He

494
00:46:44.920 --> 00:46:51.640
was actually sowing more chaos, getting these teams to be more divided. And this is where

495
00:46:52.840 --> 00:46:57.640
it's important to understand that the executives and the independent board members,

496
00:46:58.440 --> 00:47:04.920
they're all operating under this idea that they're building AGI and that AGI could either be

497
00:47:04.920 --> 00:47:13.640
devastating or utopic to humanity. And so it's not, yes, it's like any other company. And no,

498
00:47:13.640 --> 00:47:19.400
it's not like any other company. You cannot have, like in their view, you cannot have the

499
00:47:19.400 --> 00:47:26.040
degree of chaos as the pressure cooker for creating a technology that they in their conception

500
00:47:26.040 --> 00:47:32.600
could make or break the world. And so that is basically what the independent board members

501
00:47:32.600 --> 00:47:36.920
also begin to reflect on. They have these conversations amongst themselves where they're like,

502
00:47:38.760 --> 00:47:42.840
well, based on what we're hearing about Altman's behavior, like if this was an insta-cart,

503
00:47:43.720 --> 00:47:49.480
would that warrant firing him? And they concluded, maybe not. But this is not insta-cart.

504
00:47:49.880 --> 00:47:59.240
And that's why they were like, well, crap. Maybe this does rise to the bar where we should consider

505
00:47:59.240 --> 00:48:05.560
replacing him because we are ultimately building a technology that we think could have

506
00:48:06.440 --> 00:48:12.040
transformative impacts, either in the positive or negative direction. And so that is what happens.

507
00:48:12.040 --> 00:48:15.880
It's like these two executives. And then the independent board members also, they were hearing

508
00:48:15.880 --> 00:48:20.280
other feedback as well from their connections within the company with other people in the industry.

509
00:48:20.280 --> 00:48:25.000
At one point, Adam DiAngelo, who is one of the independent board members and the CEO of Cora,

510
00:48:25.720 --> 00:48:33.160
which is a tech startup in the valley, he is out of party in San Francisco. And he starts to hear

511
00:48:33.160 --> 00:48:40.360
some of these rumors that there's something weird about the way that OpenAI has structured

512
00:48:41.000 --> 00:48:47.320
its OpenAI startup fund, which was this fund that the company had created to start investing in

513
00:48:47.320 --> 00:48:55.080
other startups. And he realizes they'd never really seen documentation about how the startup

514
00:48:55.080 --> 00:48:59.640
fund had been set up from Altman. And finally, they get the documents and it turns out that OpenAI

515
00:48:59.640 --> 00:49:06.920
startup fund is not OpenAI's startup fund. It's Altman's startup fund. And this was something,

516
00:49:07.000 --> 00:49:11.800
like one of several experiences that independent board members were also having where they're like,

517
00:49:12.600 --> 00:49:20.120
there's something not right about the fact that they're continuously in consistencies between

518
00:49:20.120 --> 00:49:26.760
the way that Altman is portraying what is being done versus what is actually being done. And so

519
00:49:26.760 --> 00:49:32.360
when these two executives approach the board or the independent board members, then they're like,

520
00:49:32.440 --> 00:49:36.680
okay, this lines up with also the experiences that we've been having.

521
00:49:38.280 --> 00:49:44.760
And at that point, they then have this series of very intense discussions where they're meeting

522
00:49:44.760 --> 00:49:54.200
almost every day, talking about, should we actually really consider removing Altman? And in the end,

523
00:49:54.200 --> 00:50:00.120
they conclude, yes, we should. And if we're going to do it, we need to do it quickly because

524
00:50:00.120 --> 00:50:05.320
they were very concerned that the moment that Altman found out his persuasive abilities would

525
00:50:05.320 --> 00:50:13.320
make it impossible to do. And so they end up firing Altman without telling anyone. You know,

526
00:50:13.320 --> 00:50:20.120
they don't talk to any stakeholders to get them on the same page. Microsoft gets a call right before

527
00:50:20.120 --> 00:50:24.040
they execute the action saying, we're going to fire Altman. And Microsoft friend, that doesn't know

528
00:50:24.520 --> 00:50:31.720
lead investor in open air at the time. Yes. One of the only investors in OpenAI at the time.

529
00:50:32.920 --> 00:50:39.160
And that is what then devolves the whole thing because every single person that is affected by this

530
00:50:39.160 --> 00:50:47.000
decision is now extremely angry that they were not involved. And that is what then creates this

531
00:50:47.000 --> 00:50:52.760
campaign to bring Altman back. And then Altman is reinstalled as CEO days later.

532
00:50:53.880 --> 00:50:57.320
This company that I've just invested in is growing like crazy. I want to be the one to tell you

533
00:50:57.320 --> 00:51:00.600
about it because I think it's going to create such a huge productivity advantage for you. This

534
00:51:00.600 --> 00:51:04.840
beflows an app that you can get on your computer and on your phone on all your devices. And it

535
00:51:04.840 --> 00:51:08.600
allows you to speak to your technology. So instead of me writing at an email, I click one button on

536
00:51:08.600 --> 00:51:14.840
my phone and I can just speak the email into existence. And it uses AI to clean up what I was saying.

537
00:51:14.840 --> 00:51:18.760
And then when I'm done, I just hit this one button here. And the whole email is written for me.

538
00:51:18.760 --> 00:51:24.600
And it's saving me so much time in a day because Whisper learns how I write. So on what's app it

539
00:51:24.600 --> 00:51:28.360
knows how I am a little bit more casual, on email a little bit more professional. And also,

540
00:51:28.360 --> 00:51:32.360
there's this really interesting thing they've just done. I can create little phrases to automatically

541
00:51:32.360 --> 00:51:37.160
do the work for me. I can just say Jack's LinkedIn. And it copies Jack's LinkedIn profile for me because

542
00:51:37.160 --> 00:51:41.320
it knows who Jack is in my life. This is saving me a huge amount of time. This company is growing like

543
00:51:41.320 --> 00:51:45.240
absolute crazy. And this is why I invested in the business and why they're now a sponsor of this

544
00:51:45.240 --> 00:51:49.960
show. And Whisper Flow is frankly becoming the worst kept secret in business productivity

545
00:51:49.960 --> 00:51:58.120
entrepreneurship. Check it out now at Whisper Flow, spot W-I-S-P-R-F-L-O-W dot AI slash Stephen.

546
00:51:58.120 --> 00:52:02.760
It will be a game change if you. There's a phase a lot of companies here where they're no longer

547
00:52:02.760 --> 00:52:07.160
doing the most important thing, which is selling and they get really bogged down with Admin. And it's

548
00:52:07.160 --> 00:52:11.640
often something that creeps up slowly and you don't really notice until it's happened. Slowly,

549
00:52:11.640 --> 00:52:16.120
momentum starts to leak out. This happened to us and our sponsor pipe drive was a fix I came

550
00:52:16.120 --> 00:52:20.840
across 10 years ago. And ever since, my teams across my different companies have continued to use it.

551
00:52:20.840 --> 00:52:25.640
pipe drive is a simple but powerful sales 0M that gives you the visibility on any deals in your

552
00:52:25.640 --> 00:52:30.440
pipeline. It also automates a lot of the tedious repetitive and time consuming parts of the sales

553
00:52:30.440 --> 00:52:34.600
process, which in turn saves you so many hours every single month, which means you can get back

554
00:52:34.600 --> 00:52:38.440
to selling. Making that early decision to switch to pipe drive was a real game changer,

555
00:52:38.440 --> 00:52:43.240
and it's kept the right things front of mind. My favorite feature is pipe drives ability to sync

556
00:52:43.240 --> 00:52:48.520
your CRM with multiple email inboxes so your entire team can work together from one platform,

557
00:52:48.520 --> 00:52:53.720
and we aren't the only ones benefiting. Over 100,000 companies use pipe drive to grow their business.

558
00:52:53.720 --> 00:53:00.200
So something I've said resonates, head over to pipedrive.com slash CEO, where you can get a 30-day

559
00:53:00.200 --> 00:53:08.200
free trial, no credit card or payment required. How does a CEO of a major company get fired

560
00:53:08.200 --> 00:53:12.920
by the board? Because board members, there's a quote in your book on page 357 where you say

561
00:53:12.920 --> 00:53:17.320
about Ilya saying, I don't think Sam is the guy who should have the finger on the button for

562
00:53:17.320 --> 00:53:22.840
AGI. Now, I ask myself this question. I work with lots of people here. We have 150 people that

563
00:53:22.840 --> 00:53:30.280
work in this business, and those people know me best. They see me on camera. They see me off

564
00:53:30.280 --> 00:53:35.240
camera. So if they said that we don't think Stephen is the right person to host the die of a CEO,

565
00:53:35.240 --> 00:53:41.480
for example, it would take a lot for them to say that. They must have seen some shit off camera.

566
00:53:41.480 --> 00:53:46.600
We don't think he's the right person to be on camera of whatever reason. In the case of AGI,

567
00:53:46.600 --> 00:53:49.880
which is much more consequential than a podcast that is filmed in my kitchen,

568
00:53:49.880 --> 00:53:56.440
it almost sends a chill down one's body to think that the co-founder of a business has gone to

569
00:53:56.440 --> 00:54:00.360
the board and said, this isn't the guy to lead this consequential. And it wasn't just Ilya,

570
00:54:00.360 --> 00:54:06.120
Miriam Roddy then also said, I don't think Altman is the right guy. And then they both left. Later.

571
00:54:06.120 --> 00:54:12.840
So then Altman comes back and Loan beholds Ilya never comes back. So his concerns about the

572
00:54:12.840 --> 00:54:18.120
fact that Altman founding out would be bad for him manifested. He ended up not coming back,

573
00:54:18.120 --> 00:54:23.800
and Miriam Roddy then left shortly thereafter. Quite a lot of these people leave, don't they,

574
00:54:23.800 --> 00:54:33.640
open AI? They do. So if you consider one of the origin stories of open AI,

575
00:54:33.640 --> 00:54:40.520
is this dinner that happened at the Rosewood Hotel, which is a very swanky hotel right in the

576
00:54:40.520 --> 00:54:46.280
heart of Silicon Valley, that was one of Elon Musk's favorites whenever he was coming up from LA

577
00:54:46.280 --> 00:54:52.200
to the area. And there was this dinner that was there where Altman was intending to recruit

578
00:54:52.920 --> 00:54:59.880
the OG team that would start open AI. So he's kind of telling everyone, you might have a chance to

579
00:54:59.880 --> 00:55:06.760
meet Musk because Musk is going to come to this dinner dinner and he cold emails Ilya and gets Ilya

580
00:55:06.760 --> 00:55:11.160
to then come because and Ilya specifically wants to come because he wants to meet Musk. And he

581
00:55:11.160 --> 00:55:16.600
also emails all these other people, including Greg Rockman, Daria Amade. These people that end up

582
00:55:16.600 --> 00:55:22.920
working it. And they all almost all of them, not not every one of them, but almost all of them end

583
00:55:22.920 --> 00:55:31.080
up working at open AI. And leaving almost all of them end up leaving specifically after they clash

584
00:55:31.080 --> 00:55:39.400
with Altman. And Ilya, he left and launched a company called Safe Super Intelligence. Yeah.

585
00:55:40.280 --> 00:55:49.160
Which is, I mean, that's an indirect if I've ever heard one. If someone like co-founded this

586
00:55:49.160 --> 00:55:58.040
podcast with me and then they left and started a podcast called Safe Podcasting, I'd take that as a

587
00:55:58.040 --> 00:56:05.320
slight. I'd have people knocking on my door and I'll speak for later next. One of the things that

588
00:56:06.280 --> 00:56:14.600
is happening here is it is not a coincidence that every single tech billionaire has their own AI

589
00:56:14.600 --> 00:56:25.720
company. They want to create AI in their own image. And that's why they keep not getting along.

590
00:56:25.720 --> 00:56:31.000
And in fact, it's not just don't get along. They end up hating each other after working together.

591
00:56:31.960 --> 00:56:39.240
And then splinter off into their own organizations. So after Musk leaves, he starts XAI. After Daria

592
00:56:39.240 --> 00:56:45.000
leaves, he starts Anthropic. After Ilya leaves, he starts Safe Super Intelligence. After Mira leaves,

593
00:56:45.000 --> 00:56:55.480
she starts thinking Machines Lab. They want to have control over their own vision of this technology.

594
00:56:56.120 --> 00:57:07.080
And the best way that they have derived from their experiences of trying to put their vision into

595
00:57:07.080 --> 00:57:12.600
the arena is by creating a competitor and then competing with OpenAI and with all the other

596
00:57:12.600 --> 00:57:15.960
companies out there. Do you think some of these AICOs realize that they are quite literally

597
00:57:15.960 --> 00:57:22.440
summoning the demon as Elon said 10 years ago? But they don't really care because being the person

598
00:57:22.440 --> 00:57:30.040
that summoned the demon makes you consequential and powerful and historical, even if the outcome

599
00:57:30.040 --> 00:57:34.840
is potentially horrific, even if there's like a 20% outcome of it being horrific. I remember,

600
00:57:34.840 --> 00:57:40.600
I think it was Daria, he's the one that said there's somewhere between a 10% and 25% chance

601
00:57:41.240 --> 00:57:48.360
of things going catastrophically wrong on the scale of human civilization. 25% is a one in four

602
00:57:48.360 --> 00:57:57.880
chance. If you put bullets in a four chamber revolver and said Stephen, the upside is you could

603
00:57:57.880 --> 00:58:02.360
become a multi-gazillionaire and be remembered forever. The downside is that there a good bullet

604
00:58:02.360 --> 00:58:08.760
in your head. There is no chance that I would take that better with a 25% potential chance of

605
00:58:08.760 --> 00:58:14.040
things going catastrophically wrong. So I have a very long answer to this because

606
00:58:14.440 --> 00:58:19.800
do they know if they're summoning the demon? It really depends on what we define as summoning the

607
00:58:19.800 --> 00:58:27.560
demon. And in this particular case, to go back to what we were saying before, there's a mythology

608
00:58:27.560 --> 00:58:36.760
that the AI industry uses where summoning the demon is an integral part of convincing everyone

609
00:58:36.760 --> 00:58:41.080
that therefore they can be the only ones that are developing this technology.

610
00:58:41.160 --> 00:58:46.120
I got it. So on one end, you've got to say, if we don't shine a will, and that's terrible.

611
00:58:46.120 --> 00:58:50.120
Yeah. But if we let anyone else do it other than me, then we're fucked as well.

612
00:58:50.840 --> 00:58:54.280
Exactly. So that means that I have to do it and you have to give me money and support.

613
00:58:54.280 --> 00:59:02.040
Exactly. So when they're saying these things, we should understand it as not as like a genuine

614
00:59:02.040 --> 00:59:06.120
prediction based on what they're seeing. Because first of all, we don't predict the future. We make it.

615
00:59:07.080 --> 00:59:12.840
We should understand this as an act of speech to persuade other people into believing

616
00:59:13.400 --> 00:59:19.400
that they should see more power, more resources to these individuals. And so do they know that

617
00:59:19.400 --> 00:59:24.200
they're summoning the demon? I mean, they are purposely trying to create this

618
00:59:27.160 --> 00:59:32.760
feeling within the public that they are because it is a crucial part of their power.

619
00:59:33.720 --> 00:59:41.160
But if we were to define, just do they realize that the things that they're doing are having

620
00:59:41.160 --> 00:59:47.400
already really harmful impacts all around the world on vulnerable people, vulnerable communities,

621
00:59:47.400 --> 00:59:54.440
vulnerable countries. That's where I'm like, maybe yes, maybe no. And they don't really care. Because

622
00:59:55.800 --> 01:00:01.640
in the frame of mind, like I sometimes use the analogy that they are, I world is like, dude.

623
01:00:02.440 --> 01:00:06.120
Dude, for anyone that doesn't know, dude. Science fiction epic written by Frank Herbert.

624
01:00:06.760 --> 01:00:12.120
And it's set in this intergalactic era where there are all these houses and they're fighting each

625
01:00:12.120 --> 01:00:18.040
other for spice. So it's a call back to colonialism and empire. And they all are trying to control

626
01:00:18.040 --> 01:00:23.880
this spice. But one of the features of this story is that there are these myths that are seeded

627
01:00:23.880 --> 01:00:29.400
on the different planets about a scent, a religious myth basically about the coming of the Messiah

628
01:00:29.480 --> 01:00:36.840
that are used as a way to control the people. And politradies, when he arrives at the planet

629
01:00:36.840 --> 01:00:46.280
orakis with with the intention of trying to then fight against the empire and avenge his

630
01:00:46.280 --> 01:00:53.000
father's death, he steps into a myth that has been seeded on this planet that says that one day

631
01:00:53.000 --> 01:00:58.200
there will be a Messiah that comes and saves the planet. So he steps into the role of the Messiah

632
01:00:58.200 --> 01:01:06.600
and leans into this idea in order to better control the people and rally them behind him as a leader

633
01:01:06.600 --> 01:01:14.520
to help with this quest. He knows that it's a myth in the beginning, but because he lives and breathes

634
01:01:14.520 --> 01:01:20.360
and embodies it, it kind of starts to blur in his mind whether this is really a myth or whether

635
01:01:20.360 --> 01:01:27.480
he's really the Messiah. And this is what I think happens in the AI world. On one hand,

636
01:01:28.360 --> 01:01:35.480
there are all these executives that actively engage in myth making because I have all these

637
01:01:35.480 --> 01:01:40.840
internal documents that I write about in the book where they are very keenly aware of how to

638
01:01:40.840 --> 01:01:49.160
bring the public along with them by showing them dazzling demonstrations of the technology by using

639
01:01:49.160 --> 01:01:57.160
crafting a mission that will sound really good and make people give more leniency to their

640
01:01:57.160 --> 01:02:04.040
companies. So they know they're doing the myth making and also I think many of them lose themselves

641
01:02:04.040 --> 01:02:10.760
in the myth because they have to live and breathe and embody it day in and day out. And so when

642
01:02:10.760 --> 01:02:18.360
Dario says he thinks that 10 to 25% of the future could be catastrophic or whatever the probability

643
01:02:18.360 --> 01:02:24.920
is 10 to 25%. He is actively engaging in the myth making, but also he's losing himself in the myth.

644
01:02:24.920 --> 01:02:29.640
Like I think if you were to ask him, do you genuinely believe that? He would be like yes, I genuinely

645
01:02:29.640 --> 01:02:36.360
believe that because there's been a blurring of when he's saying something just to say something

646
01:02:37.000 --> 01:02:45.080
versus when he actually believes what he's required to believe in order to then continue

647
01:02:46.040 --> 01:02:51.880
doing the things that he's doing. And this is the whole psychology of cognitive

648
01:02:51.880 --> 01:02:56.120
dissonance right where you the brain struggles to hold too conflicting world views at the same time.

649
01:02:56.120 --> 01:03:02.360
So it's incentivized or it endeavors to dismiss one. So if you want to be a healthy person but

650
01:03:02.360 --> 01:03:06.840
also a smoker and I pointed out smoking is bad for you. The first words out of your mouth are going

651
01:03:06.840 --> 01:03:14.120
to be yes, but yeah, it helps me with stress. Yes, but I only do it when I think I don't know,

652
01:03:14.120 --> 01:03:18.600
I kind of see that at the moment because these companies have to raise extortion at like huge

653
01:03:18.600 --> 01:03:23.960
amounts of money to fund their AI research and they're building out all of these data centers.

654
01:03:24.760 --> 01:03:28.360
So when they're out in the public, they're always fundraising. All of these major companies are

655
01:03:28.360 --> 01:03:32.280
fundraising all the time at the moment. So you can't be fundraising and saying I'm going to destroy

656
01:03:32.280 --> 01:03:35.320
your children's future potentially. There's 25% chance that your children aren't going to

657
01:03:36.040 --> 01:03:41.480
have a great life, which might be the truth. I mean that is actually what they say. This is

658
01:03:41.480 --> 01:03:46.600
what famously Dario Amade does. He does that. The others sound not doing that as much anymore.

659
01:03:46.600 --> 01:03:54.040
Yes, and it's because it goes back to each of them kind of distinguishes themselves a little bit as

660
01:03:55.160 --> 01:04:01.320
the brand that they need to project. Do you think any of them are more have a stronger moral compass

661
01:04:01.320 --> 01:04:07.800
than others? Because I think Dario often gets the credit for having more of a backbone and being

662
01:04:07.800 --> 01:04:14.520
more conscious of implications. He does get a lot of credit for that. He's from Claude and

663
01:04:14.520 --> 01:04:22.440
Anthropic for anyone that doesn't know. I don't think it truly matters that question. The answer

664
01:04:22.440 --> 01:04:29.080
is that question because to me, even if you were to swap all the CEOs for someone that people would

665
01:04:29.080 --> 01:04:35.480
say is better at running these companies, it doesn't fix the problem that I identify in the book,

666
01:04:36.280 --> 01:04:40.840
that there is a system of power that has been constructed where these companies and the people

667
01:04:40.840 --> 01:04:46.120
running these companies get to make decisions that affect billions of people's lives around the world

668
01:04:46.120 --> 01:04:53.320
and those billions of people do not get any say in how it goes. Those people, they can go to the polls,

669
01:04:53.320 --> 01:04:58.520
right? So if the public are sufficiently educated, they can go to the polls and pick a leader that

670
01:04:58.520 --> 01:05:05.320
says they're going to legislate or pass laws or try and pass laws. Yes, but at the speed

671
01:05:05.320 --> 01:05:11.080
and pace at which these companies operate and at the sheer scale and size, they're able to also

672
01:05:11.080 --> 01:05:15.480
spend extraordinary amounts of money, hundreds of millions in this upcoming midterms,

673
01:05:15.480 --> 01:05:20.040
to try and kill every possible piece of legislation that gets in their way and craft legislation

674
01:05:20.040 --> 01:05:27.880
that would codify their advantage. And so to me, I think sometimes as a society, we obsess a little

675
01:05:28.120 --> 01:05:35.800
bit with, are these leaders good or bad people? And to me, the bigger question is,

676
01:05:36.520 --> 01:05:42.760
is the governance structure that we've created a sound one or that allows broad participation

677
01:05:42.760 --> 01:05:47.880
or an anti-democratic one that has consolidated this decision-making power in the hands of the few,

678
01:05:47.880 --> 01:05:53.720
because no person is perfect. I don't care who is on the top of these companies,

679
01:05:53.720 --> 01:05:59.320
they're not going to have the ability to make decisions on behalf of so many people around the

680
01:05:59.320 --> 01:06:06.680
world who live and talk and have a culture and history that are fundamentally different from them

681
01:06:06.680 --> 01:06:14.520
without things going wrong. And so that is why throughout history, we've moved from empires

682
01:06:14.520 --> 01:06:22.920
to democracy. It's because empire as a structure is inherently on sound. It does not actually

683
01:06:22.920 --> 01:06:29.080
maximize the chances of most people in the world being able to live dignified lives.

684
01:06:29.720 --> 01:06:33.640
I'm going to try and take on their point of view. So this is me playing devil's advocate, okay.

685
01:06:34.680 --> 01:06:43.080
But Karen, if the US don't continue to accelerate their research with AI, at some point, China's model

686
01:06:43.080 --> 01:06:48.120
is going to become so smart and intelligent that we're basically going to have to rent it off

687
01:06:48.520 --> 01:06:54.440
them and they'll get the scientific discoveries. They'll discover the new era of autonomous weapons

688
01:06:54.440 --> 01:07:02.600
and we will be their backyard. And like logically, that argument does appear to be pretty true.

689
01:07:02.600 --> 01:07:07.080
No, it's not. If we scale up, if we just imagine any rate of change with this intelligence,

690
01:07:07.080 --> 01:07:14.200
at some point, we're going to come to a weapon that could theoretically disable all of the United

691
01:07:14.200 --> 01:07:19.560
States electricity, their weapons systems. It would know exactly how to disable the United States

692
01:07:19.560 --> 01:07:23.880
from a side of perspective because it would be that smart. All you've got to imagine is any rate

693
01:07:23.880 --> 01:07:30.520
of improvement or sort of a long period of time. So this is a theory that might be true.

694
01:07:31.400 --> 01:07:38.840
And if it's true, any theory might be true. But again, going to this point of even if it's

695
01:07:38.840 --> 01:07:41.320
a small percentage, it's worth paying attention to on the other side of the foot,

696
01:07:42.280 --> 01:07:47.880
this is a theory that people talk about. It could be the case that the most intelligent

697
01:07:47.880 --> 01:07:53.720
civilization is going to be the superior civilization. Logically, that's a pretty something to say,

698
01:07:53.720 --> 01:08:00.760
no? So there's a lot of a lot of fundamentals in this argument that would need to be true in

699
01:08:00.760 --> 01:08:06.120
order for this to be a viable argument. And let's knock them down one by one. So the first one is

700
01:08:06.920 --> 01:08:13.240
that these systems are intelligent. And that just scaling them is going to bring us more

701
01:08:13.240 --> 01:08:21.240
intelligence. So far so true? No, it's actually not. Because first of all, again, we don't actually know

702
01:08:22.200 --> 01:08:28.360
if these systems are like intelligence is not, it's not like the right analogy almost. It's sort of like,

703
01:08:30.440 --> 01:08:34.840
it's like is a calculator, a calculator can do math problems faster than a human. Does that make it

704
01:08:34.840 --> 01:08:39.320
intelligent? It has a narrow intelligence because they're solving a narrow problem, which is like

705
01:08:39.320 --> 01:08:46.520
1 plus 1 equals 2. And these systems, they actually also are quite narrowly intelligent. In the sense

706
01:08:46.520 --> 01:08:50.360
that even though these companies say that they're everything machines that can do anything for

707
01:08:50.360 --> 01:08:55.720
anyone, they actually can only do some things for some people. This is like the jagged frontier

708
01:08:55.720 --> 01:09:00.280
of these AIR models. Like some of the capabilities are quite good. Other capabilities are not that

709
01:09:00.280 --> 01:09:05.080
good. You know why that happens? Because the company can only focus on advancing certain types

710
01:09:05.080 --> 01:09:09.640
of capabilities. I can't literally focus on advancing all types of capabilities. They have to

711
01:09:09.640 --> 01:09:13.800
actually set their mind to advancing a certain by gathering the data that is needed for that

712
01:09:13.800 --> 01:09:22.840
capability by getting a bunch of human contractors to annotate and train the model to do that exact

713
01:09:22.840 --> 01:09:32.440
thing. And so scaling these models is actually a perpendicular question to, are we actually getting

714
01:09:34.040 --> 01:09:39.000
more cyber capability specifically and more military capability specifically?

715
01:09:39.000 --> 01:09:44.040
I would argue that most of the top people in AI believe that the intelligence is going to

716
01:09:44.040 --> 01:09:47.880
continue to scale for some time. A lot of them do. Like Jeffrey Hinton does.

717
01:09:47.960 --> 01:09:54.280
And again, it's back to his hypothesis about how human intelligence works and what the appropriate

718
01:09:54.280 --> 01:10:00.520
model of the brain is. His hypothesis throughout his career has been the brain is a statistical engine.

719
01:10:00.520 --> 01:10:06.440
But that's his hypothesis and that is not universally agreed upon, especially among people that are

720
01:10:06.440 --> 01:10:10.920
not in the AIR world. When you talk with neuroscientists and psychologists, people who actually study

721
01:10:10.920 --> 01:10:16.040
human intelligence in the human brain, that is where you start to get a lot of debate and disagreement

722
01:10:16.120 --> 01:10:24.440
about this particular view that Hinton has. And so this is kind of like one of the, one of the things

723
01:10:24.440 --> 01:10:31.400
is like AI is already being used in the military and has been used in the military for a long time.

724
01:10:31.400 --> 01:10:41.400
But it's specifically accelerating large language models isn't just the only path for getting

725
01:10:41.400 --> 01:10:47.400
military, like the companies would have to choose to specifically pick military capabilities to

726
01:10:47.400 --> 01:10:52.600
accelerate, not just like Shiong Aron Tellet. It's like, you know what I'm saying? Like they create

727
01:10:52.600 --> 01:10:58.040
this myth that they are actually pushing the frontier of all of the capabilities of the model.

728
01:10:58.040 --> 01:11:02.760
But that's not what's actually happening internally. And I have, I had hundreds of pages of documents

729
01:11:02.760 --> 01:11:07.880
on like how they were specifically training models. They pick what capabilities they want to advance.

730
01:11:07.880 --> 01:11:13.000
And you know how they pick them? It's based on which industries would be able to pay them the

731
01:11:13.000 --> 01:11:22.120
most money for their services. So they pick finance, law, medicine, healthcare, commerce. It's not

732
01:11:22.120 --> 01:11:29.000
actually intelligent like a, like a, a baby where you, the more that you, that the baby grows up,

733
01:11:29.000 --> 01:11:32.440
they start having this like general, these general abilities.

734
01:11:32.440 --> 01:11:39.000
I think I have dragon intelligence. I wasn't going to say it. But I think I know a little,

735
01:11:39.000 --> 01:11:44.120
I know a little bit about, no, I know a lot about a little bit. Yeah, but it's, but you also have

736
01:11:44.120 --> 01:11:47.960
the capability to learn and acquire knowledge by yourself. And you also have the ability to choose

737
01:11:47.960 --> 01:11:51.880
what you're going to learn and acquire by yourself. It's not easy. And it takes a lot more time

738
01:11:51.880 --> 01:11:56.920
than these models, it seems, less compute. And you can learn how to drive in one place and then

739
01:11:56.920 --> 01:12:01.960
immediately know how to drive in another place. These models cannot do that. Every time a self-driving

740
01:12:02.040 --> 01:12:08.440
car is shifted to another location, it has to completely retrain on that location. It's like all

741
01:12:08.440 --> 01:12:11.560
the self-driving cars. I mean, we're sitting in Austin right now and there's all these self-driving

742
01:12:11.560 --> 01:12:17.640
cars that are driving through Austin. But when one of them learns they all learn, which is,

743
01:12:17.640 --> 01:12:24.600
well, it's just because it's, it's an operating system that has an AI model as part of it and

744
01:12:24.600 --> 01:12:28.840
you're training the AI model and then you deploy the AI model across all the self-driving cars.

745
01:12:28.840 --> 01:12:34.600
This is a big advantage because if one optimist robot learns one thing in one factory,

746
01:12:34.600 --> 01:12:39.000
they all learn it. And imagine that. Imagine if humans, if we all learn what all the other humans

747
01:12:39.000 --> 01:12:42.760
learned, that would be, that would give us such an unbelievable competitive advantage. I mean,

748
01:12:42.760 --> 01:12:45.640
one of the ways we did that is through communication. Or it could not because they could be learning

749
01:12:45.640 --> 01:12:50.120
the wrong thing, which has also happened again and again with these technologies is that all of them

750
01:12:50.120 --> 01:12:54.440
that learn the wrong thing and they all have the same failure mode. I mean, part of the resilience

751
01:12:54.520 --> 01:12:58.840
of human society is that we do have different expertise and we also have different failure modes.

752
01:12:59.400 --> 01:13:02.840
I think sometimes we hold AI models to a higher standard than we hold humans to.

753
01:13:03.560 --> 01:13:07.400
And in a way, because I, I'd hear on stage, we're in, we're in Austin at the moment and I'd hear

754
01:13:07.400 --> 01:13:12.520
people go, ah, but you know, then AI models, they hallucinate sometimes. I'm like, have you met a

755
01:13:12.520 --> 01:13:17.960
human? Like, I hallucinate all the time. I can barely spell all due math.

756
01:13:17.960 --> 01:13:24.760
Yes, but it's, it's once again like using this analogy that was specifically picked in the early

757
01:13:24.760 --> 01:13:30.760
days of the field as a way to market these technologies. Like, we're repeatedly using the intelligence

758
01:13:30.760 --> 01:13:38.520
analogy and relating these machines to human intelligence as a way to try and gauge whether or not

759
01:13:38.520 --> 01:13:42.840
it is good or worthy or capable in society. I think the output is the thing that really

760
01:13:42.840 --> 01:13:46.840
Matt is the most consequential, which is like, okay, it might have a different brain, a different system,

761
01:13:46.840 --> 01:13:51.640
but it doesn't arrive at the same capability. Like, does it, is it able to do surgery on someone's

762
01:13:51.640 --> 01:13:56.680
brain? Is it able to drive a car? Like, my car drives itself in, in Los Angeles. I don't touch a

763
01:13:56.680 --> 01:14:00.680
steering wheel and I can drive for many, many hours. And in here in Austin, I just saw the ones the

764
01:14:00.680 --> 01:14:04.920
other day where they've removed the steering wheel and the pedals, the new cybercabs. So I go, it

765
01:14:04.920 --> 01:14:08.600
doesn't really matter if it's using a different system. If it's navigating through the world as a car,

766
01:14:08.600 --> 01:14:15.240
it has a better safety record than human beings. Then as far as I'm concerned, intelligence or not,

767
01:14:15.240 --> 01:14:19.960
it's like, you know, but that was not the original argument that you made, which was like,

768
01:14:19.960 --> 01:14:23.800
these systems are just generally going to become more intelligent across different things.

769
01:14:24.600 --> 01:14:28.680
Based on the prediction, this is a prediction that you're making, right? Like that. And this is

770
01:14:28.680 --> 01:14:34.440
a prediction that all the AI, it is making, Daria's making, Elon's making, Zuckerberg's making,

771
01:14:34.440 --> 01:14:38.280
almonds making, Dermis is making. And do you know what the common future of all of them is?

772
01:14:39.160 --> 01:14:44.760
They profit enormously off of this myth. Elon has recently spearheaded the construction

773
01:14:44.760 --> 01:14:50.440
of Colossus, a massive supercomputer in Memphis, housing 100,000 GPUs, specifically to scale up.

774
01:14:50.440 --> 01:14:55.080
They're growing API models faster than their competitors. It appears that they've all converged

775
01:14:55.080 --> 01:14:59.960
around this idea that you can brute force your way to greater, more generalized intelligence.

776
01:15:00.760 --> 01:15:05.800
They've converged around the idea that you can brute force your way into models that they can sell

777
01:15:05.800 --> 01:15:10.680
to people for automating certain tasks that are that are financially lucrative.

778
01:15:10.680 --> 01:15:14.200
And I heard Elon say that if you're a surgeon out, there's just no point. He was like,

779
01:15:14.200 --> 01:15:18.920
don't train to be a surgeon. He says in a couple years' time, Optimus and AI generally are

780
01:15:18.920 --> 01:15:22.600
going to be better than any surgeon that's ever lived. Do you think these things are true?

781
01:15:22.600 --> 01:15:27.400
Well, you know, I'm pretty sure it was Hinton that famously slushed and famously said,

782
01:15:27.400 --> 01:15:32.200
there would be no need for radiologists anymore. There would be no need for radiologists anymore.

783
01:15:33.000 --> 01:15:37.000
And he said a deadline that we've already passed. I don't remember how many years

784
01:15:38.600 --> 01:15:42.440
radiology is doing great as a profession. Do you think it will be in five years?

785
01:15:43.160 --> 01:15:47.560
Okay. So this once again goes back to this question of like, why do we build technology

786
01:15:47.560 --> 01:15:54.760
and why should we specifically be building AI? Okay. And for me, like the whole project of technology

787
01:15:54.760 --> 01:16:00.440
development advancement is not to advance technology for technology, say, it's to help people.

788
01:16:01.640 --> 01:16:07.480
And there have been lots of research that has shown that actually the best outcomes for people

789
01:16:07.480 --> 01:16:15.240
in a healthcare setting is for the radiologist to have the AI model in their hands

790
01:16:16.360 --> 01:16:24.840
and for the human expert to use the AI model as a tool as an input into their judgment.

791
01:16:24.840 --> 01:16:30.840
And it is that combination that leads to the most accurate and early diagnosis

792
01:16:31.400 --> 01:16:35.240
of certain types of cancer that then help improve the prognoses of the patient.

793
01:16:35.240 --> 01:16:39.720
Do you believe that in the coming years, all the cars, pretty much all the cars in the

794
01:16:39.720 --> 01:16:44.840
road will be driving themselves? No. You don't think so. How come? Because of the way the technology

795
01:16:44.840 --> 01:16:52.840
works, because these are statistical, I mean, currently the way that AI models are primarily

796
01:16:52.840 --> 01:16:57.400
developed, they're statistical engines, you have what's called a neural network, which is a

797
01:16:57.400 --> 01:17:03.720
piece of software that has a bunch of densely connected nodes and... But parameters,

798
01:17:03.800 --> 01:17:09.000
is what they call promises? Yeah, pretty much. And you're just pumping a bunch of data into it and

799
01:17:09.000 --> 01:17:14.600
then it's analyzing the data and creating all of these, finding all these correlations in the data,

800
01:17:14.600 --> 01:17:19.720
finding all these patterns. And then it's through those patterns that the machine is then able to

801
01:17:20.280 --> 01:17:24.520
act autonomously, right? And so the way that they're turning itself to our own cars,

802
01:17:25.240 --> 01:17:30.360
they're recording all this footage. And then they have tens of thousands or hundreds of thousands

803
01:17:30.360 --> 01:17:38.040
of human contractors that draw literally around every single vehicle in the footage,

804
01:17:38.600 --> 01:17:43.800
every single pedestrian, every single traffic light, every single lane marking, and label it

805
01:17:43.800 --> 01:17:50.040
exactly as such, so that then it's fed into an AI model that can identify all of these

806
01:17:50.040 --> 01:17:56.280
different components. And then it's connected to another piece of software that is not AI,

807
01:17:56.280 --> 01:18:02.360
that's saying, okay, if the AI model recognizes the pedestrian, we do not run over the pedestrian.

808
01:18:03.400 --> 01:18:12.040
If the AI model recognizes a red traffic light, we stop. And so the thing about statistical

809
01:18:12.040 --> 01:18:16.440
engines is that it's based on probabilities. It's not based on deterministic logic.

810
01:18:16.760 --> 01:18:23.000
So systems make errors all the time. And it's impossible. It is

811
01:18:23.880 --> 01:18:27.960
technically impossible to get them to stop making errors.

812
01:18:28.520 --> 01:18:32.680
Humans make errors way more than systems in this case. Yeah.

813
01:18:32.680 --> 01:18:37.640
Like the safety record is like, isn't it like 10 times more safe to be driven in a Tesla

814
01:18:37.640 --> 01:18:41.320
with autonomous driving than it is for a human to drive? It depends on the place.

815
01:18:41.320 --> 01:18:45.720
It depends on whether the Tesla was trained to specifically navigate the place that you're driving.

816
01:18:46.680 --> 01:18:54.360
Because if it's in Mumbai, in some place in Vietnam, no, it would not be safer. I would

817
01:18:54.360 --> 01:19:00.440
much rather be driven by someone that has been driving in that place their whole life.

818
01:19:00.440 --> 01:19:05.000
I'm not arguing against like the fact that in certain places where the car has been explicitly

819
01:19:05.000 --> 01:19:10.120
trained to drive in this place that it has a better safety record than the humans that are driving

820
01:19:10.200 --> 01:19:17.720
in that place. But you specifically asked if I think that all of the most cars in the world

821
01:19:17.720 --> 01:19:21.880
in the US, less than the United States, because we're here. I don't actually think that it's

822
01:19:21.880 --> 01:19:25.560
like imminently on the horizon. Ten years? No, I don't think so.

823
01:19:25.560 --> 01:19:28.600
I sat with Dara from Uber. He's pretty convinced that his 9 million

824
01:19:28.600 --> 01:19:34.280
carriers will be replaced by autonomous vehicles. I mean, how long has Delta driving cars been

825
01:19:34.760 --> 01:19:41.240
invested in Tesla? It's been more than 10 years. And what percentage of cars right now are

826
01:19:41.240 --> 01:19:47.800
autonomous on the US roads? I mean, so part of it is it's actually not a technical problem.

827
01:19:48.680 --> 01:19:52.760
Part of it is also a social problem. Do people even trust getting into these vehicles?

828
01:19:52.760 --> 01:19:58.600
Part of it's also a legal problem, which is if the self-driving car kills someone,

829
01:19:59.240 --> 01:20:03.480
which it has happened. Yeah, it has happened. Who is responsible?

830
01:20:04.280 --> 01:20:09.320
So in the case in LA, it was both Tesla and the driver, because the driver dropped their phone,

831
01:20:09.320 --> 01:20:13.800
they looked down, and this was a couple of years ago, I believe, and they went to grab their phone

832
01:20:13.800 --> 01:20:19.240
and they hit someone. And so it went to court and they were held both responsible, both the driver

833
01:20:19.240 --> 01:20:27.560
and Tesla. In terms of Tesla, pretty much everyone that gets the car, it comes with autonomy now

834
01:20:27.560 --> 01:20:31.560
for pretty much most people, I believe. Partial autonomy. Yeah, it's called full self-driving at the

835
01:20:31.560 --> 01:20:36.600
moment where it's like- I mean, yes, it is called full self-driving supervised, where you kind of

836
01:20:36.600 --> 01:20:39.960
have to be looking in the direct- you have to be looking in the right direction, but- Yes, so it's

837
01:20:39.960 --> 01:20:46.360
partial autonomy. And here in Austin, it's full autonomy, because there's no steering wheel on the

838
01:20:46.360 --> 01:20:51.400
new car, so you can't drive it anyway. But it is, you know, the model Y is the undisputed,

839
01:20:51.400 --> 01:20:57.480
higher-selling car, best-selling car in the world across all brands. Well, I guess my point here

840
01:20:57.480 --> 01:21:03.960
is, like, these predictions where they say AI is going to completely change transportation and

841
01:21:03.960 --> 01:21:07.240
driving, it's going to completely change, lawyers aren't going to have jobs, accountants aren't going

842
01:21:07.240 --> 01:21:11.800
to have jobs. Do you believe that they are true? Do you believe that there's going to be mass

843
01:21:11.800 --> 01:21:17.000
job displacement? Okay, so I do think that there is going to be huge impacts on employment,

844
01:21:17.080 --> 01:21:23.720
and we are ready seeing those impacts. It is not simply because the AI models are just automating

845
01:21:23.720 --> 01:21:30.920
those jobs away. It is specifically because the models are improving in certain capabilities

846
01:21:30.920 --> 01:21:37.240
based on what the companies that are developing them choose to improve them on. And executives at

847
01:21:37.240 --> 01:21:43.960
other companies are then deciding to fire or lay off their workers because they think that AI

848
01:21:43.960 --> 01:21:48.440
can replace the worker irrespective of whether that might be true. And there, you know,

849
01:21:48.440 --> 01:21:52.360
if there have been cases of like the Clarner CEO who laid off a bunch of people thinking that you

850
01:21:52.360 --> 01:21:56.280
would replace everyone with AI, and then it didn't actually work, and he had to ask some people to

851
01:21:56.280 --> 01:22:00.680
come back. I actually DMed him about this. If you're hearing this, this is because I've DMed Sebastian

852
01:22:00.680 --> 01:22:05.560
and he's fine with me sharing this. He said, because I've heard his name mentioned a lot. And so

853
01:22:05.560 --> 01:22:10.120
when we talked about AI in the past and people mentioned Sebastian and Clarner as the example,

854
01:22:10.120 --> 01:22:14.760
I wanted to clarify with him what the truth was. He said, it's great to hear from you. I think

855
01:22:14.760 --> 01:22:19.000
sometimes people struggle with two things can be true at the same time. I think it might be time

856
01:22:19.000 --> 01:22:24.440
to come back on your podcast. To your point, this is the medium misinterpreting my tweet. We are

857
01:22:24.440 --> 01:22:30.120
doubling down on AI more than ever. Clarner is shrinking with almost 100 employees per month due to AI.

858
01:22:30.840 --> 01:22:39.880
We used to be 7,400 at the peak a year ago, 5,500. Now we're 3,300. And by the end of summer,

859
01:22:39.880 --> 01:22:47.400
so this was last year will be 3,000 people. AI handles 70% of our customer service conversations

860
01:22:47.400 --> 01:22:52.520
at this moment. This is because we have realized that with AI, the production cost of software comes

861
01:22:52.520 --> 01:22:57.400
down to almost zero, just like manufacturing used to be all handcrafted. And then the machines came,

862
01:22:57.400 --> 01:23:02.600
code used to be all handcrafted up until a few years ago. And now it is machine produced.

863
01:23:03.320 --> 01:23:09.880
And ultimately, we pay people more than ever for the unique handcrafted man-made stuff.

864
01:23:09.880 --> 01:23:14.680
Clarner is a bank. People will want to connect to humans, not only machines. They want us to be

865
01:23:14.680 --> 01:23:21.400
personable, relatable, even flawed. So we need to make sure while we are automating, replacing

866
01:23:21.400 --> 01:23:27.080
with AI in parallel, we make sure we offer a super available human experience.

867
01:23:27.800 --> 01:23:33.080
I'm really glad you read this because I think it touches on some really important nuances to

868
01:23:34.920 --> 01:23:39.960
the AI, like the impact that AI is going to have on employment. So I think there's often

869
01:23:39.960 --> 01:23:47.160
these binary narratives. It's like AI is going to come for every job. Or people say AI is not

870
01:23:47.160 --> 01:23:52.520
actually working and it's not actually coming for jobs. And the reality is it's coming for jobs.

871
01:23:52.520 --> 01:23:58.040
There are definitely jobs that are being automated away because of the capabilities of their

872
01:23:58.040 --> 01:24:01.800
models. And there's also jobs that are being lost because executives are deciding to lay off

873
01:24:01.800 --> 01:24:05.560
the workers. Even if the models don't match the capabilities because it's good enough,

874
01:24:05.560 --> 01:24:08.920
like they would rather have the good enough model for way cheaper.

875
01:24:08.920 --> 01:24:12.600
Or they made a mistake with hiring. They blurred their team and it's a great convenience thing

876
01:24:12.600 --> 01:24:17.480
to say. Exactly. But clearly, we're already seeing impacts on the job market.

877
01:24:17.480 --> 01:24:25.160
Like the US Jobs report that came out earlier this year showed that there has been a decline

878
01:24:25.800 --> 01:24:33.320
in hiring, is a slowdown in hiring across especially white collar professional industries.

879
01:24:33.320 --> 01:24:37.160
And you saw Anthropics report, didn't you, this week? The TLDIRs, it matches kind of what you

880
01:24:37.160 --> 01:24:41.320
were saying where they anthropic looked at exactly how people were using their models.

881
01:24:41.880 --> 01:24:47.320
And they looked at what people are saying. Yeah. And they said that there's been a 40% reduction

882
01:24:47.320 --> 01:24:50.680
in an entry level jobs in particular. And then they made this graph, which has gone viral

883
01:24:50.680 --> 01:24:55.720
over the internet. The red shows where we are now in terms of capability. And based on how people

884
01:24:55.720 --> 01:25:01.080
are currently using the models, they extract it out that the blue part will be the disrupted parts.

885
01:25:01.080 --> 01:25:06.440
This is the things that they say AI can do right now. But people don't realize it yet. So if you look

886
01:25:06.440 --> 01:25:10.920
at it, it's like it's kind of all the stuff you'd expect. Yeah. It's the physical real world human

887
01:25:10.920 --> 01:25:16.280
stuff, which robots maybe can do someday like construction or agriculture that are untouched.

888
01:25:16.360 --> 01:25:22.440
But like office in admin, like saying finance stuff, math. And you notice that these are all

889
01:25:22.440 --> 01:25:28.920
of things that I just named that they purposely finance, math, law, media and art, that's me cooked.

890
01:25:28.920 --> 01:25:34.040
Yeah. So obviously, of the admin, I mean, they do focus a lot on like assistant type

891
01:25:35.000 --> 01:25:41.080
and managerial work. So, but the other thing that the carno CEO said was,

892
01:25:41.480 --> 01:25:47.880
but people also want human experiences. So it's not actually just about the capabilities

893
01:25:47.880 --> 01:25:54.200
of the models. It's also about what people want. Like some things they would turn to AI for

894
01:25:54.200 --> 01:25:59.160
and some things they wouldn't irrespective of whether or not AI is capable of doing it.

895
01:25:59.880 --> 01:26:06.920
But because of a preference that they want human to human interaction. And so what we're seeing

896
01:26:06.920 --> 01:26:13.480
right now is, yeah, the thing that happens with every wave of automation, which is that there

897
01:26:13.480 --> 01:26:19.960
is a bunch of entry-level work that gets automated away. And there are also new jobs created. But

898
01:26:19.960 --> 01:26:26.440
the jobs that are created are one in one of two categories. There are people that get even higher

899
01:26:26.440 --> 01:26:30.600
skilled jobs. And what he was saying, like we pay people more for like the handcrafted

900
01:26:31.240 --> 01:26:38.200
now. And there's also the people who get way worse jobs. And so there was this amazing article

901
01:26:38.200 --> 01:26:43.320
in New York magazine that was talking about how a lot of people are getting laid off.

902
01:26:44.200 --> 01:26:50.440
And then they end up working in data annotation, which is the labor that I've been referring to

903
01:26:50.440 --> 01:26:55.560
throughout this conversation that companies need in order to teach their models the next thing

904
01:26:55.560 --> 01:27:01.720
that the companies are trying to automate. And so like a marketer gets laid off and then they go

905
01:27:01.720 --> 01:27:08.680
and work for a data annotation firm to train the models on the very job that they were just laid off

906
01:27:08.680 --> 01:27:16.920
in, which will then perpetuate more layoffs if that model then develops that skill. And the article

907
01:27:16.920 --> 01:27:25.560
was talking about how this has become a huge catch all for a lot of people that are struggling

908
01:27:25.560 --> 01:27:31.080
with finding job opportunities right now, including like award-winning directors and Hollywood

909
01:27:31.080 --> 01:27:34.520
that are actually secretly doing this data annotation work to put food on the table.

910
01:27:35.560 --> 01:27:42.120
And so when they talk about there's going to be mass unemployment and then there's going to be

911
01:27:42.120 --> 01:27:46.760
some new jobs created that we can't even imagine. I think a lot of these narratives rarely talk

912
01:27:46.760 --> 01:27:51.800
about like first of all, why are some jobs going away? It's not just because of the model capabilities,

913
01:27:51.800 --> 01:27:55.960
also because of executive choices and because of the rhetoric that they use if they want to just

914
01:27:55.960 --> 01:28:03.000
downsize. But the other thing that is rarely talked about is the jobs, a lot of the jobs that are

915
01:28:03.000 --> 01:28:10.280
created are way worse than the jobs that were there. And it breaks the career ladder. So it's the

916
01:28:10.280 --> 01:28:16.920
entry level in the mid-tier jobs like it gouged out. It's higher order jobs and then way more

917
01:28:16.920 --> 01:28:24.840
lower order jobs that get created. And so how do people continue to progress in their careers?

918
01:28:24.840 --> 01:28:28.520
There's no more rungs on the ladder. I actually don't know the answer to this question and I've been

919
01:28:28.520 --> 01:28:31.720
furiously trying to find a good answer to this question because I can't, you know,

920
01:28:32.360 --> 01:28:37.400
everything is theory and for my audience, I would say most of my audience don't run businesses.

921
01:28:37.400 --> 01:28:41.400
A lot of them do a lot of them spy too, but they don't run businesses. So they're also in the

922
01:28:41.400 --> 01:28:44.920
land of theory. They're hearing lots of different things. Jack Dorsey does his tweets saying he's

923
01:28:44.920 --> 01:28:48.840
halving his head count because of AI. They don't know what's true. They don't know the internal

924
01:28:48.840 --> 01:28:52.840
economics at Jack's company and did he bloat the company during the pandemic and he's just using

925
01:28:52.840 --> 01:28:56.920
this as an excuse to make this share price spike seven points because his investors now think

926
01:28:56.920 --> 01:29:01.800
they're an AI company, whatever. It's hard to pass through. So eventually I go, okay, what am I doing?

927
01:29:02.680 --> 01:29:07.240
I have hundreds of team members, probably 70 companies I invest in, maybe five or six that I'm

928
01:29:07.240 --> 01:29:12.040
like the lead shareholder in. What am I actually doing on a daily basis right now? I'm also,

929
01:29:12.040 --> 01:29:17.160
I also consider myself the head of recruitment, but in the last month in particular, I have met

930
01:29:17.160 --> 01:29:21.960
extremely capable candidates in terms of cultural alignment, hard work, those kinds of things,

931
01:29:21.960 --> 01:29:26.520
but I've had to take a great deal of pause because when I run the experiment of can I get an AI agent

932
01:29:26.600 --> 01:29:32.200
to do that exact same thing? The answer is increasingly us, especially in a world of open

933
01:29:32.200 --> 01:29:39.560
clause. And so what I'm curious, like, yeah, now you confront this decision where you're seeing

934
01:29:39.560 --> 01:29:46.920
in this short term period, you could just choose the AI agent. And in the long term period,

935
01:29:48.040 --> 01:29:55.160
there is no career ladder. So, so who are you promoting into these senior roles? Like, how do you

936
01:29:55.160 --> 01:29:58.440
resolve it for your own company? Yeah, it's a good question. So there's kind of two ways I'm

937
01:29:58.440 --> 01:30:03.000
thinking about it. I think really deep expertise is very, very valuable because if you're now the

938
01:30:03.000 --> 01:30:08.520
orchestrator of potentially AI agents, it's really about having a deep understanding of the right

939
01:30:08.520 --> 01:30:12.920
question to ask. And that's someone who has deep expertise on something. So I need my CFO,

940
01:30:13.560 --> 01:30:17.800
because if she's going to be orchestrating our team of agents that might be doing financial analysis

941
01:30:17.800 --> 01:30:24.520
or whatever else, she needs to understand what to tell them to do in our company. And in turn,

942
01:30:24.520 --> 01:30:29.400
financial analysts can't do that. They need the 50 odd years of experience that Claire has.

943
01:30:30.040 --> 01:30:36.440
On the other end, I need CAS. CAS is 25. CAS knows everything about AI agents. He's a young

944
01:30:36.440 --> 01:30:41.160
Japanese kid who's highly, highly curious. On the weekend, he's building AI agents to solve

945
01:30:41.160 --> 01:30:46.520
problems in my life. I need those two kinds of thinking, which is highly proficient agent

946
01:30:46.520 --> 01:30:51.000
maxing young kids, or they don't necessarily need to be young, but really lean in high curiosity,

947
01:30:51.000 --> 01:30:54.920
that's creating a force model to plan my business and then any deep expertise. Now,

948
01:30:54.920 --> 01:31:00.040
everything else outside of there is another one I thought of another group is like people with

949
01:31:00.040 --> 01:31:06.840
extremely great IRL people skills. Because we do meet people in real life. We greet you when

950
01:31:06.840 --> 01:31:11.240
you arrive here. We greet when we go for lunch with big clients that we have, whether it's Apple

951
01:31:11.240 --> 01:31:18.200
or LinkedIn or whoever it might be, we need to smooze. And we have teams who are in person in

952
01:31:18.280 --> 01:31:22.360
the office, so we do a lot of stuff IRL. And increasingly, we're building communities even for this

953
01:31:22.360 --> 01:31:25.560
show. We're doing community events all around the world, so many people that are good at that as

954
01:31:25.560 --> 01:31:30.440
well. IRL bringing people together in real life and organizing stuff. Those are the three groups of

955
01:31:30.440 --> 01:31:38.840
people that I'm like, you know, irreplaceable right now. And if you were to all the roles that

956
01:31:38.840 --> 01:31:42.440
could be done by AI agents, if we were to place them with AI agents, do you think you would still

957
01:31:42.440 --> 01:31:48.600
have these three roles, pools of people to hire and promote into the three critical things that

958
01:31:48.600 --> 01:31:55.880
you need in the long term? If things carry on at the current rate of trajectory, one could assert

959
01:31:55.880 --> 01:32:00.440
that even those roles would experience pressure. If you just imagine it like people think of things

960
01:32:00.440 --> 01:32:04.600
either statically or linearly or exponentially, you imagine an exponential rate of improvement,

961
01:32:04.600 --> 01:32:09.000
which is kind of what I've seen, even like a 10% compounding rate of improvement. At some point,

962
01:32:12.920 --> 01:32:19.240
at some point, I think what remains is actually the IRL irreplaceably human stuff,

963
01:32:19.240 --> 01:32:23.960
human to human. Our Maslowian needs of being in person like we are now aren't going to change. We

964
01:32:23.960 --> 01:32:28.680
need connection. Humans get very sick when they don't have other human beings in their life and

965
01:32:28.680 --> 01:32:34.120
strong deep relationships. So that stuff is going to matter a whole lot. I have this contrarian

966
01:32:34.120 --> 01:32:38.520
weird take that actually maybe this is the first technology that's going to deliver on the promise

967
01:32:38.520 --> 01:32:42.200
of making us human and connected because we're going to be rendered useless at everything else

968
01:32:42.200 --> 01:32:45.640
other than what humans are good at. Because all the other technology said, oh, we're going to make

969
01:32:45.640 --> 01:32:49.720
you more connected, connecting the world and they disconnected the world and isolated the world.

970
01:32:49.720 --> 01:32:53.000
But maybe this is the one it's so intelligent now that it doesn't need us to fuck around in

971
01:32:53.000 --> 01:32:59.880
spreadsheets anymore. Do you see that actually happening in real time right now that it's making us

972
01:32:59.880 --> 01:33:07.400
more able to be in person connected with one other, having deeper social community engagements?

973
01:33:08.680 --> 01:33:14.040
Yes. I'll give you some data points. Okay. Data point number one. The financial times release the

974
01:33:14.040 --> 01:33:20.760
report on social media usage and what they saw is 2022 is the peak and it's plateaued ever since.

975
01:33:20.760 --> 01:33:25.400
The generation that's plateaued the fastest and heading down is the younger generations. The

976
01:33:25.400 --> 01:33:30.680
boomers are still off to the races, right? On Facebook and stuff. And then you look at the way

977
01:33:30.680 --> 01:33:35.720
Jennifer are using social media. They're not posting as much. They call it posting zero. They're

978
01:33:35.720 --> 01:33:39.000
scrolling sometimes, but they're in dark social environments like WhatsApp and Snapchat and I

979
01:33:39.000 --> 01:33:43.400
message. They're not like performing to the world. They also value IRL experience as much more

980
01:33:43.400 --> 01:33:47.720
than any other generation. They're like not getting smashed. We're seeing every brand has a run club.

981
01:33:47.720 --> 01:33:52.440
We're seeing, I mean, run clubs explode exploding around the world and we're seeing this real sort of

982
01:33:53.640 --> 01:33:59.960
sort of almost like innate realization that like technology led us down at some fundamental level.

983
01:33:59.960 --> 01:34:04.200
Like dating apps led us down. Social networking kind of has led us down. And we're seeing I think

984
01:34:04.920 --> 01:34:08.200
maybe a biification of society where a lot of people are going, fact this, like I want to go

985
01:34:08.200 --> 01:34:12.920
back to what it is to be a human. And I would imagine that in such a world where intelligence

986
01:34:12.920 --> 01:34:18.040
is so sophisticated that we no longer needed to sit at laptops. I'm like, I think screen time's

987
01:34:18.040 --> 01:34:20.440
going to continue to fall. I think you go into an office. You're not going to see people

988
01:34:20.440 --> 01:34:25.080
sit at laptops. You're going to see something completely different. And I think maybe,

989
01:34:25.960 --> 01:34:29.720
you know, and then we talk about robots and Optimus robots. Elon says there'll be 10 billion

990
01:34:29.720 --> 01:34:36.600
Optimus robots. Elon has been wrong with timing before. He's almost never been wrong on the big

991
01:34:36.600 --> 01:34:43.160
things completely. He's just, his timing has got a bad track record. So I think he's probably

992
01:34:43.160 --> 01:34:47.240
right. You know, I think I've got some people on the way from Boston Dynamics and these other big

993
01:34:47.240 --> 01:34:51.080
companies like Scale AI and they're actually bringing the robots here to show it like folding

994
01:34:51.080 --> 01:34:54.760
laundry during the dishes. I'm not saying that's what I would want in my home, but I think factory

995
01:34:54.760 --> 01:34:57.960
work is going to completely change. I think a lot of manual labor is going to completely change.

996
01:34:58.440 --> 01:35:04.440
I think we're going to be forced to do what only we can do. Sebastian, who's the CEO of Klanah,

997
01:35:04.440 --> 01:35:12.440
has actually just called me. Hello, Sebastian, you're right. I'm good. How are you?

998
01:35:14.040 --> 01:35:17.320
A little while. It has been a while since you're on the show. I was just saying we do need to get

999
01:35:17.320 --> 01:35:22.360
you back on. I just had a couple of simple questions because, you know, I do a lot of interviews and

1000
01:35:23.080 --> 01:35:26.840
Klanah's always mentioned because I think the media has said that you like double down an

1001
01:35:27.000 --> 01:35:30.840
year than you reversed because it didn't work out. So I know I spoke to you a while ago and we

1002
01:35:30.840 --> 01:35:34.440
exchanged a couple of DMs about it, but that was more than, it was almost a year ago now.

1003
01:35:35.000 --> 01:35:39.880
So I just wanted to get an update on Klanah's business, AI agents and all of that, if possible.

1004
01:35:39.880 --> 01:35:46.360
First and foremost, we were early on released AI to support our customer service, which had that

1005
01:35:47.160 --> 01:35:53.880
initial benefit of more calls being dealt with by AI, which customers liked because those calls

1006
01:35:53.880 --> 01:35:59.160
or chat messages were much, much faster and more qualitative. Then since then, that has actually

1007
01:35:59.160 --> 01:36:05.480
expanded slightly. What we did, however, try to communicate as well is that we believed in a world

1008
01:36:05.480 --> 01:36:12.440
of where AI is cheap and available. The value of human interaction will be regarded as higher.

1009
01:36:12.440 --> 01:36:19.720
So the future of customer service VIP is a human. We have then hence doubled down on providing more

1010
01:36:19.800 --> 01:36:26.120
that. But at the same time, the efficiency gains within the company has continued. I mean, we used to

1011
01:36:26.120 --> 01:36:33.320
be about 6,000 people and now we are less than 3,000, which is two, three years since we stopped

1012
01:36:33.320 --> 01:36:39.400
recruiting. And at the same time, our revenue has doubled. So you can clearly see that AI has

1013
01:36:39.400 --> 01:36:46.920
allowed us to do more with less people, but we have avoided layoffs instead relied on natural

1014
01:36:46.920 --> 01:36:52.920
nutrition when people kind of move on to other jobs. I mean, from my perspective, we will continue

1015
01:36:52.920 --> 01:36:57.560
to be very, you know, not really recruit much of me. We recruit a little bit here and there,

1016
01:36:57.560 --> 01:37:03.320
but we expect that kind of natural nutrition of 10, 15 percent per year to continue

1017
01:37:04.040 --> 01:37:09.320
to become fewer. I think the big breakthrough was really in November, December, last year,

1018
01:37:09.320 --> 01:37:17.640
where even the kind of more most skeptical engineers who are like very well renowned and appreciated

1019
01:37:17.640 --> 01:37:22.840
like the founder of Linux and stuff like that basically said that coding has now been resolved.

1020
01:37:23.400 --> 01:37:28.440
And hence is not, you know, you don't need to code anymore. And that was kind of a common sentiment.

1021
01:37:28.440 --> 01:37:34.600
So I think in coding, that's definitely an engineering work that has been a tremendous shift

1022
01:37:34.600 --> 01:37:40.680
in the last six months. What do all these people go do? Sebastian? I am optimistic. I mean,

1023
01:37:40.680 --> 01:37:45.720
I think obviously people will have a lot of opinions about this topic, but I still believe that

1024
01:37:45.720 --> 01:37:52.920
we are going to move towards a richer society. Now in the short term, there could be more worry

1025
01:37:52.920 --> 01:37:58.520
about what happens if people don't get a job and so forth. But I think in the longer term, I am

1026
01:37:58.520 --> 01:38:03.720
optimistic what it means for society and humanity. Thank you so much, Seb. I'll chat to you soon.

1027
01:38:03.720 --> 01:38:05.160
Thank you for taking the time. I appreciate you, mate. Thanks.

1028
01:38:09.160 --> 01:38:13.880
You know the little traditional SIM card that goes inside of our phones. They haven't changed at all

1029
01:38:13.880 --> 01:38:18.440
since they were invented in the 90s. You have this physical piece of plastic that means you're locked

1030
01:38:18.440 --> 01:38:24.520
into one carrier, one network, and the second you cross a border that carrier can start charging you

1031
01:38:24.520 --> 01:38:28.760
whatever they want. But there are alternatives. And today's sponsor, Sally, is one of them.

1032
01:38:28.840 --> 01:38:35.160
It's an eSIM app that gives you a safe and secure data connection in over 200 destinations.

1033
01:38:35.160 --> 01:38:39.160
All of their eSIMs have built in cybersecurity, which is great if you're traveling for work and

1034
01:38:39.160 --> 01:38:43.560
looking at confidential material. I've been using Sally whenever I travel because the connection

1035
01:38:43.560 --> 01:38:48.760
is always reliable and it saves me a ton of roaming fees. It also means I don't have to deal with all

1036
01:38:48.760 --> 01:38:52.840
of the faths that surround sorting out a SIM everywhere I go. If you want to give it a try,

1037
01:38:52.840 --> 01:38:58.360
download the Sally app from the app store now and scan the QR code on screen. And if you want 15%

1038
01:38:58.360 --> 01:39:07.320
off your first purchase, use my code, D-O-A-C, when you get to check out, that's D-O-A-C for 15% off.

1039
01:39:09.240 --> 01:39:13.560
So this is something that I've made for you. I've realised that the driver's here audience are

1040
01:39:13.560 --> 01:39:18.120
strivers, whether it's in business or health. We all have big goals that we want to accomplish.

1041
01:39:18.120 --> 01:39:23.400
And one of the things I've learnt is that when you aim at the big, big, big goal, it can feel

1042
01:39:23.400 --> 01:39:28.520
incredibly psychologically uncomfortable because it's kind of like being stood at the foot of

1043
01:39:28.520 --> 01:39:33.560
Mount Everest and looking upwards. The way to accomplish your goals is by breaking them down into

1044
01:39:33.560 --> 01:39:39.400
tiny small steps and we call this an R team the 1%, and actually this philosophy is highly

1045
01:39:39.400 --> 01:39:44.760
responsible for much of our success here. So what we've done so that you at home can accomplish any

1046
01:39:44.760 --> 01:39:50.200
big goal that you have is we've made these 1% diaries and we've released these last year and

1047
01:39:50.200 --> 01:39:54.760
they all sold out so I asked my team over and over again to bring the diaries back but also to

1048
01:39:54.760 --> 01:40:00.680
introduce some new colours and to make some minor tweaks to the diaries. So now we have a better

1049
01:40:00.680 --> 01:40:06.360
range for you. So if you have a big goal in mind and you need a framework and a process and some

1050
01:40:06.360 --> 01:40:11.560
motivation that I highly recommend you get one of these diaries before they all sell out once

1051
01:40:11.560 --> 01:40:16.520
again. And you can get yours at the diary.com and if you want the link, the link is in the description

1052
01:40:16.680 --> 01:40:24.200
below. Any thoughts? Well I actually had thoughts on something that you said before he called

1053
01:40:24.920 --> 01:40:29.400
which is you were saying that the Gen Z years like there's this trends that they're actually

1054
01:40:29.400 --> 01:40:33.640
disconnecting from technology so they're becoming more in person and then there's this other

1055
01:40:33.640 --> 01:40:36.920
class of workers that are actually leaning into the technology but then becoming more human

1056
01:40:36.920 --> 01:40:42.200
because they're leaning into the technology because they're realizing that they should actually

1057
01:40:42.200 --> 01:40:47.480
just be spending more time doing in person to person interactions rather than steering on a

1058
01:40:47.480 --> 01:40:51.080
screen and so they're no longer doing the typing and whatever. I really want to go back to this

1059
01:40:51.080 --> 01:40:57.480
New York magazine piece that just came out because what you're describing is true for very specific

1060
01:40:57.480 --> 01:41:02.840
category of people which is often like the business owners and leadership within companies that

1061
01:41:02.840 --> 01:41:08.440
actually can make these decisions on how they spend their time and what they ultimately do with

1062
01:41:08.440 --> 01:41:16.920
their time but what the piece talks about is the working class like people like people who are not

1063
01:41:16.920 --> 01:41:24.840
business owners that are then having to experience being laid off and then working for the data

1064
01:41:24.840 --> 01:41:31.800
annotation industry which is now one of the top jobs on LinkedIn by the way. The yeah so LinkedIn had

1065
01:41:31.800 --> 01:41:39.080
a report that showed the top 10 jobs with the highest growth in the last year and data annotation

1066
01:41:39.080 --> 01:41:44.840
is on that list. And for anyone who doesn't know what data annotation is. Yeah so data annotation is

1067
01:41:44.840 --> 01:41:53.320
the process of teaching these chatbots or any AI system to do what they ultimately are able to do.

1068
01:41:53.320 --> 01:41:57.880
So the fact that chat GPT can chat is because there were tens of thousands or hundreds of thousands of

1069
01:41:57.880 --> 01:42:04.040
people that were literally typing into a large language model and showing it this is how you're

1070
01:42:04.040 --> 01:42:10.680
supposed to then respond when a user types in a prompt like this. Before they did that work,

1071
01:42:11.240 --> 01:42:16.200
chat GPT didn't exist like it just it would just you would prompt the model and the model would

1072
01:42:16.200 --> 01:42:21.080
generate some text that was not in dialogue with the person it would kind of generate something

1073
01:42:21.080 --> 01:42:24.840
that was adjacently related. Is this what they call reinforcement learning where you kind of you

1074
01:42:24.840 --> 01:42:29.080
give it like? It's a part of the process of reinforcement learning so you do data annotation which

1075
01:42:29.080 --> 01:42:36.920
is literally showing lots of different you know examples of things that you want the model to know

1076
01:42:36.920 --> 01:42:41.640
and then reinforcement learning is getting the model to then train on those examples iteratively

1077
01:42:41.640 --> 01:42:48.120
in a way that then gives the model some of those capabilities. And what the New York magazine

1078
01:42:48.120 --> 01:42:54.200
piece highlighted is many many of the people that are getting laid off now or are struggling to

1079
01:42:54.200 --> 01:42:59.960
find work and these are highly educated people. They're college graduates, PhD graduates,

1080
01:42:59.960 --> 01:43:06.680
law degree graduates, doctors and again like award winning directors that are that are then

1081
01:43:07.320 --> 01:43:11.800
struggling to find employment in the economy because the economy has been very much

1082
01:43:11.800 --> 01:43:19.080
restructured by AI. They are then finding themselves being serving this industry and the industry

1083
01:43:19.080 --> 01:43:26.360
is designed in a way that is extremely inhumane because what the companies the companies that

1084
01:43:26.360 --> 01:43:30.680
use these data annotation services like there's these third party providers that are data annotation

1085
01:43:30.680 --> 01:43:39.800
firms. An open AI, a GROC, a Google they will hire these firms to then find the workers to perform

1086
01:43:39.800 --> 01:43:46.280
the data annotation tasks that they need. For these firms these third party firms they are incentivized

1087
01:43:46.280 --> 01:43:52.360
to pit workers against each other because they want this data annotation to happen at speed

1088
01:43:52.360 --> 01:43:57.320
and as cheaply as possible so that they can also compete with one another in this middle layer

1089
01:43:57.320 --> 01:44:05.720
to get the the bit the the contract from the client. And so all of these workers that were

1090
01:44:05.720 --> 01:44:11.240
interviewed for this New York magazine story talk about how they actually no longer have an ability

1091
01:44:11.320 --> 01:44:18.840
to be human because they are waiting at their laptop to be pinged on slack for when a project is

1092
01:44:18.840 --> 01:44:23.160
going to open up for data annotation because they've tried job hunting they literally can't find

1093
01:44:23.160 --> 01:44:27.080
anything else. This is the thing that's going to help them put food on the table for their kids

1094
01:44:27.080 --> 01:44:33.320
and there was this one woman who said like I have so much anxiety about when the project is going

1095
01:44:33.320 --> 01:44:38.680
to come when it's going to leave that when the project came it was right when my kid was coming

1096
01:44:39.400 --> 01:44:44.280
off of school and I just started tasking furiously because I don't know what's going to go and I

1097
01:44:44.280 --> 01:44:48.920
need to earn as much money as possible in this window of opportunity so then my when my kid came

1098
01:44:48.920 --> 01:44:56.520
home and tried to talk to me I screamed at my child for for distracting me and then she was like

1099
01:44:56.520 --> 01:45:03.960
I've become a monster and I am not even allowed to go to the bathroom or take care of my kids let alone

1100
01:45:04.120 --> 01:45:10.760
myself because this industry that is absorbing more and more of the workers that are being

1101
01:45:10.760 --> 01:45:20.920
laid off is mechanizing my life atomizing my work devaluing my expertise and then harvesting it

1102
01:45:21.640 --> 01:45:27.160
for the perpetuation of this machine that all of these AI executives are saying is then going to

1103
01:45:27.160 --> 01:45:33.800
come for everyone else's jobs and so what you were saying about these this class of workers

1104
01:45:34.920 --> 01:45:40.280
the business owners that get to become more human because there are all of these AI models

1105
01:45:40.280 --> 01:45:46.760
now doing the task that they don't have to do anymore it is at the cost of the vast majority of

1106
01:45:46.760 --> 01:45:52.440
people who are not business owners that are struggling to find work getting absorbed into

1107
01:45:52.440 --> 01:45:58.920
the work of then providing these technologies that the business owners can use and instead of

1108
01:45:58.920 --> 01:46:05.480
becoming more human they feel like their humanity has been squeezed and diminished and

1109
01:46:06.680 --> 01:46:12.440
they have no ability to have control agency and dignity in their lives anymore I think this is

1110
01:46:12.440 --> 01:46:17.240
a big I think this is a big question that kind of pertains to this graph here which is you know all

1111
01:46:17.320 --> 01:46:22.760
of these people if we believe anthropics prediction of who will be disrupted these people in these

1112
01:46:22.760 --> 01:46:29.800
industries like arts media legal and life and social sciences architecture and engineering

1113
01:46:29.800 --> 01:46:36.040
computer and maths business and finance and management and also office and admin these people if

1114
01:46:36.040 --> 01:46:40.600
we believe this would have to retrain at something else and unlike the industrial revolution where you

1115
01:46:40.600 --> 01:46:45.480
might get 10 20 years to retrain because factories take a long time to build the distribution layer

1116
01:46:45.480 --> 01:46:50.440
that AI sits on top of his the open internet so this is why I chat you can go pop and get hundreds

1117
01:46:50.440 --> 01:46:53.880
of millions of users in no time at all and become the fastest growing company of all time

1118
01:46:55.160 --> 01:47:00.600
one of my fears is that this disruption takes place at a speed where we can't transition

1119
01:47:01.720 --> 01:47:07.960
and that was you know that I think you you set that sentence in the passive voice the transition

1120
01:47:07.960 --> 01:47:15.400
would happen at a speed but who is driving that speed it's the companies the companies and their

1121
01:47:15.400 --> 01:47:21.560
race with one another yeah and so they are driving the transition to happen at a speed at which

1122
01:47:22.440 --> 01:47:29.640
it would be really hard to take care of all of the people that would be bulldozed over by this is

1123
01:47:29.640 --> 01:47:33.080
one of the crazy questions that no one can answer for me when I sit with these people that are

1124
01:47:33.080 --> 01:47:37.080
AICO so I go so what happens to the people if this is if you agree that this is going to happen at

1125
01:47:37.080 --> 01:47:41.960
super speed you know I spoke to that CEO of Uber Dara who said very similar things to what you're

1126
01:47:41.960 --> 01:47:47.560
saying is you know they'll be data labeling jobs for example for the drivers but they can't all

1127
01:47:47.560 --> 01:47:52.280
become data labelers and there's a question around meaning and purpose and fulfillment and that comes

1128
01:47:52.280 --> 01:47:57.960
from losing your meaning in life I sit also sit here with so many people who talk about how their

1129
01:47:57.960 --> 01:48:03.000
father lost their job being Iran or some some other country and came to the United States and had

1130
01:48:03.000 --> 01:48:08.280
to be a toilet cleaner on particular case was a doctor in Iran but came to the US and was a

1131
01:48:08.280 --> 01:48:13.160
toilet cleaner and had to deal with the sense of shame that that particular person fell in the

1132
01:48:13.160 --> 01:48:17.080
lack of dignity that that caused and how that made that person self esteem feel in the depression

1133
01:48:17.080 --> 01:48:23.720
alcoholism that transpired from that if this happens at a large scale across society there's going

1134
01:48:23.720 --> 01:48:28.840
to be a ton of consequences like that I mean this is this is like the core themes of my work and

1135
01:48:28.840 --> 01:48:33.560
the reason why I'm critical of these companies is that they are creating technologies in a way

1136
01:48:34.120 --> 01:48:41.080
that creates the haves and have nots in an extreme form that we have that it's it's exacerbating

1137
01:48:41.080 --> 01:48:47.400
the inequality that we already see in the world like the people who have things will have way

1138
01:48:47.400 --> 01:48:52.360
more riches they'll have way more free time they'll be allowed to be more human but the people who

1139
01:48:52.360 --> 01:49:01.400
don't have things are even being squeezed even more and it's not just from a work perspective I

1140
01:49:01.400 --> 01:49:07.560
mean I talk in my book also about the environmental and public health crisis that these companies have

1141
01:49:07.560 --> 01:49:17.400
created where they are building these colossal super computer facilities and and in communities

1142
01:49:17.480 --> 01:49:22.200
like communities all around the world and they specifically pick some of the most vulnerable

1143
01:49:22.200 --> 01:49:28.120
communities we're sitting in Texas right now open AI's largest one of its largest data center

1144
01:49:28.120 --> 01:49:33.480
projects is being built in Abelian Texas as part of the Stargate initiative which was an effort

1145
01:49:33.480 --> 01:49:38.360
announced at the beginning of Trump's second administration to spend 500 billion dollars on AI

1146
01:49:38.360 --> 01:49:46.520
computing infrastructure this facility consumes will when it's finished will consume more than a

1147
01:49:46.520 --> 01:49:53.960
gigawatt of power which is over 20 percent yeah over 20 percent so this is actually a little bit

1148
01:49:53.960 --> 01:49:59.080
inaccurate now this was something that circulated online for a while but there's updated numbers

1149
01:49:59.080 --> 01:50:03.320
just for someone that can't see because they're listening to Spotify or something it's a picture

1150
01:50:03.320 --> 01:50:09.640
of the size of this facility so this is not the Abelian Texas one this is a meta facility

1151
01:50:09.640 --> 01:50:14.760
so let's just talk about opening our facility in Texas that one would be the size of central park

1152
01:50:15.720 --> 01:50:23.480
and it would run a million computer chips and it would require the power of more than 20 percent

1153
01:50:23.480 --> 01:50:28.600
of New York City do you know one of the things which I found confusing so I'd like to like

1154
01:50:28.600 --> 01:50:33.000
alleviate the distance is I thought you were saying earlier that you didn't think the job disruption

1155
01:50:33.000 --> 01:50:43.240
promises were real no what I was saying is that when we talk about what these executives predict

1156
01:50:43.240 --> 01:50:49.640
about the future we need to understand that they are ultimately trying to influence the public

1157
01:50:49.640 --> 01:50:53.640
in a way that allows them to continue maintaining control over the technology so

1158
01:50:53.640 --> 01:50:57.240
objectively do you think that the job disruption that they talk about yeah I mean I

1159
01:50:57.240 --> 01:51:03.560
mean I mentioned real well I don't want to comment specifically on like this chart but it's like

1160
01:51:03.560 --> 01:51:08.040
we've already seen in job reports that there is a restructuring of the economy happening right now

1161
01:51:08.840 --> 01:51:14.040
but but going back to like the data so this super computer facility it's a meta super computer

1162
01:51:14.040 --> 01:51:21.240
facility is being built in Louisiana and it would be four times the size of the Abelian Texas one

1163
01:51:21.880 --> 01:51:26.520
and use half of the average power demand of New York City so it's one fifth the size of Manhattan

1164
01:51:26.520 --> 01:51:31.000
this makes it seem like almost all of Manhattan but it's it would be one fifth the size of Manhattan

1165
01:51:31.080 --> 01:51:38.760
when these facilities go into these communities what happens power utility increases grid reliability

1166
01:51:38.760 --> 01:51:45.880
decreases the facilities also need fresh water to generate the power for powering them as well as

1167
01:51:45.880 --> 01:51:51.240
fresh water to cool and there have been lots of documented stories of communities that are already

1168
01:51:51.240 --> 01:51:55.400
really constrained in their fresh water resource they're under a drought when a facility comes in

1169
01:51:55.400 --> 01:52:00.360
and then there are people the community is actually like competing with this facility for fresh water

1170
01:52:00.360 --> 01:52:06.200
I talk about one of those communities in my book and also sometimes these facilities instead of

1171
01:52:06.200 --> 01:52:13.240
connecting to the grid they instead of a power plant pops up next to it so in Memphis time to see

1172
01:52:13.240 --> 01:52:21.080
where Mosque built Colossus the super computer for training Grock he used 35 methane gas turbines

1173
01:52:21.080 --> 01:52:27.000
to power the facility this is a working class community a black and brown community a rural community

1174
01:52:27.000 --> 01:52:34.040
that was not even told that they would be the hosts of this facility and they discovered it because

1175
01:52:34.040 --> 01:52:40.360
they literally smelled what seemed like a gas leak in all of their living rooms and that's when

1176
01:52:40.360 --> 01:52:48.040
they discovered that these methane gas turbines were taking away their right to clean air and this

1177
01:52:48.040 --> 01:52:54.040
is a community that's already been facing a history of environmental racism they had already had

1178
01:52:54.040 --> 01:53:01.080
lots of struggles to access their right to clean air and now there's this huge super computer

1179
01:53:01.080 --> 01:53:07.880
that's landed in their midst that is pumping thousands of tons of toxins into their air

1180
01:53:07.880 --> 01:53:14.200
exacerbating the asthmatic symptoms of the children exacerbating the respiratory illnesses of other

1181
01:53:14.200 --> 01:53:22.360
people that it's one of the communities that has the highest rates of lung cancer and so

1182
01:53:22.440 --> 01:53:27.240
and the super computers taking their jobs and then they also have super computers taking their

1183
01:53:27.240 --> 01:53:32.840
jobs so so this is what I mean it's like the halves and have-nots are fundamentally

1184
01:53:33.640 --> 01:53:41.640
being pulled apart even further like if you in this version of Silicon Valley's future are in

1185
01:53:41.640 --> 01:53:49.560
the misfortunate category of being a have-not we are talking about you now getting a job that is

1186
01:53:49.560 --> 01:53:55.560
way worse than what you had because you might be doing data annotation and you might be treated

1187
01:53:55.560 --> 01:54:00.920
as a machine rather than as a human to extract value the value of your labor for perpetuating

1188
01:54:00.920 --> 01:54:07.160
this labor automating machine that these people are building you might be competing with these

1189
01:54:07.160 --> 01:54:13.400
facilities for fresh water resources they're also polluting your air your bills have increased

1190
01:54:13.400 --> 01:54:21.480
so the affordability crisis is getting worse like how is that making people able to be more human

1191
01:54:21.480 --> 01:54:30.280
what do we do about it yes okay so one of the analogies that I always use is AI is like the word

1192
01:54:30.280 --> 01:54:35.160
transportation transportation can literally refer to everything from a bicycle to a rocket

1193
01:54:35.800 --> 01:54:41.560
and we have nuanced conversations about transportation where we always say we need a transition

1194
01:54:41.640 --> 01:54:48.040
our transportation towards more sustainable options we need a transition towards you know

1195
01:54:48.040 --> 01:54:55.080
public transport electric vehicles and we know we don't ever say everyone should get a rocket to do

1196
01:54:55.080 --> 01:55:00.440
every to serve all of the transportation needs right like we're in Austin if you use the rocket

1197
01:55:00.440 --> 01:55:05.560
to fly from Dallas to Austin like that would just make knots no sense it's just a disproportionate

1198
01:55:05.560 --> 01:55:12.200
use of resources to get the benefit of getting from point A to point B this how we should think

1199
01:55:12.200 --> 01:55:18.200
about AI so all of the models that we've been talking about I like to think of them as the rockets

1200
01:55:18.200 --> 01:55:24.200
of AI they use an extraordinary amount of resources and they provide benefit some dramatic benefit

1201
01:55:24.200 --> 01:55:30.040
to some people but they're also great exacting an extraordinary cost on a large swap of people

1202
01:55:30.040 --> 01:55:39.240
because of the like the cost of developing this technology why don't we build more bicycles

1203
01:55:39.240 --> 01:55:47.240
of AI this is things like deep minds alpha fold which is a system that predicts how proteins

1204
01:55:47.240 --> 01:55:53.080
will fold based on amino acid sequences is really important for accelerating drug discovery for

1205
01:55:54.440 --> 01:55:59.640
understanding human disease and it won the Nobel Prize in chemistry in 2024 and the reason why

1206
01:55:59.640 --> 01:56:06.680
it's a bicycle of AI is because you're using small curated datasets you're just you just have data

1207
01:56:06.680 --> 01:56:13.880
that has amino acid sequences and protein folding so that means you need significantly less

1208
01:56:14.600 --> 01:56:18.760
computational resources to develop the system which means significantly less energy which means

1209
01:56:18.760 --> 01:56:24.920
less emissions so on and so forth and you're providing enormous benefit to people it feels like

1210
01:56:25.000 --> 01:56:31.640
the horses left the stable in this regard because they've already taken people's IP they've taken

1211
01:56:31.640 --> 01:56:36.840
media they they train on this podcast we know they do because it shows that they do I think there's

1212
01:56:36.840 --> 01:56:40.680
a button actually the backend of YouTube now that allows you just to click it and it says we will

1213
01:56:40.680 --> 01:56:46.680
train on your YouTube channel so the horse is kind of like if the horse truly had left the

1214
01:56:46.680 --> 01:56:52.760
stables they wouldn't have to train on anything anymore why is it that their appetite for data has

1215
01:56:52.760 --> 01:56:58.120
actually expanded it's because in order to build the next generations of their technologies in

1216
01:56:58.120 --> 01:57:05.800
order to have the technologies continue to be relevant and continue to update with the pace of new

1217
01:57:05.800 --> 01:57:12.200
knowledge creation and societies involvement they need to train again and again and again and again

1218
01:57:12.200 --> 01:57:17.800
and why are they employing actually more and more and more data annotation workers over time

1219
01:57:17.880 --> 01:57:24.120
it's because they need more of that work over time I mean I've been reporting on the data

1220
01:57:24.120 --> 01:57:30.520
annotation work for over seven years now and it's not gone down it's gone it's increased

1221
01:57:30.520 --> 01:57:35.160
do you think there's any chance of it going down do you think there's any chance of this sort of

1222
01:57:35.160 --> 01:57:41.880
brute force scaling approach where you take data you take computational power energy and you you know

1223
01:57:41.880 --> 01:57:48.040
you have the data lablers and you know building out more and more parameters for the models

1224
01:57:48.040 --> 01:57:51.400
do you think there's any chance it's going to stop or go in a different direction other than

1225
01:57:51.400 --> 01:57:56.360
the one that's going in now I would love to reframe the question and say what should we be doing

1226
01:57:56.360 --> 01:58:02.040
in this moment where it's not going down where we do recognize that actually these companies

1227
01:58:02.040 --> 01:58:08.680
in this moment need continued resources inputs and labor to perpetuate what they are doing yeah

1228
01:58:08.840 --> 01:58:14.920
because this sounds like stop and I just feel like stop is like a hard it feels like I just think

1229
01:58:14.920 --> 01:58:18.760
you know with the government in place they're supporting these companies like crazy globally this

1230
01:58:18.760 --> 01:58:24.120
is happening some like stop doesn't feel I always say we need to break up the empire and we need

1231
01:58:24.120 --> 01:58:30.920
to develop alternatives and we are already seeing a flourishing of incredible grassroots movements

1232
01:58:30.920 --> 01:58:35.720
that are applying an enormous amount of pressure to the way that the empire is trying to unfold

1233
01:58:36.600 --> 01:58:42.440
it's agenda 80% of Americans in the most recent poll think that the AI industry need to be

1234
01:58:42.440 --> 01:58:48.040
regulated yes when was the last time that 80% of Americans were on the same side of an issue no

1235
01:58:48.040 --> 01:58:51.640
yeah when I had these conversations on the podcast the comments section are clear yeah there's

1236
01:58:51.640 --> 01:58:55.640
no there's no disagreement there's no one in there going oh no I think they should crack on yeah

1237
01:58:55.640 --> 01:59:01.320
dozens dozens of protests against data centers have broken out all around this country in the US

1238
01:59:01.400 --> 01:59:06.760
all around the world so what do we do about it so these are think people that are doing something

1239
01:59:06.760 --> 01:59:13.800
about it they are actually reasserting their agency and exercising democratic contestation against

1240
01:59:13.800 --> 01:59:18.600
the ways that the empires are going about their business what goal should we be aiming at so if I

1241
01:59:18.600 --> 01:59:22.440
said to my audience Janet at home because this is kind of what I see in the comments is hopelessness

1242
01:59:22.440 --> 01:59:28.840
it's like what can I do I'm just uh yeah well well well the goal is not that we completely get rid

1243
01:59:28.920 --> 01:59:33.400
of this technology the goal is that these companies need to stop being empires and the way I define

1244
01:59:33.400 --> 01:59:38.760
like a typical business versus an empire is that the empires are predicated on this idea that they

1245
01:59:38.760 --> 01:59:43.640
do not have to provide a fair exchange of value with the workers who work for them or the people who

1246
01:59:43.640 --> 01:59:48.040
use them or all of the other people that are involved in like the supply chain of producing and

1247
01:59:48.040 --> 01:59:52.680
deploying these technologies they can extract and exploit and extract and exploit and get more

1248
01:59:52.680 --> 01:59:59.000
value than what they offer whereas typical businesses there is a fair exchange you buy a service

1249
01:59:59.000 --> 02:00:03.320
you feel like you got the same amount of value as the service they you provide it but like for

1250
02:00:03.320 --> 02:00:07.240
these data annotation workers for example they do not feel in any way that they're being paid the

1251
02:00:07.240 --> 02:00:12.520
same value that they provide to these companies so that's like for me the north star is like we

1252
02:00:12.520 --> 02:00:20.280
should be pushing back and holding accountable these companies when they operate in an imperial way

1253
02:00:20.280 --> 02:00:24.920
and that's what we've seen with all of these people that are now literally protesting in the streets

1254
02:00:24.920 --> 02:00:30.600
against data centers and having an enormous effect by the way actually stalling data center projects

1255
02:00:30.600 --> 02:00:35.800
and also completely banning data centers from being developed in their localities we're seeing that

1256
02:00:35.800 --> 02:00:40.600
with artists and writers that are suing these companies for intellectual property infringement and

1257
02:00:40.600 --> 02:00:46.600
creating a huge public conversation about what is it that we actually how do we actually want to

1258
02:00:46.600 --> 02:00:53.480
protect our intellectual property it's like I three weeks ago I met Megan Garcia who was the mother

1259
02:00:53.480 --> 02:01:01.480
of seulsets or the third who was the 14-year-old who died by suicide after being sexually groomed by

1260
02:01:01.480 --> 02:01:11.720
characterized trap-on and she when that happened I mean obviously was incredibly devastated by what had

1261
02:01:11.720 --> 02:01:17.720
happened to her son she also decided to do something about it she sued the companies and that

1262
02:01:17.720 --> 02:01:23.640
lawsuit then sparked many other parents and families who were actually experiencing similar things

1263
02:01:23.640 --> 02:01:30.920
to sue these companies as well that has created an enormous public conversation about what these

1264
02:01:30.920 --> 02:01:37.880
companies are actually doing when they exploit and they extract what is the cost to the lives of

1265
02:01:37.880 --> 02:01:43.000
people around the world including children so what do you think my audience should do if they

1266
02:01:43.000 --> 02:01:48.120
if they agree with everything written in your book age empire of AI dreams and nightmares and

1267
02:01:48.120 --> 02:01:52.280
some ornaments open AI if they agree with everything said here if they agree with everything we've

1268
02:01:52.280 --> 02:01:56.840
discussed today they can send about their kids they they don't want everyone to become data

1269
02:01:56.840 --> 02:02:01.640
labelers they don't think that's a particularly great solution what can they actually go and do

1270
02:02:02.120 --> 02:02:07.320
when I was writing the book the only discourse that was happening was this is the best thing since

1271
02:02:07.320 --> 02:02:14.920
sliced bread because of all the actions of these people like saying when they're they're they're not happy

1272
02:02:14.920 --> 02:02:19.720
with the things that these companies are doing we now have 80% of Americans that want to regulate

1273
02:02:19.720 --> 02:02:25.960
this industry and so I would say to people think about all the ways that your life intersects

1274
02:02:25.960 --> 02:02:32.760
with the resources and that the air industry needs to perpetuate what they do and also the spaces

1275
02:02:32.760 --> 02:02:38.600
that they would need to deploy these technologies to continue having broad based adoption in

1276
02:02:39.480 --> 02:02:47.000
their work so you're a data donor to these companies you could withhold that data and that's

1277
02:02:47.000 --> 02:02:50.840
what those artists and writers are are doing like they're suing these companies to withhold to

1278
02:02:50.840 --> 02:02:55.720
try and create mechanisms by which that data would then be withheld you probably have a data

1279
02:02:55.720 --> 02:03:01.880
center popping up around you if you're at a school environment or a company environment you're

1280
02:03:01.880 --> 02:03:06.760
probably having a discussion in those environments right now about what should the AI adoption policy B

1281
02:03:07.480 --> 02:03:12.920
and these companies they like I was talking with some open employees just the other day

1282
02:03:13.720 --> 02:03:20.440
and they were telling me that it's understood internally that the revenue targets for the company

1283
02:03:21.240 --> 02:03:28.520
are extraordinary and they need things to go flawlessly for it to all work out and so

1284
02:03:29.480 --> 02:03:35.240
they would need every single person to adopt this every single space to adopt this they would need

1285
02:03:35.240 --> 02:03:39.640
to be able to build their data centers at the speed that they're trying to build them and so what I

1286
02:03:39.640 --> 02:03:45.320
would say to everyone of your viewers is let's not make it go flawlessly if we don't agree with what

1287
02:03:45.320 --> 02:03:53.400
they are doing I can't go and then let's build alternatives because the thing is what I'm saying

1288
02:03:53.400 --> 02:03:59.000
is not that these technologies don't have utility it's that specifically the political economy that

1289
02:03:59.000 --> 02:04:05.000
has emerged to support the production of these technologies right now is exacting a lot of harm

1290
02:04:05.000 --> 02:04:12.520
on people but we have research that shows that the very same capabilities could be developed with

1291
02:04:12.520 --> 02:04:18.840
much more efficient methods with much less resource consumption and we have a lot of different

1292
02:04:18.840 --> 02:04:24.280
other AI systems outer disposal that are like the bicycles of AI that we also know provide

1293
02:04:24.280 --> 02:04:30.200
extraordinary benefit at very little cost so let's break up the empire and let's forge new paths

1294
02:04:30.200 --> 02:04:35.480
of AI development that are probably beneficial to everyone it's strange I'm quite I think I'm

1295
02:04:36.040 --> 02:04:42.520
I'm I've trained myself to deal with dichotomies in my head and this for me is such is a dichotomy

1296
02:04:42.520 --> 02:04:47.640
where I as a CEO and as a founder is an entrepreneur and someone that loves technology I think it's

1297
02:04:47.640 --> 02:04:52.040
incredible it's absolutely incredible AI it's just so amazing and incredible the things it's

1298
02:04:52.040 --> 02:04:57.560
an enabled me to do and create because it's designed to enable people like you and my car driving

1299
02:04:58.200 --> 02:05:05.480
in the morning and being safer incredible um I think you know the billion odd people that use AI

1300
02:05:05.480 --> 02:05:08.920
tools or chat to beauty or whatever it might be they'd probably say that it's added value to their

1301
02:05:08.920 --> 02:05:13.400
life but and this is the part that people find confusing that you can and I like I invest in

1302
02:05:13.400 --> 02:05:18.200
companies that are you know heavily using AI but and the big part is is it possible to think

1303
02:05:18.200 --> 02:05:24.200
that is true and also think that there are significant unintended consequences which technology

1304
02:05:24.200 --> 02:05:28.760
and history of technology should have taught us to take a moment to pause to talk about because

1305
02:05:28.760 --> 02:05:34.760
I think this is absolutely like you can have both of these things in your head and what I'm saying

1306
02:05:34.760 --> 02:05:41.560
is that this tension doesn't have to be a tension because we could actually preserve the utility

1307
02:05:41.560 --> 02:05:46.600
and benefits of these technologies but actually develop and design them in a different way that

1308
02:05:46.600 --> 02:05:51.000
doesn't have all of these unintended consequences yes and I think there needs to be a big social

1309
02:05:51.000 --> 02:05:54.760
conversation which is why I have so many conversations about AI in the show like there needs to be a big

1310
02:05:54.760 --> 02:06:00.840
social conversation uh conversation about being intentional about the social impact um the social

1311
02:06:00.840 --> 02:06:06.040
and environmental impact and that conversation is not being had in the in government from what I

1312
02:06:06.120 --> 02:06:10.520
can see the conversation takes place in the industry and actually trying to pull it out of the

1313
02:06:10.520 --> 02:06:15.000
industry and open people's minds to it is hopefully what we've been doing over the last couple of

1314
02:06:15.000 --> 02:06:20.120
months with the subject because I think it's actually been it is but it's been happening everywhere

1315
02:06:20.120 --> 02:06:25.800
outside of the industry and for local governments and state level governments there have been huge

1316
02:06:25.800 --> 02:06:29.800
conversations about this everywhere like I've been on book tour I've been to dozens of cities

1317
02:06:29.960 --> 02:06:37.480
around the world people are having these crucial conversations everywhere I have not gone to a single

1318
02:06:37.480 --> 02:06:41.800
city yeah it's everywhere even here in South bias yeah I'm going to a single city where the room is

1319
02:06:41.800 --> 02:06:47.000
not packed and people are not wrestling with the same exact questions as every other person in every

1320
02:06:47.000 --> 02:06:51.640
other room that I've been in speaking of Pat rooms now you've got to go because you've got to talk

1321
02:06:51.640 --> 02:06:55.640
today so I'm going to we've got our last question which is the closing tradition on this podcast

1322
02:06:55.640 --> 02:07:02.200
how would your advice to a friend with a terminal diagnosis differ from what you would do yourself

1323
02:07:03.160 --> 02:07:07.320
that's a great question differ from what you would do yourself oh my god I haven't

1324
02:07:08.840 --> 02:07:17.080
I would tell them like enjoy like live life for yourself and take it easy and yeah I

1325
02:07:18.360 --> 02:07:21.800
am not taking it easy well I think it's a good thing you're not taking it easy because you're

1326
02:07:21.800 --> 02:07:25.800
leading a conversation which is incredibly important and I think that's the thing I think the

1327
02:07:25.800 --> 02:07:31.320
conversation is the important thing and so you know because of algorithms and echo chambers it's so

1328
02:07:31.320 --> 02:07:36.520
rare to have a conversation these days especially a long form one like this so I think they're so important

1329
02:07:36.520 --> 02:07:42.280
in your book is for anyone that's curious about I think a lot of people would have learnt a lot of

1330
02:07:42.280 --> 02:07:46.360
stuff today because I sit here with an interview AI people all the time and I've learnt so much today

1331
02:07:46.440 --> 02:07:52.280
from reading your book and the extensive objective perspective that your book takes you're able to

1332
02:07:52.280 --> 02:07:55.800
unravel all of these stories that we sometimes see in tweets and we don't know if they're true or

1333
02:07:55.800 --> 02:07:59.240
not because you've gone and met the people and you've done your research and you're incredibly

1334
02:07:59.240 --> 02:08:06.280
intelligent person extremely intelligent person who clearly has humanities interests as your north

1335
02:08:06.280 --> 02:08:10.360
star and that shows up and everything you do and everything you say so please continue to fight in

1336
02:08:10.360 --> 02:08:14.520
the way that you are because it's an incredibly important one and it's people like you that are

1337
02:08:15.240 --> 02:08:22.440
I think galvanizing the world to take the collective action that we're starting to see everywhere

1338
02:08:22.440 --> 02:08:28.600
yeah empire of AI dreams and nightmares in Sam Holtman's Open AI by Karen Howe. I'll link it below for

1339
02:08:28.600 --> 02:08:31.560
anyone that wants to read this book I highly recommend you do it's a New York Times bestseller for

1340
02:08:31.560 --> 02:08:52.760
good reason. Karen thank you thank you so much Stephen