Latest / The Smart Cookies podcast / AI empires are here - what are they after? With Nick Srnicek

Transcript

AI empires are here - what are they after? With Nick Srnicek · The Smart Cookies podcast

0:04
Nick Srnicek is Senior Lecturer in Digital Economy
0:06
at King's College London. He's also the author
0:09
of Platform Capitalism, After Work, A History
0:12
of the Home, and The Fight for Free Time. And
0:15
the most recent book, Silicon Empires, The Fight
0:18
for the Future, not out just yet, I believe,
0:21
but soon, definitely soon enough to hopefully
0:24
warrant this conversation. Nick, welcome to the
0:27
show. Thank you very much for having me. So this
0:31
isn't going to be a conversation about the technological
0:33
aspect of AI and its progress. However, for the
0:36
sake of discussion, I do have to ask you whether
0:40
you believe that current AI is groundbreakingly
0:44
transformative or not, because there seems to
0:46
be like two camps and they both seem to have
0:49
pretty convincing arguments. So in one corner,
0:51
you have folks saying AI is literally five seconds
0:54
away from becoming Skynet and murdering us all.
0:57
And then on the other. There's other people saying
1:01
it's a massively overhyped guessing machine,
1:04
nothing more than a glorified autocorrect. So
1:08
I was wondering, which camp do you fall into?
1:12
Yeah, neither is the short answer. So let me
1:17
set it up. I think that there's a variety of
1:19
different camps, and you set it up quite nicely
1:20
there. The one side says... These large language
1:24
models, generative AI, chat GPT, they're what
1:27
are called stochastic parrots, which I think
1:29
is a beautiful term. And the idea is that they
1:32
are effectively machines which are sophisticated
1:36
text prediction machines. They predict, you know,
1:39
you type in something and they'll predict what
1:41
comes after that. And they're very, very good
1:43
at it. but they're no better than a parrot at
1:47
sort of repeating what humans can do. So they're
1:50
stochastic parrots, and they're sort of based
1:51
on this probability distribution of how language
1:54
works. In that case, you know, these machines
1:57
are fancy. They can do some interesting seeming
2:02
things, but there's no real intelligence there,
2:04
and there's nothing beyond some sophisticated
2:06
pattern matching. On the other hand, you have
2:09
people, and there's a lot of people within the
2:11
sort of advanced AI labs who will say, This is
2:14
the foundation for artificial general intelligence
2:17
and then something like artificial super intelligence
2:20
and AI, which is better than humans at every
2:25
conceivable sort of task that we could imagine.
2:28
And neither of these, I think, really get at
2:31
what they actually do, these machines. I don't
2:34
think that these large language models are the
2:36
foundation for artificial super intelligence.
2:39
They might be part of something like that. And
2:41
I think. Something like artificial superintelligence
2:44
is at least theoretically possible, but these
2:47
are not sufficient to get us there. On the other
2:50
hand, the stochastic para -critique, I think,
2:52
misses that things are evolving. These machines
2:56
are getting better and better at the sort of
2:59
tasks that we want them to do. And I think it
3:02
underplays the significant elements of what they
3:05
actually can do already. So my sort of path is
3:08
in between, and there's a recent paper that came
3:11
out. I forget the two authors' names offhand,
3:14
but it's called AI as a Normal Technology. And
3:18
here, AI is a sort of transformative technology.
3:21
akin to something like the steam engine or electricity
3:24
or general purpose computing, but it's not something
3:28
which is going to completely revolutionize the
3:31
world. It's not artificial general intelligence,
3:33
but it will have significant economic, political,
3:36
and social impacts that will crucially take time
3:40
to unfold. So it's not something which is going
3:43
to happen, as you might hear from AI research
3:46
labs, it's not going to happen over the next
3:48
year. or two, it might take a decade or two decades
3:51
for this stuff to unpack. I just have to ask
3:53
this. Have you seen the infamous recent interview
3:55
with Peter Thiel, Palantir's founder that he
3:59
gave to New York Times? I mean, it's hard to
4:01
watch it because the guy just sucks any sort
4:03
of charisma out of the room instantly. I couldn't
4:06
watch the whole thing. It was provoking my acid
4:08
reflux. But I definitely saw the part where he
4:11
was asked, this is all over TikTok now, where
4:14
he was asked by the journalist whether he thinks
4:16
the human race should survive. And Peter Thiel
4:19
totally did not say. Yes, before he proceeded
4:23
to babble on about the Antichrist and God knows
4:25
what else. It was really odd. So I'm very confused.
4:29
But I thought this might be a good starting point
4:32
to ask, what's actually making these sweaty,
4:36
awkward big tech or AI CEOs tick? What are they
4:39
thinking? I think it's worth saying that there's
4:44
not one unified sort of vision behind a lot of
4:47
these leading AI people. And I think... oftentimes
4:50
the criticisms of them sort of take the eccentricities
4:54
of Peter Thiel and say, well, this is all Silicon
4:56
Valley. I think Peter Thiel's approach is quite
5:00
unique. I think somebody like Elon Musk has a
5:03
different vision. And I think somebody like,
5:06
you know, the CEOs of Microsoft and Google and
5:10
Amazon have a completely different vision of
5:12
what's going on here. So in my book, Silicon
5:16
Empires, I try and distinguish between all of
5:18
these different groups. And I think there's two
5:19
core groups to think about. One is what we might
5:23
call a sort of classic neoliberal focused on
5:26
a global digital market sort of institution.
5:30
And this is the big tech companies that we normally
5:32
associate things like Apple and Amazon and Microsoft
5:36
and Meta and Google. These companies have built
5:40
their empires off of a global neoliberalized
5:43
world, and they want to see that continue. On
5:47
the other hand, you have groups which are much
5:50
more prone to a sort of clash of civilizations
5:53
or sort of geopolitical or sort of patriotic
5:57
vision of what AI is supposed to be about. And
5:59
they will talk your ear off about. AI, US versus
6:04
China sort of AI race. I think Peter Thiel falls
6:07
into that and Palantir is a major example of
6:09
it. But you also have people like Palmer Luckey
6:12
from Enduro. You have people like Alexander Wang
6:15
from Scale AI, which is he's now a part of Meta.
6:19
These people will vociferously argue that this
6:23
is a race to AGI that the US needs to win as
6:28
a sort of ostensibly democratic country but that's
6:31
increasingly a sort of crazy claim to make and
6:35
for them it is an existential sort of issue about
6:37
who is going to win this race so for them it's
6:40
not about a global neoliberalized world it is
6:43
much more about american supremacy first and
6:47
foremost over other countries so different camps
6:52
but is there so i'm guessing the answer is no
6:54
but i want to ask anyway There isn't a coherent
6:57
set of ideas that are maybe a bit deeper than
7:00
just the US supremacy over China animating these
7:04
guys. It seems like it's a much more banal quest
7:07
for power than occasionally, in Peter Thiel's
7:10
case, dressed up in some nonsense about the Antichrist
7:13
and Elon's case about colonizing Mars or whatever.
7:17
Yeah, I think there is some genuine beliefs on
7:21
their part that... They do see themselves as
7:24
part of this larger, you know, almost cosmic
7:27
scale sort of project, you know, interplanetary
7:30
civilization and things like that. I think there
7:32
is probably a general belief at some level that
7:36
that is what they're aiming for. And this is
7:38
also partly what motivates like the immense sacrifice
7:41
of people and environments and communities today
7:44
is because for them. this sort of virtuous future
7:48
goal justifies any sort of sacrifices that are
7:52
made today, particularly if it's not them. But
7:56
at the same time, I do think that there is multiple
7:58
different sort of ideologies at play. There's
8:00
that sort of belief that AGI is the next stage
8:07
of evolution, that, you know, humanity is a sort
8:12
of passing phase and perhaps this is not a bad
8:14
thing, at least in some of these ideologies.
8:17
You have that interplanetary civilization idea
8:20
that you sometimes hear Musk talking about. But
8:23
I think first and foremost, it is a matter of
8:25
power. And I think a lot of the behaviors that
8:28
we see, we can understand by just saying who
8:32
is benefiting from this in terms of power and
8:34
profit. That to me fundamentally answers a lot
8:37
of the questions about what is going on in the
8:39
AI world today. Yeah, I mean, Elon Musk also
8:43
keeps harping about the Roman Empire. And then
8:46
there's also this, I don't know if you've heard
8:48
of him, this Curtis Yarvin fella, apparently
8:51
some sort of an American monarchist or something
8:54
that's supposedly a bit of an ideological inspiration
8:57
to a lot of these tech pros. I don't know if
8:59
that's true and who's advocating for some sort
9:00
of neo -feudalism or something. Have you heard
9:04
of that about him? Yeah, I mean, Curtis Yarvin,
9:06
he's a curious figure. I think in the most part,
9:09
his influence is self -declared rather than actual.
9:14
I don't think, you know, these sort of leading
9:19
figures, Trump and... J .D. Vance. I mean, there's
9:24
some connections there, but I don't think Yarvin
9:27
has any sort of direct influence on what's going
9:29
on in those cases. He's just a curious sort of
9:32
side figure, which has presented himself as a
9:35
much more influential character than he in fact
9:37
is. Okay. That at least makes him a little bit
9:42
more funny. Probably not intentionally. Okay,
9:46
what about this transhumanism thing? a quest
9:50
to extend life and possibly live forever. Is
9:55
that a real kind of ideological tenet behind
9:59
some of these companies? Yeah, I think absolutely.
10:01
It's a part of Silicon Valley's sort of culture.
10:05
You've got that crazy guy whose name I can't
10:08
remember at the moment. Brian Johnson, I believe
10:10
his name is. Yes, yes, exactly. Who's so obsessed
10:14
with this, this longevity and, you know, I don't
10:17
know, someday is probably going to get hit by
10:18
a car. But there is, I think there is this longer,
10:23
there's a longer history of this sort of thinking.
10:26
And it's not just in Silicon Valley. You can
10:28
go back and you can find it in the sort of Russian
10:30
Cosmos movement and things like that. And, you
10:33
know, trying to supersede death is a longstanding
10:37
fascination of humanity. So I don't think it's.
10:41
entirely unique to Silicon Valley or some sort
10:44
of eccentricity of that particular culture. I
10:48
think it's been there for a long time. But what
10:50
you have in Silicon Valley is people with the
10:52
power, the money, the time, and just the general
10:57
resources, the technological resources to be
11:00
able to follow this sort of goal in a way which
11:03
is more advanced than. what other people have
11:07
been able to do. So I think these people are
11:09
following along a long sort of philosophical
11:12
and ideological history of these things, but
11:16
just taking it in a much more impactful way than
11:20
what we've seen for most of history. But do you
11:24
think these guys actually read like Mark Zuckerberg,
11:26
Peter Thiel, Elon Musk, did they have some sort
11:29
of like... education in humanities and they can
11:32
track these ideas in their own mind or is this
11:36
some sort of a patchwork and they're just taking
11:38
what's useful you know at the moment as some
11:40
sort of like window dressing when they have to
11:43
say something but in reality it's more of a quest
11:46
for power like you said i mean is elon musk like
11:49
reading hegel in his spare time Or Sam Altman.
11:55
I don't know. I have a hard time believing that.
11:57
I think they've got some sort of half -baked
12:00
ideas from when they were teenagers, and they've
12:02
sort of drawn upon that as a foundation for their
12:05
worldviews. I mean, certainly the sophistication
12:09
of what they actually write and say doesn't seem
12:12
to go beyond what a teenager could say. So their
12:17
worldviews are not sophisticated in any significant
12:19
way. But there are these, you know, Brian Johnson
12:22
sort of characters, which are... perhaps more
12:25
invested in it as an ideological project. And
12:28
I think, you know, part of the risk of criticism
12:31
is to pinpoint these very social media friendly
12:36
characters and take them as representative of
12:39
a much broader thing. I think it's really problematic
12:41
to do that. They are curiosities and they're
12:45
representative of something, but how representative
12:48
are they of Silicon Valley in general or the
12:51
tech world more broadly, I think is much more
12:54
debatable. So I think some caution is needed
12:56
there. I see. So you don't think that these guys
12:59
that we've mentioned, for example, are not like
13:01
the driving force behind this entire AI progression
13:05
of movement that's happening right now. It's
13:07
a much broader scope. It's more structural. In
13:11
general, yes. In general, yes. I mean, I do think
13:14
there is something like effective altruism's
13:18
philosophical movement, which seems to have been
13:20
quite pervasive amongst the tech scene. Sorry,
13:22
what's that? I've never heard of that before.
13:25
So effective altruism, it's stemmed from a variety
13:30
of different websites sort of trying to think
13:33
rationally about things, but thinking rationally
13:36
in a very sort of... rudimentary sense of that.
13:41
But the idea was that you could sort of calculate
13:45
the best ways to act in the world to provide
13:49
the most benefits for the most people. Very utilitarian
13:54
in its sort of view, but also taking into account
13:57
these sort of, for instance, very long perspectives.
14:01
So like one of the famous sort of claims being
14:03
made is that, well, If AGI, for instance, or
14:08
ASI can resurrect all of humans in the future,
14:13
we have to take into account the interests and
14:17
the values of that sort of group of people in
14:19
the future. And how does that weigh upon what
14:22
sorts of decisions we should be making today?
14:25
So there's this real sort of emphasis on calculating
14:29
all these different things and thinking, what
14:31
is the rationally best way to... uh to act in
14:34
the world it's been really influential in the
14:37
crypto movement so there's a variety of famous
14:40
figures within the crypto movement to subscribe
14:42
to this viewpoint at least sort of on a superficial
14:46
level um perhaps as window dressing for just
14:50
making loads of money and saying that they were
14:52
doing good at the same time but it has been a
14:55
sort of quite pervasive idea. And I do think
14:58
a number of tech people have picked it up as
15:00
a sincere belief of how to organize their worldview
15:03
and how to justify their actions that they're
15:05
making. So I think that sort of philosophical
15:09
approach does have some impact in the world.
15:13
What's that called again? What kind of altruism?
15:17
Effective altruism. Effective altruism. I've
15:21
never heard of it anyway. Maybe that's perhaps
15:23
on me. Okay. I believe the premise of your new
15:26
book silicon empire is that far from being an
15:29
actually disruptive technology in any meaningful
15:31
sense ai is in effect designed to consolidate
15:35
the power of big tech can you give us like an
15:39
example that's already perhaps manifesting that
15:41
yeah so disruptive in the sense of does it disrupt
15:46
the existing tech giants right so historically
15:49
there has been You know, new sort of foundational
15:53
technologies come along and they disrupt the
15:55
existing giants. So personal computing is, you
16:00
know, supporting Microsoft's monopoly within
16:03
the 1990s and the 2000s. And then suddenly you
16:06
get a shift to mobile computing and the form
16:09
of smartphones and things like that. And that
16:12
leads to a massive disruption for Microsoft.
16:14
It leads to a massive disruption for Facebook
16:17
because their advertising has been based upon,
16:19
you know, ads served through a personal computer
16:23
rather than through a mobile phone. So all of
16:26
these businesses sort of have this moment where
16:28
they're potentially going to collapse and there'll
16:31
be sort of new upstarts emerging to take their
16:33
place. Similar sort of arguments have been made
16:36
about AI. The idea being that, again, AI is a
16:39
new disruptive technology and that traditional
16:42
business models, something like, for instance,
16:44
Google and Search, which generates most of Google's
16:48
money, that maybe AI is going to disrupt that.
16:51
and that people will no longer use search engines
16:53
instead they'll just go to chat gpt to find answers
16:56
to information that they need and so therefore
16:58
google's business will be disrupted and they'll
17:02
collapse and some new upstart will come in like
17:04
chat gpt or perplexity or something like that
17:08
So that's the story that gets told a lot. And
17:10
this is partly justification as well for saying,
17:13
well, big tech is not a problem. We just need
17:16
to allow technology to develop. And these companies
17:19
will get disrupted as a sort of natural progression
17:22
of technology. This isn't what we're seeing with
17:25
AI, though. And the key reason why is because
17:28
AI in its contemporary form, which is deep learning
17:32
since 2012 and since about 2017, has been increasingly
17:36
about what's called the transformer architecture,
17:39
which is the sort of key technical advance behind
17:43
chat GPT and all the generative AI that we see
17:46
today. That stuff, though, requires a few things.
17:50
It requires huge amounts of data, which... the
17:53
big tech companies are well -placed to have.
17:55
It requires vast amounts of computing power.
17:59
So Mark Zuckerberg yesterday was saying he's
18:04
going to spend hundreds of billions of dollars
18:06
on data centers for AI computing. So huge amounts
18:12
of money, which is only available to the big
18:13
tech companies. And it also requires incredible
18:17
amounts to be spent on the human labor behind
18:21
it. Now, this ranges from paying data labelers
18:25
and content moderators and paying people to provide
18:29
feedback on AI, which is usually done very cheaply
18:33
and spread around the world into really low wage
18:35
countries. But it also requires, as we've seen
18:38
in the past few weeks, huge amounts of money
18:40
to be spent on AI engineers. So one of the big
18:43
stories in the past few weeks in AI is that Mark
18:45
Zuckerberg has been personally approaching people
18:48
from OpenAI, people from Anthropic, people from
18:51
Google DeepMind, and saying, I will give you
18:54
a $100 million signing bonus. You know, just
18:58
astonishing amounts of money. We're talking like
19:01
sports superstars level of money in order to
19:05
move them to meta, to Facebook. So all of this,
19:09
again, requires... money that's only available
19:12
to big tech companies. It's not even really available
19:15
to open AI, to Anthropic, to all these sort of
19:19
startup AI companies which have been so successful.
19:22
They don't have that access to that much capital.
19:26
They would love to have it, but they don't have
19:28
hundreds of billions of dollars to be throwing
19:30
around. They have to go and rely upon other companies
19:33
in order to do that sort of thing for them. So
19:36
this is why AI is dependent on big tech. It's
19:39
not disrupting their position at all. It's consolidating
19:41
their position of centrality within this ecosystem.
19:44
So my book ends up talking about, well, they're
19:48
consolidating this power. How are they then reaching
19:51
out to expand their power as well? By the way,
19:57
what is the AI -propelled future that these companies
20:00
are advertising? I mean, what is AI going to
20:03
help us do? Because it's hard to find that really
20:06
clearly articulated anywhere. Yeah, it partly
20:10
depends on who you end up talking to. So if you
20:13
talk to a company like Facebook and based on
20:18
social media, based on sort of personal connections,
20:22
the answer for them is that. AI will provide
20:26
people with companionship. Zuckerberg has said
20:29
this, that there's a loneliness epidemic, which
20:32
is true, partly fostered by social media, and
20:36
that AI companions will be the answer to this
20:39
loneliness epidemic. So it fits in with his business
20:42
model. To investors, he'll also say, We'll provide
20:46
better advertising tools and we'll provide more
20:48
personalized advertising for people. And so therefore,
20:51
investors are saying, yeah, Facebook's going
20:53
to make more money. So it's great for us. So
20:56
that's the sort of vision that they have. If
20:59
you talk to somebody like Sam Altman from OpenAI,
21:02
who is... You know, regardless of whether or
21:07
not he believes it, the very function of seeking
21:11
out venture capital requires that you hype up
21:13
your business as much as possible. So for him,
21:16
AI is all about the sort of utopian futures where
21:19
we have massive abundance and we're simply distributing
21:23
out this absolute abundance to everybody. It's
21:26
the end of scarcity. The only questions then
21:29
are questions of distribution and not questions
21:32
of production or anything like that. So it is
21:36
a completely fantastical vision, and it brings
21:39
along a lot of people. And by all accounts, Sam
21:42
Altman is a very charismatic and convincing person
21:44
to investors, and therefore he can raise tens
21:47
of billions of dollars quite easily. So the story
21:51
depends on sort of the economic position of a
21:53
company. If you're social media, it's one thing.
21:55
If you're seeking after venture capital, it's
21:57
another thing. If you're organizing the world's
21:59
knowledge like Google, it's a completely different
22:01
sort of thing. But again, the answer I think
22:04
comes down a lot to what is the economic rationale?
22:07
The ideology sort of springs from that rather
22:10
than ideology determining the behaviors. Seems
22:15
that the word charisma has lost all meaning because
22:18
what I... watch sam altman i definitely do not
22:22
see a charismatic man but to each to each their
22:24
own by the way you mentioned that zuckerberg
22:26
said that you know we're we're all gonna have
22:29
five virtual ai friends and to me that just i
22:33
mean my translation to that is that these guys
22:35
are just trying to obliterate any sense of community
22:38
probably because they never had one but uh so
22:41
this is not something that should be said out
22:43
loud and yet they are saying this stuff out loud
22:46
and it's always something more horrifying than
22:50
the last thing that they said. I think it really
22:53
speaks to the sort of insularity of their position.
22:58
The fact that they say this stuff and they think,
23:00
oh, this will be great. And then everybody else
23:02
sort of recoils in horror, like, no, that's not
23:04
okay. Yeah, it shows a lot. It reveals a lot
23:09
about them, I think. So I believe it's a well
23:12
-established fact that the productivity of...
23:15
Workers globally has increased exponentially
23:18
over the last 50 years, and yet the wages have
23:20
remained stagnant while the costs are obviously
23:23
rising. Do you expect that AI will further deepen
23:28
this divide? Most likely, yes. I would say productivity
23:32
hasn't actually increased that much. And I think
23:35
one of the core economic problems for the global
23:38
economy... is precisely that, this sort of surge
23:42
of productivity growth that we saw in the frontier
23:44
countries from 1920s till about the 1970s, that
23:49
productivity growth has been dwindling since
23:52
then. So we still see productivity growth, but
23:55
not anywhere near the extent that we saw for
23:58
this brief period in the 20th century. That has
24:03
major impacts because as productivity growth
24:05
slows, It means the potential for an economic
24:10
pie that grows for everyone increasingly becomes
24:13
difficult. And you end up with a sort of economic
24:16
system which becomes increasingly more zero sum.
24:19
And so this is partly why I think geopolitical
24:25
issues are on the rise recently, is because of
24:28
a sort of general... either unconscious recognition,
24:32
you know, you sort of recognize the symptoms
24:34
of this increasing stagnation or an explicit
24:37
recognition that this is just the future. But
24:40
countries are increasingly focused on more economic
24:43
nationalism rather than the sort of economic
24:46
globalization, which we were told for 40, 50
24:49
years was to the benefit of everybody. Increasingly,
24:52
it is a much more economically nationalist world.
24:55
And crucially, this isn't just the fault of Donald
24:58
Trump. Which I think sometimes it gets chalked
25:00
up to because Biden continued it. You know, it
25:03
was a bipartisan American policy to continue
25:07
export controls, to put more barriers on trade
25:10
around the world. And this is just, I think,
25:13
the new normal sort of post neoliberal period
25:17
that we're entering into because of this sort
25:20
of stagnation. Now, AI enters into the situation
25:23
and. Again, the promise from these companies
25:27
is that AI is going to massively increase productivity.
25:30
And as a result, we'll have this massive abundance
25:33
of goods. And therefore, everybody will be able
25:35
to get whatever they've ever wanted, you know,
25:37
perfect utopia. The reality is that productivity
25:41
increases are incredibly difficult, even with
25:44
massively transformative technologies, things
25:46
like electricity, things like the steam engine.
25:49
It takes decades. for these things to have their
25:53
productivity, the sort of capability that they
25:57
have to increase productivity. It takes decades
26:00
to fulfill that. And simply because the general
26:05
technology itself is one thing, but to actually
26:08
adopt it into an actually existing business is
26:11
a completely different thing. And one of the
26:13
most famous stories from economic history is
26:15
that factories, for instance, manufacturing productivity.
26:20
It took decades for factories to reorganize around
26:23
the steam engine because they were simply sort
26:27
of plopped in as a new power source to replace
26:29
the old power sources. But that didn't maximize
26:32
the productivity enhancements that could be done.
26:35
What had to be done was you had to reorganize
26:37
the entire production workflow. You had to reorganize
26:41
the actual physical spaces of these factories,
26:43
which meant that The productivity enhancements
26:46
didn't come from old factories, which had already
26:48
been built. They came from new factories that
26:50
were completely redesigned around what this new
26:53
power source enabled. And all of this meant that
26:56
there was a complete, you know, it didn't happen
26:59
overnight. It took years and years and decades
27:01
eventually for all of this stuff to happen. Same
27:04
thing is going to happen with AI. You can see
27:07
it right now. You know, people are still wondering,
27:09
what is AI good for? How do we actually use it?
27:12
This is... What happens with every massively
27:16
transformative technology? Its real productive
27:19
uses aren't immediately apparent and they take
27:21
time for this stuff to be figured out and time
27:23
for businesses and organizations and social institutions
27:27
and laws and all of this stuff to be reconfigured
27:30
in order to take advantage of these things. So,
27:33
yeah. productivity may increase, but it's not
27:35
going to be a sudden increase. It'll be something
27:37
that takes years and decades to unfold. But what
27:41
do you think that's going to do to the labor
27:42
market and to jobs? Because some of these AICOs
27:47
themselves... have said recently, hey, in two
27:50
years, I think 30 % of the jobs are just going
27:53
to be gone. And I was just watching an interview
27:55
with Jeffrey Hinton, one of the godfathers of
27:58
AI, and he said, yeah, you better become a plumber
28:00
now because you're probably going to be unemployed
28:02
in 18 months or something. So these are pretty
28:05
startling predictions. I think there will be
28:10
significant changes to the labor market. I think
28:13
a number of jobs will massively change. But again,
28:16
it's not going to happen that quickly unless
28:19
there's some major technical advance that nobody
28:22
really foresees at the present moment. I think
28:25
specific areas of the labor market will be pretty
28:29
significantly impacted. I mean, they keep mentioning
28:31
like entry -level white -collar jobs. Yeah, this
28:36
is a curious one because there is ways in which
28:40
AI can sort of take some of this work and do
28:43
some of this work. It requires more checking
28:47
by senior people than it does to, you know, to
28:52
oversee an entry level person. And it also completely
28:56
disrupts this sort of progression that's built
28:59
up in organizations that you come in in a junior
29:02
role. You learn not just that particular, you
29:05
know, the tasks involved in that job, but you
29:07
learn how the entire organization works. And
29:10
so you still need that sort of. that sort of
29:14
pathway for people to come in. And it may not,
29:18
that sort of entry -level position may not look
29:19
the same in the future, but there still has to
29:22
be a spot for that entry -level position because
29:24
you need to introduce people to the organization.
29:26
You need to have them have an understanding of
29:29
the entire scope of the organization eventually
29:32
as they move into a senior role. So organizations
29:36
will have to work this out. And again, this is
29:38
partly why these things take time to really get
29:40
put into place. Because it can't just simply
29:44
replace entry -level workers because in five
29:46
years' time, you'll have no people to promote
29:48
to senior roles. Exactly, yeah. It is, yeah,
29:52
it's something which will require change. I don't
29:57
see it leading to mass automation. I think in
29:59
certain areas, it will lead to significant automation.
30:02
Translation is a major one. Transcription is
30:05
another one. Generic sort of content production
30:09
of graphic design and things like that, I think
30:12
is also. increasingly under a lot of threat from
30:15
this sort of stuff. But yeah, as a sort of like,
30:20
will it lead to mass unemployment? Will it lead
30:22
to 30 % unemployment? No, no. We always create
30:26
new jobs as well, which is another thing to think
30:28
about. Yeah, some jobs will be taken away, but
30:31
there will be new jobs created. We really need
30:34
to be thinking about how do we support people
30:36
through this. And this doesn't just mean job
30:38
training, which is what every politician will
30:40
tell you as like everybody can agree on that.
30:42
It means proper welfare systems designed to support
30:45
the fact that people are being tossed out of
30:48
jobs at a high rate and they need time to be
30:51
able to find new jobs and things like that. Yeah.
30:55
So I just was thinking when you said like. politicians
30:58
saying like job training and stuff, but there
31:00
was some Congress hearing about AI. I don't know
31:02
exactly what it was in the US. I've seen a video
31:05
of it. And the Congresswoman that was talking
31:08
about it kept referring to AI as A1. So she wasn't
31:13
really even know what's happening in the room.
31:18
So yeah, good luck there. But yeah, by the way,
31:22
I talked to a lot of people in creative fields
31:25
and they keep saying, The thing that you mentioned
31:28
as well, that only the top level people, the
31:31
most talented ones, are going to be able to keep
31:35
their jobs. But if that's the case, how do you
31:37
ever become a graphic designer in the first place
31:40
or a photographer or whatever, you know, a top
31:42
level one, if you're not able to go through any
31:46
sort of like professional experience as one?
31:48
I don't get it. Yeah, I mean, I don't have an
31:53
easy answer to that either. I think... One thing
31:56
to note is that people do a lot of this stuff
31:57
as well just for hobbies, particularly something
32:00
like photography. It's a hobbyist thing as well.
32:04
So there is still scope for people to do this
32:06
in their free time and to enjoy it just as a
32:08
personal venture and also use AI tools to be
32:13
able to sort of do more than they would be able
32:15
to do by themselves. That doesn't answer the
32:18
question of jobs. It doesn't answer the question
32:20
of employment and income and things like that.
32:24
people will continue to do the stuff as a hobby
32:26
as well. I mean, if AI actually replaces 30 %
32:31
of workers or 40 % or whatever it's going to
32:33
be, how will these companies make any money if
32:35
nobody has any? That's a bit of a pickle, isn't
32:38
it? Yeah, yeah. I mean, you don't believe that's
32:41
going to happen, but they keep coming out with
32:44
these sort of dramatic predictions. I think it's
32:48
why a lot of them end up pushing for something
32:49
like a universal basic income as a response.
32:52
So universal basic income, give everybody a certain
32:56
amount of money every single month, no questions
32:58
asked. And this then provides the sort of lubricant
33:03
for the economy. People go out and they consume.
33:05
They have enough money to survive. They keep
33:08
the economy ticking over. It's a simple solution,
33:12
at least sort of on a superficial level, but
33:15
I think it's... It underestimates, I think, a
33:19
lot of the power issues involved. And I think
33:21
it underestimates sort of the different elements
33:25
that are needed for, to keep people surviving,
33:28
basically. Can you talk more about the, you said
33:32
that they underestimate the power issues underneath.
33:35
Can you talk more about that? Yeah. So the basic
33:38
idea, if you believe that most of the labor force
33:42
is going to be automated away. So let's say 30
33:46
% to, you know, some predictions are up to like
33:48
100 % eventually. What does this effectively
33:52
mean? You've got a load of people who are living
33:55
off a UBI, which is provided by a government.
33:58
And you've got a handful of people who control
34:02
these AI companies and are gaining all the resources
34:05
and power and the wealth. And the redistribution
34:09
of a UBI comes from those people in power. But
34:14
those people in power ultimately don't need the
34:17
rest of the world, and they certainly don't need
34:18
six or seven or eight billion of us basically
34:22
surviving off of their charity. So that's where
34:26
the power issue sort of comes in. If you really
34:28
believe these sort of mass unemployment hypotheses,
34:34
it does become an issue of loads of people who
34:37
are surplus to the economy surviving off of the
34:41
charity of a handful of people. And that's not
34:44
a sustainable situation. You might have, if you're
34:48
lucky, some benevolent trillionaire who is deciding
34:53
to do this. And then they pass away and it turns
34:56
out their child is much worse and decides that,
34:58
no, actually, I don't care about these people.
35:01
It's not sustainable in the long term. So there's
35:04
these sorts of power issues, even at a much sort
35:07
of less radical level. So UBI, it's like pensions,
35:13
for instance. So every single year in workplaces
35:16
or at a national level, there's political debates
35:20
about how much do we increase the pensions by
35:22
at a nominal level because inflation has gone
35:26
up. And if you don't change those pension levels,
35:28
it means the real value of the pensions has gone
35:31
down. So you have to increase it in line with
35:33
inflation every year, at least, otherwise the
35:36
value is going away. Same thing holds for UBI.
35:39
You can imagine a situation where we get a UBI
35:41
implemented. It's, I don't know, here in the
35:45
UK, let's say 2 ,000 pounds a month or something
35:49
like that. Great, wonderful for this year, but
35:52
inflation will quickly eat that away. I mean,
35:54
good luck with that in London. Yeah, very much
35:59
so. But it means that you have to constantly
36:02
fight for it every single year. And again, who
36:05
has power there? Who is able to exert the force
36:08
on, you know, what level is that UBI set year
36:12
after year after year? How do you maintain the
36:14
value of it? So this is where I think these issues
36:17
of power really get, you know, they become important
36:21
to think about. You can't just say UBI is the
36:23
answer and that's it. There's power involved.
36:26
Yeah, exactly. I mean, if that happens and I'm
36:30
getting like 1 ,000 euros per month, but I know
36:34
that I'm getting this from some... fucking trillion
36:36
i'm sorry it's for some trip from some trillionaire
36:39
and that i'm living off of his charity i'm not
36:42
going to be very happy in fact i'm my blood is
36:44
already boiling just talking about it right now
36:46
so breathe your world well i mean The sort of
36:51
hypothetical scenarios that have been set out
36:53
as well is that, well, if you've got advanced
36:55
general, advanced superintelligence, you know,
36:59
you've also got control over massively advanced
37:02
security systems, which means that the average,
37:04
you know, the public, the average person stand
37:07
no chance against overturning that sort of power.
37:10
You know, they'll have drones surrounding their
37:12
underground lair. Oh, fantastic. It's completely,
37:16
you know, wild. But if you buy into the scenario
37:19
of an artificial superintelligence, superintelligence,
37:21
this sort of stuff does become seemingly more
37:24
possible. In this sort of race that's happening
37:29
right now, supposedly between China and the US,
37:32
it's a useful excuse to not have any sort of
37:35
controls and brakes on any of this, right? Because
37:39
if we say, oh, let's implement some controls
37:42
on the development of generative AI, these guys
37:44
can immediately say, oh, what about China? We
37:46
can't lose to China and vice versa, right? Yeah,
37:52
exactly. And this is explicitly what the Trump
37:55
administration has been saying. You know, there
37:58
was an international summit in Paris a month
38:01
or two ago, and the US didn't sign up to the
38:05
international summit because they said it was
38:08
going to slow down AI innovation. You know, it
38:11
was an impediment within this AI race. As much
38:16
as I do sort of think that there is something
38:18
like an AI race going on, even sort of calling
38:23
it a race risks bringing it into existence along
38:27
with all of these dynamics of racing towards
38:29
it. Now, I think China is somewhat interesting
38:31
because China doesn't seem to be as much invested
38:38
in the race. sort of idea, as the U .S. does,
38:43
partly out of necessity. You know, they can't
38:47
get the latest NVIDIA chips. They can't sort
38:49
of meet the same sort of scale that the U .S.
38:51
can do. So they can't compete at that sort of
38:53
frontier level. Yet, they developed this sort
38:56
of really impressive deep -seek model. Nevertheless,
38:59
even though they can't have access to these NVIDIA
39:03
chips and stuff. They did. They did. And this
39:06
was sort of one of the... the shocks was that
39:09
they did it so cheaply. The sort of headline
39:13
figure was something like $5 million that they
39:15
managed to produce this model. But it's worth
39:19
noting that $5 million was the cost of one training
39:22
run. But usually there's sort of multiple training
39:25
runs. It doesn't account for the cost of the
39:27
hardware, which nobody explicitly knows what
39:30
they trained it on. But the estimates are anywhere
39:32
from like 10 ,000 to 100 ,000 GPUs, which we're
39:36
talking hundreds of millions of dollars for that
39:38
sort of thing. Plus, you've got to pay for the
39:40
AI people actually building it. And then this
39:43
$5 million figure was for the... what's called
39:46
a base model. It wasn't for the reasoning model,
39:49
which gained all this attention as being really,
39:51
really sophisticated. So again, it was cheap.
39:55
We know that. And it was very, very good, better
39:59
than a lot of people thought that China could
40:00
do, but it wasn't that cheap. And they basically
40:05
relied upon a lot of NVIDIA GPUs that they sort
40:09
of got before all these export controls were
40:12
put in place. Sorry, but I rudely interrupted
40:14
you when you were in the middle of explaining
40:18
how China is not as invested as the US. Please
40:21
go on. Yeah, so it's hard to find all that many
40:28
examples of sort of the AI race discourse in
40:32
China. They're also not nearly as concerned about
40:35
things like AGI, ASI. The rhetoric isn't about
40:41
that sort of thing. Partly it is a cultural difference.
40:45
I think Silicon Valley is very invested in this
40:48
idea. And the number of sort of Chinese investors
40:52
who are invested in the same idea just aren't
40:55
as many. There are some, and DeepSeek is sort
40:58
of part of that, but there's not nearly as many.
41:01
The other side is that I think China is much
41:04
more interested in... not in innovating at the
41:09
very edges of AI, which is an extremely expensive
41:12
and difficult thing to do. They're more interested
41:15
in how do we actually take this AI and apply
41:17
it to all these different industries. This is
41:20
where I think China's doing really well and where
41:24
they're explicitly turning their focus at not
41:26
just the sort of national level government, but
41:29
also at provincial levels and municipal levels
41:32
as well. You see loads of projects to say, We
41:36
want to support businesses adopting AI, experimenting
41:38
with AI, learning how to adapt it. So that sort
41:42
of diffusion process, which, again, normally
41:45
takes years and decades to do, that they can
41:48
sort of speed run through that and try and integrate
41:51
AI into their economy as quickly as possible.
41:53
So I think they are sort of running a race, but
41:56
it's a very different race than what the US is
41:58
going for. It's a race where they're running
42:00
off in different directions effectively. Okay.
42:03
So it's not like the Cold War and the space race
42:05
and all that stuff. It's a bit different. I mean,
42:08
on some sort of level, there's concerns about
42:11
geopolitical power. There's concerns about military
42:13
power and economic power. But the pathway to
42:16
get there is quite different for both of these
42:18
countries. By the way, you said it yourself that
42:22
a lot of these AI companies are building loads
42:26
of these data centers, you know, like it's nobody's
42:28
business. And these have a massive environmental
42:31
impact. Are these companies addressing any of
42:36
this? So Elon Musk is probably the worst of this
42:41
because he is, I mean, it's the worst of many
42:45
things, but he's the worst of this as well. I
42:47
was thinking the same thing. Yeah. Not the best
42:51
fellow out there. Yeah, yeah. So in order for
42:56
his company, XAI, to catch up to OpenAI and Google
43:01
and Anthropic, basically rapidly rolled out a
43:04
massive, massive data center. At the time, it
43:06
was the biggest data center in the world. Rolled
43:09
out this massive data center. But the issue was,
43:11
of course, how do you get power that quickly?
43:15
Because power takes time. These data centers
43:18
can't rely on solar power because you need to
43:20
be able to have it 24 -7. Oh, I didn't know that.
43:24
That's interesting. Yeah, you can't have a training
43:26
run that's interrupted constantly every night.
43:29
So you need that sort of continuous power. You
43:31
need massive amounts of power. So let me give
43:35
an example. So Facebook, Zuckerberg yesterday
43:37
said they're planning a five gigawatt power data
43:40
center. The average nuclear power plant produces
43:43
about one gigawatt. So we're talking five nuclear
43:46
power plants worth of energy going towards one
43:49
data center. It's insane. So you've got this.
43:53
And Musk, what he effectively did was just rolled
43:56
this thing out as quickly as possible. And in
43:57
order to get sufficient energy, just rolled out
44:00
a bunch of extremely polluting energy sources.
44:05
It seems like probably broke a number of laws
44:08
in the process. I don't know what's going to
44:10
come of that. Probably nothing because it's Elon
44:13
Musk. But yeah, so he just doesn't care at all
44:17
about the environmental impact. He's rolling
44:19
out AI models with absolutely zero apparent concerns
44:24
about AI safety. So every other... organization
44:28
will do some like they'll do some red teaming
44:31
uh they'll do some you know trying to prevent
44:35
these models from doing you know horrible things
44:37
uh they'll at least do some efforts towards that
44:40
whereas xai seems to not care at all which is
44:44
how you end up with the mecca hitler situation
44:46
and things like that um you know grok convinced
44:50
that it's mecca hitler and turns massively anti
44:52
-semite And curiously, it has also been shown
44:56
to when you ask Grok about a controversial political
44:59
topic, the first thing it will do is go and see
45:01
what Elon Musk has said on Twitter about the
45:03
topic. So it's using Elon Musk as a source of
45:06
truth for all of these things. But this is what
45:09
happens when you roll these things out incredibly
45:11
quickly. And I think Musk is the purest example
45:15
of that. Other companies are rolling out really,
45:18
really quickly to the point where they've had
45:20
to break their net zero targets. They said they
45:23
were going to have net zero, I think, usually
45:24
by 2030. And they basically said, we can't do
45:27
that anymore. On the good side, they are largely
45:32
looking at renewable energy sources in order
45:35
to get their energy, or if not renewable. non
45:41
-carbon intensive stuff, things like nuclear
45:44
and stuff like that. So they are trying to roll
45:47
out energy sources, which are much cleaner than
45:49
what Elon Musk has done. And in China, there's
45:53
a mandate that they have to be relying upon relatively
45:56
green sources of energy, that you can't just
45:59
rely upon coal or something like that. So on
46:04
the plus side, the AI boom means that there is
46:09
a wave of investment in clean energy that just
46:12
wasn't going to happen otherwise and in my sort
46:16
of like my sort of like biggest hopes for the
46:20
future is that if we have a system where the
46:23
ai bubble collapses the data centers go down
46:26
but all those renewable energy sources that we've
46:29
built up to power them continue on uh and actually
46:32
we end up in a much better energy circle situation
46:35
in five years time than we we ever thought was
46:37
possible um you know capitalism rolling out green
46:41
energy not to combat climate change but to build
46:43
better ai but in the end maybe it'll help with
46:46
climate change that's my that's my hope fingers
46:49
crossed these companies are like they're trying
46:52
to meet their climate goals i think there is
46:57
From Google and Microsoft in particular, there
46:59
is a real desire to meet those sorts of goals,
47:02
whether that's because of corporate pressure,
47:05
investor stakeholder pressure, or just a general
47:08
recognition that climate change is a serious
47:11
issue that needs to be tackled. But at the same
47:13
time, they're also compelled by capitalist competition
47:17
to roll out data centers as quickly as possible.
47:21
Relying on the 10 -year timeline for a new nuclear
47:23
energy source to come on board isn't possible.
47:27
They can't rely upon wind power and solar power
47:30
to do all this stuff either. It can be a part
47:32
of the answer, but not the complete answer. So
47:35
they will have to rely upon dirty energy, and
47:37
they recognize that. So I think they are somewhat
47:40
tied. But again, it's that competition aspect,
47:43
that AI race aspect, which is really driving
47:46
this. Yeah, ironic that Elon Musk is the worst
47:51
when it comes to these sort of AI guardrails,
47:53
when he was one of those 100 prominent people
47:56
signing the letter to... stop the development
48:00
of AI and he was like you know parading with
48:02
this petition all over the place and but it's
48:05
not surprising at all because he's just an insufferable
48:08
I think that was more a matter of to give him
48:10
six months to catch up yeah yeah of course yeah
48:12
obviously you just said something that piqued
48:15
my interest you said that you're hoping that
48:17
the AI bubble will burst do you think that could
48:22
potentially happen Is that a real possibility?
48:25
As we say back home in Eastern Europe, from your
48:28
mouth to God's ears. Let me distinguish because
48:35
I think AI as we have it today, generative AI,
48:38
transformer architecture, I think it is transformative.
48:41
I think it will have massive economic impacts.
48:44
It will transform many elements of society and
48:47
the economy. But again, that will take decades.
48:49
And investors don't like to wait decades for
48:52
their... investments to make a return. They're
48:55
more interested in quarterly profits. They're
48:57
more interested in how do I hop on the biggest
49:01
thing right now. So I think there will be somewhat
49:07
of an investor reckoning at times. And I think
49:10
you can see that if you read the sort of quarterly
49:13
earnings calls from these companies, the questions
49:16
are always about from the investors, when do
49:20
we get our money back? What is the bottom line
49:23
here? And there's a lot of defensive justification
49:26
from the executives to say that this will pay
49:29
off. This will pay off. It might not. It might.
49:34
I'm not going to put that out of possibility.
49:35
It might in the next five years actually lead
49:37
to the hundreds of billions of dollars that are
49:40
being spent. And Goldman Sachs said a trillion
49:44
dollars is going to be spent. Maybe. It will
49:48
make that back. I'm not going to put that out
49:50
of the realm of possibility. But it's also somewhat
49:54
unlikely, in which case investors may decide
49:57
to turn against these companies and pull back
49:59
some of that spending on these data centers and
50:03
AI training runs and AI research and things like
50:06
that. So in that case, there would be a sort
50:09
of what's called an AI winter. which throughout
50:12
the period of research into AI, there's been
50:15
moments where everybody gets really hyped up.
50:17
This new technology comes along. People are like,
50:19
this will solve all these economic issues. And
50:22
then it doesn't. And then the funding retreats
50:24
and you lead to an AI winter where AI research
50:27
sort of dwindles in academia for a while. That
50:30
could happen again. But I don't think it's going
50:34
to be a sort of permanent thing. I think the
50:36
nature of AI as a general... So speaking of AI
50:42
from early neural networks onwards, it continues
50:46
to grow and develop. And eventually some sort
50:48
of new technique will be created. Some new possibility
50:52
will come about. And even if there's an AI winter,
50:56
people will be like, okay, this is the new way.
50:58
And you'll get tens, hundreds of billions of
51:00
dollars pouring back into the field. So yeah,
51:03
short term, there could be an AI winter. But
51:05
long term, I think AI is still economically really
51:08
impactful. Yeah. Bad news. I was kind of hoping
51:13
that it would be like the smartphone, you know,
51:15
that there was this huge boom, but like the last
51:18
three iPhones or whatever, basically the same.
51:21
The camera is a little bit better, but it's not
51:23
a completely new transformative machine anymore.
51:26
And nobody's expecting it to be anymore. So,
51:28
I don't know. I guess we'll see. Yeah. By the
51:34
way, I also wanted to ask the following. Every
51:38
single person that I talk to kind of hates AI
51:42
and wants this development to stop. And yet we're
51:45
getting a new chat GPT update every couple of
51:48
months, which in turn makes us feel even more
51:50
powerless because we don't want this to happen,
51:52
you know, to be completely at the mercy of these
51:55
companies. What is to be done? Yeah, it's difficult.
52:01
There's no real way to stop this constant pushing
52:05
out of new products. You can not use it. You
52:12
can feel free as an individual to not use it.
52:15
But it's far too embedded in the economy already
52:18
for that to make much of a difference. If you're
52:22
not using it, your co -workers and friends and
52:24
family are probably using it. So I don't think
52:32
that sort of individual action is going to have
52:34
much impact. on what these companies do. I think
52:40
some of the best that we can hope for is strict
52:44
regulations around the environmental impacts
52:47
of these things. So mandate that it has to be
52:52
clean energy that they're using, for instance.
52:54
Don't do like what Elon Musk did and just fire
52:57
up a bunch of gas generators to support your
52:59
hobby. We need laws around, for instance, copyright
53:06
that are protecting artists and giving them some
53:12
sort of fair remuneration for these sorts of
53:15
uses being made of their property. We also need
53:21
rules around the low -wage workers who are involved
53:24
in this stuff. So people typically in a place
53:27
like Kenya who are doing a lot of the sort of
53:32
background work to build these things by providing,
53:36
for instance, judgments on what is a good AI
53:39
response and what is a bad AI response, often
53:42
getting paid very poorly for it, or who are having
53:45
to do sort of content moderation for these things
53:48
to make sure that they're not producing MechaHitler
53:50
and all this nonsense. So those sorts of things,
53:54
I think, are what we can sort of push for. But
53:58
again, I don't think that's sort of, I'm not
54:00
using... AI is sufficient and I don't think it's
54:04
going to have much impact on what these companies
54:06
are doing. That's like recycling to fight climate
54:09
change, I suppose. Yeah, very much so. Coming
54:13
back to the original villain of this story, Peter
54:16
Thiel, what are some of the military applications
54:18
of this stuff? I actually don't know what Palantir
54:22
is doing because it always gets me depressed
54:24
when I start watching these videos, but what
54:26
are they doing? I do know that they signed some
54:29
sort of massive contract with the Pentagon or
54:31
something. Yeah. And a bunch of the AI companies
54:35
just signed a Department of Defense contract
54:37
yesterday for $800 million, I think, collectively.
54:40
Yeah. Wonderful. So in my book, I recount this
54:45
quite curious change. Because when 2024 started,
54:51
every major AI company explicitly said, you cannot
54:56
use our AI for military or harmful purposes.
54:59
So it was written into their rules, their foundational
55:02
principles that you couldn't use their AI for
55:04
these sorts of purposes. Over the course of the
55:08
next 12 months, every single one of those companies
55:10
changed those rules around military use. And
55:14
all of them started making contracts with the
55:17
military. Now, I think for the startups, groups
55:20
like OpenAI and Anthropic, The idea here is that
55:25
you need money, and the military is a great source
55:28
of money that is not going to ask a lot of questions
55:33
about, is this profitable? They'll just ask whether
55:36
or not it helps us in our killing people projects.
55:42
As for what is actually being used, the actual
55:44
use cases, I think, are somewhat obscure because
55:48
it is top secret stuff. But we know, for instance,
55:53
that they're used to analyze satellite imagery.
55:57
So instead of having humans pour through this
55:59
massive amount of data, you use AI to pour through
56:02
all this imagery and try and figure out what
56:07
is a target, what's not. That's been used already
56:10
by Israel and Gaza and their genocide against
56:12
Gazan people. We also know that they're using
56:18
this AI stuff increasingly in... in drone technology,
56:23
but also there's work on fighter jets that don't
56:29
have any pilot inside of them. So I think Anduril
56:33
is working on that sort of stuff and a couple
56:36
of other companies. So increasingly, you know,
56:39
military armaments of the future are not involving
56:44
any humans, except at some sort of decision -making
56:49
level, hopefully. But more and more of that decision
56:52
-making will be offloaded to AI. And then I've
56:55
seen as well, there's the intelligence community,
56:57
for instance, trying to find like financial connections
57:02
amongst terror groups and things like that. Using
57:06
generative AI as a way to try and find subtle
57:09
patterns that they wouldn't be able to do otherwise.
57:14
So yeah, that's sort of the use is being made
57:17
of it. And it's a bit terrifying. to be honest.
57:22
Yeah, I was just about to ask. These are all
57:23
great tools also for a totalitarian regime. Coupled
57:29
with AI, it really does seem like nightmare fuel.
57:31
Does it keep you awake at night sometimes? I
57:36
mean, I do worry about the world that my kids
57:37
are coming into. Politically, economically, socially.
57:45
I mean, for me, one of the biggest worries. The
57:48
recent Iran -Israel conflict was, I think, really
57:51
illuminating for me because you've got this long
57:55
range war going on, Iran shooting long range
57:58
ballistic missiles, Israel firing back with its
58:01
anti -missile systems. And ultimately what that
58:06
battle was going to come down to was who runs
58:08
out of weapons first, because there's only a
58:11
certain amount of those missiles and they're
58:12
costly and they take time to produce and things
58:14
like that. So the battle was basically who has
58:17
the most military resources and how long can
58:19
that last? My real concern is that since World
58:23
War II, Europe has not been all that invested
58:26
in building up their militaries. Beyond basic
58:29
sort of security, there's not been a huge reinvestment,
58:32
particularly in Germany for obvious reasons.
58:36
That's all changing now. We have every single
58:38
European country deciding we are going to spend
58:41
a lot more of our GDP on the military. So that
58:44
situation that we saw in Israel and Iran, that
58:46
maybe the battle can only last a couple of weeks.
58:50
That's going to be changing as people build up
58:52
these military resources. And again, for the
58:55
military companies, battles are profitable. They
58:58
need more missiles. They need more technology.
59:01
They're immensely profitable. And I think there's
59:03
a real powerful interest group growing in these
59:07
sorts of hot conflicts. So it's really worrying,
59:12
the sort of conditions that are being laid down
59:14
right now. We can't do much on an individual
59:16
level, but what about on a collective level?
59:18
Yeah, on a collective level, I think there's
59:21
the things I mentioned. So environmental laws,
59:23
copyright laws, workers' rights. There's also
59:27
at an individual level, I think in workplaces,
59:30
you can collectively have some impact on. how,
59:35
when, and what sort of AI is being implemented
59:38
into your workplace. That's not going to solve
59:41
these massive social problems, but it will have
59:43
impacts in individual workplaces that I think
59:45
could be very beneficial to workers. That sort
59:49
of thing, I think, is quite important. And maybe
59:53
as the sort of negative impacts of AI become
59:56
more obvious, you might again see the sort of
1:00:00
reaction against social media that we're seeing
1:00:03
right now. So it's become increasingly obvious
1:00:07
that social media has had a load of negative
1:00:09
impacts on society, on individuals, on mental
1:00:12
health, on physical health. And there's a real
1:00:14
movement to sort of push people away from social
1:00:16
media. We might see a similar sort of thing happen
1:00:19
with AI as its impacts become more obvious to
1:00:22
the public. But I don't think the conditions
1:00:26
are there for that at the moment right now. Yeah,
1:00:29
and even so, despite the fact that there's a
1:00:31
push against using social media, people are still
1:00:33
using social media all over the place. Nick,
1:00:38
thank you so much for this conversation. Where
1:00:39
and when can people get your book? I think it's
1:00:42
available for pre -order right now. We're going
1:00:44
to include the link. I think so as well, yeah.
1:00:48
So Silicon Empire is the fight for the future
1:00:50
of AI. It should be out in October of this year.
1:00:54
It should be available via... all the usual places
1:00:57
online, and hopefully in bookstores as well.
1:00:59
All right. Thanks again. Thank you so much for
1:01:03
listening. Please subscribe to the podcast on
1:01:04
YouTube. Follow and rate on Spotify and Apple
1:01:07
Podcasts marketing tutorials. Insist, it's really
1:01:10
important that I hassle you about that, so I'm
1:01:12
doing it. Also, if you'd like to, please support
1:01:14
me on Patreon. This is a one -person podcast.
1:01:17
I know I say this every time, but I do everything
1:01:20
by my little old self. There's no team behind
1:01:22
me. It's very, very sad, I know. But you can
1:01:25
help by hopping on Patreon and finding the smart...
1:01:28
cookies podcast on there and then last but certainly
1:01:30
not least i have to thank my existing patrons
1:01:33
without you there would be no smart cookies which
1:01:36
makes you the real actual producers of this show
1:01:40
so eternal thanks to thank you so much your support
1:02:04
means the world to me