Silicon Empires: How Big Tech Is Re-Mapping Geopolitics and Labor with Nick Srnicek

WEBVTT
00:00:48.369 --> 00:00:50.530
Hello and welcome to Varm Vlog.
00:00:51.009 --> 00:00:55.969
I am happy to welcome kind of back Dr.
00:00:56.210 --> 00:00:57.649
Nick Cernick.
00:00:57.969 --> 00:01:01.329
I say kind of back because you have been a guest on one of my shows.
00:01:01.490 --> 00:01:08.210
In fact, you were one of my first five independent interviews, but you haven't been a guest on this show.
00:01:09.650 --> 00:01:12.930
And so just a little bit of background of Dr.
00:01:13.170 --> 00:01:14.769
Cernick's background.
00:01:15.090 --> 00:01:18.769
He is a writer, political economist, and academic.
00:01:19.010 --> 00:01:24.609
He is currently the senior lecturer in digital economy in the digital humanities department at King College London, I think.
00:01:24.849 --> 00:01:26.129
Is that still correct?
00:01:26.450 --> 00:01:27.250
Yeah, that's right.
00:01:27.409 --> 00:01:27.650
Okay.
00:01:27.810 --> 00:01:29.090
Academic positions move around.
00:01:29.170 --> 00:01:30.530
So I'm always checking.
00:01:31.490 --> 00:01:39.090
He's known for contributions to contemporary political economy and the Marxist or Marxian sphere.
00:01:40.210 --> 00:01:55.169
And he is also focused on anti-work politics and was briefly, although interestingly, we can talk about this in the context of your current book, associated with left accelerationism as opposed to its more common right-wing variant.
00:01:56.689 --> 00:02:09.810
He is the author of numerous articles, but four books, Inventing the Future of Post-Capitalism War Without Work, with Alex Williams from 2015, which is what I interviewed you about about a decade ago.
00:02:10.050 --> 00:02:18.530
The book Platform Capitalism from 2016, After Work, A History of Home and the Fight for Free Time with Helen Hester, you wrote in 2023.
00:02:18.610 --> 00:02:22.770
And this current book, which is Silicon Empires and the Fight for the Future of AI?
00:02:24.770 --> 00:02:37.330
You also contribute to international policy discussions, and you've you've given reports to organizations that actually matter, like the United Nations Trade and Development Committee, and you've spoken to the European Parliament.
00:02:37.569 --> 00:02:41.490
So it's nice to have you back on a show with me.
00:02:41.729 --> 00:02:51.330
I I wanted to ask you before we really get into it, has your optimism about technology changed in the last decade and a half?
00:02:51.889 --> 00:02:53.650
This is a first question.
00:02:54.449 --> 00:02:59.810
Yeah, I mean, on a fundamental level, I don't think it's changed all that much.
00:03:00.209 --> 00:03:20.689
Because I think one of the things of inventing the future was precisely that it was an argument about the potentials of the technology that was coming about, but it was also a sort of warning about that's not the direction that we're headed in, and to also try and give an argument for how we can sort of maneuver things to head in a different direction.
00:03:20.930 --> 00:03:37.729
And in many ways, you know, platform capitalism was sort of written as in my own sort of way, a response to inventing the future, thinking about well, if we do want to, you know, use this technology to liberate us from wage labor, for instance, what are some of the major impediments to that project?
00:03:37.969 --> 00:03:42.769
And at that time, it was the big tech companies embodied within platform business models.
00:03:43.009 --> 00:03:52.609
So platform capitalism was sort of, you know, here's here's the optimism of inventing the future, but here's platform capitalism sort of saying, here's the challenges in our way.
00:03:52.769 --> 00:03:58.289
Uh and I sort of see silicon empires in many ways as perpetuating that sort of project.
00:03:58.370 --> 00:04:02.530
And then, you know, uh after work has been my my optimistic book as well.
00:04:04.689 --> 00:04:10.129
So you one for the optimist and one for the pessimist seems to be your current pattern.
00:04:10.370 --> 00:04:11.569
Um, yeah.
00:04:13.329 --> 00:04:21.490
You talk a lot about the generative AI stack and how the power is concentrated in physical infrastructure and not just code.
00:04:21.730 --> 00:04:35.170
One thing I found very good about the current response to data centers, which is in the United States, is overwhelmingly negative because they're being handled in a piss-poor way because of capital allocation.
00:04:35.329 --> 00:04:39.329
And I am at I I in Utah am weirdly at the center of a lot of that.
00:04:40.530 --> 00:04:54.850
Where you know you're putting heat generating, water-intensive, massive infrastructure in the middle of a mountainous desert in the worst drought conditions we've had in 150 years.
00:04:55.650 --> 00:05:07.409
But I have been trying to get people to think about this in a more broad way, because part of the problem is people haven't been looking at the physical infrastructure of the internet in general.
00:05:07.569 --> 00:05:13.009
Like the data centers have gotten them to look, but I'm like, but this isn't just true for AI guys.
00:05:13.089 --> 00:05:19.969
This has been true for server farms and all kinds of stuff for a long time.
00:05:20.209 --> 00:05:32.930
And yes, AI and to a lesser degree, Bitcoin and NFTs got people to look at it, but it's not unique to this pro to AI except for scale.
00:05:33.490 --> 00:05:39.730
I wanted you though to talk about maybe a bit how we could reimagine our relationship with this.
00:05:39.810 --> 00:05:52.129
I've looked at some stuff with China where they're, you know, talking about building these AI data centers off coast so that they have the cooling of the ocean right there.
00:05:52.290 --> 00:06:20.129
That of course will have some not insignificant environmental impact, but probably less so than, I don't know, dropping two hyperscale data centers that require more power than the entirety that my state generates, double the empower, just one of them requires double the power the entirety of my state generates and requires more water than we have in a situation of massive drought.
00:06:21.089 --> 00:06:33.089
So if we were to democratize this technology, which I've tried to get people to think about instead of thinking about just being anti-AI, like what could we do?
00:06:33.250 --> 00:06:36.449
Yes, there are cognitive effects to AI if you use it incorrectly.
00:06:36.930 --> 00:06:38.129
Some of them are quite serious.
00:06:38.370 --> 00:06:55.089
Yes, there's all kinds of downstream effects, but I I have wondered about both thinking about this as a way to break corporate choke points, but also how could we rethink the current infrastructure of this technology?
00:06:55.329 --> 00:06:58.209
And that's gonna be my first question, so I'm gonna hand it over to you.
00:06:59.250 --> 00:07:00.849
Yeah, yeah, it's a big question.
00:07:01.009 --> 00:07:10.610
And I think you're right to sort of emphasize that a lot of these negative impacts that we see with data centers today are not unique to the AI period, the generative AI period.
00:07:10.849 --> 00:07:22.209
So you have people like Mel Hogan who in the whole sort of critical data center studies have been going on for more than a decade now, looking at the impact of these things, and particularly things like streaming.
00:07:22.290 --> 00:07:30.769
You know, the sort of major negative impacts beforehand were Netflix, YouTube, podcasters, all of this sort of stuff.
00:07:31.089 --> 00:07:39.170
You know, that that still has a major negative impact, but the sort of acceleration of it with generative AI has been massive, absolutely intense.
00:07:39.329 --> 00:07:45.329
Uh, and I think the scale of it all is, you know, is far beyond what we were sort of worried about beforehand.
00:07:45.730 --> 00:08:03.490
And then in terms of thinking about it in you know, sort of a maybe a more positive sense, I mean my sort of take, and this is maybe a hot take on the data center protests, is that you know, I think the best elements of them are not a sort of complete rejection of data centers.
00:08:03.569 --> 00:08:10.050
And I think data centers have, you know, some positive functions that can be be implemented within them.
00:08:10.209 --> 00:08:19.170
They can have some utility, they have you know all the sort of negative impacts of any sort of major piece of infrastructure, but a lot of that can be offset.
00:08:19.250 --> 00:08:29.170
And I think the real value of the data center protests has been that until they started, these hyperscalers were just coming in and basically doing whatever they wanted.
00:08:29.409 --> 00:08:43.170
You know, you've got people like Elon Musk with his data centers in in Memphis, you know, one of the poorest areas uh of the country, and basically just installing these natural gas generators that are not even legal to use.
00:08:43.490 --> 00:08:50.930
He's been using them illegally with massive environmental repercussions just because he's been trying to get these things going as quickly as possible.
00:08:51.250 --> 00:08:57.889
The data center protests, however, have enabled people in communities to say, actually, you know what?
00:08:58.210 --> 00:09:00.850
We want these things on our terms.
00:09:01.090 --> 00:09:04.050
And if you don't accept our terms, you can go someplace else.
00:09:04.129 --> 00:09:05.409
You know, we will just ban them.
00:09:05.570 --> 00:09:07.970
And I think there are good terms that can be built into them.
00:09:08.129 --> 00:09:17.649
So things like you know, north northern Virginia is the data center hub of the world, they get most of their tax revenues through data centers.
00:09:17.810 --> 00:09:23.970
And there's no there's no reason why other locations couldn't get major tax revenues from these things.
00:09:24.210 --> 00:09:29.570
And that's about the only thing they can productively offer a region because they don't really bring that many jobs.
00:09:29.890 --> 00:09:31.970
They're sort of the construction period, and then that's it.
00:09:32.129 --> 00:09:34.129
They run pretty minimally afterwards.
00:09:34.450 --> 00:09:40.370
The other benefit that they can bring is that you can force them to upgrade the local electricity grid.
00:09:40.530 --> 00:09:52.290
Uh, you can force them to build sufficient and in fact beyond what is needed for their data center, building energy, renewable energy, in the local community.
00:09:52.530 --> 00:09:59.330
So you can actually be having these things adding to the electricity that's available rather than simply taking away from it.
00:09:59.650 --> 00:10:19.490
And the amount of money that is being tossed around for these things means that there is lots of resources to be building, you know, upgrading electrical grids, building new renewable energy sources, and doing so on terms of the local community rather than simply saying, like, it's either the hyperscaler's way or it's nothing.
00:10:19.730 --> 00:10:26.770
So I think the real power of these protests has been that it's become a massive political force with real political sway.
00:10:29.890 --> 00:10:31.970
I think that's uh pretty fair.
00:10:32.050 --> 00:10:46.930
I've also been trying to get people to parse out their intense hatred of AI, which I get, by saying one, that most of the job loss that you're seeing actually isn't attributable to AI.
00:10:47.010 --> 00:10:58.610
It is actually using the AI to hide other downsizes or overhiring that happened during the beginning of the COVID recovery period.
00:10:59.490 --> 00:11:13.890
And two, that it's actually also true that a lot of industry has had to scale back its promises on using AI because they were frankly, they were using in ways that I don't even know why they thought it would be successful at that.
00:11:14.050 --> 00:11:29.890
It seemed like uh despite all even Marxist belief in the rationality of capitalism, that some of this stuff was basically a hype scale.
00:11:30.370 --> 00:11:35.170
But I actually have, you know, I'm I've been made to use AI in my job.
00:11:35.410 --> 00:11:43.650
Some some of it has been helpful immensely for things like increasing accessibility, being compliant with the ADA.
00:11:44.050 --> 00:11:55.570
I now for this podcast can produce transcripts and summaries and things to aid people for screen readers and stuff like that in about 20 to 30 minutes to an hour.
00:11:55.730 --> 00:12:00.930
And it would used to be like, I need to hire four people for a week, and I'm not doing that because I don't have that kind of money.
00:12:01.330 --> 00:12:23.170
And so the intensity of the hatred sometimes has led people to over deny the usefulness of the technology, but it also does seem in this, you know, the these large tech platforms have been, I think, massively overselling things.
00:12:23.330 --> 00:12:29.330
They've also done things that don't make sense unless you assume they're utterly cynical.
00:12:29.490 --> 00:12:36.290
So, like, talk about we're developing a technology that we can completely destroy the world, and oh my god, I might destroy the world, but we're gonna do it anyway.
00:12:36.530 --> 00:12:39.810
And you're just like, there's no way you actually believe that, or you would stop.
00:12:41.330 --> 00:12:55.410
You know, maybe I'm wrong, maybe they are that craven, but I I don't also don't see, you know, don't see the threat of like AG, you know, uh general AI as really on the horizon.
00:12:55.490 --> 00:13:17.330
The power costs seem to be astronomical, but something more scalable is possible, which is self-reparing, so self-learning AI, which doesn't really think like a human, but can approximate human functions in some ways because of the complex complexity of the math going into the neural networks, right?
00:13:17.650 --> 00:13:32.129
So I was gonna ask you before we get into some of the resource demands, how would you get people to look more realistically at the capacity of this technology?
00:13:32.210 --> 00:13:35.090
Because people say, oh, it's the decreasing originality.
00:13:35.170 --> 00:13:42.210
And I I hate to be a bummer for people, but I'm like, originality is always combining past things into new things.
00:13:42.370 --> 00:13:51.090
It's like like the completely original idea does not seem to be, I'm not saying they don't necessarily exist, but they don't seem to be viable.
00:13:51.170 --> 00:14:15.650
Like if you can't scope an idea back to an idea that's already has an analogous form in general society, that idea does not seem to take, you know, and I have a you know kind of Marxist reason for believing this too, which is like, oh, you know, if it doesn't function in the current mode of production and the current social relations, then a technology or an idea for a technology is just not viable, even if it's possible.
00:14:16.850 --> 00:14:33.650
But I but how would we, you know, how would we get people to look realistically at what these things can do if we divorce them from the current tech platforms where they're trying to we'll talk about what they're trying to do in a minute, but I was just gonna ask you that first.
00:14:34.610 --> 00:14:35.490
Yeah, yeah.
00:14:35.650 --> 00:14:51.010
I mean, this is one of the things I was trying to argue against, I guess, in my book is a sort of knee-jerk dismissal of this technology as being sort of peer hype and you know, empty and just, you know, completely useless.
00:14:51.250 --> 00:15:11.970
And I think there's good reasons for believing that because you look at a lot of the people who are involved, and particularly the sort of so-called neo-clouds, they've got Bitcoin and blockchain pedigrees, and you know, they've got many of the same people who are hyping up metaverse and Bitcoin and all these other things, you know, five years ago, are now on a generative AI hype train.
00:15:12.050 --> 00:15:15.250
And you sort of say, well, it's just the newest fad.
00:15:15.970 --> 00:15:17.330
I don't think that's actually the case.
00:15:17.490 --> 00:15:24.210
I do think that this technology is extremely capable, and I think its capabilities are increasing quite rapidly.
00:15:24.370 --> 00:15:38.530
Uh and just to very quickly on that, the the question of sort of originality, because I think you're absolutely right that originality is a sort of combinatorial process, you know, something of taking all these disparate elements, combining them in new ways, and producing something novel from that.
00:15:38.850 --> 00:15:53.890
But my sort of test for originality with these systems has always been something akin to like, could you feed it the history of you know rock music up until the prog rock period of like the 1960s, 70s?
00:15:54.129 --> 00:15:57.410
And could it generate punk music as a result?
00:15:57.650 --> 00:16:04.450
You know, with no knowledge of punk music, would that that sort of novel genre leap ever emerge from these systems?
00:16:04.530 --> 00:16:07.570
And that for me is like a real marker of originality.
00:16:07.810 --> 00:16:13.410
I don't think those systems can, these systems can do it yet, but it's not necessarily an impossibility.
00:16:13.650 --> 00:16:17.810
But you know, the the key point being that they are extremely capable, they are extremely useful.
00:16:17.970 --> 00:16:24.450
I think the vast majority of people, when they use it for their sort of personal lives, they find a lot of utility in it.
00:16:24.610 --> 00:16:29.730
And this is just sort of the simple chat bots that have been around since you know 2022.
00:16:30.129 --> 00:16:36.370
And then we sort of get into the agentic AI stuff, which has emerged from really December of 2025.
00:16:36.610 --> 00:16:39.970
You know, it's been around for a very short period of time.
00:16:40.210 --> 00:16:44.290
And this stuff is becoming extremely capable in a lot of other areas as well.
00:16:44.530 --> 00:16:55.810
And I think, you know, the risk is that the left sort of dismisses it as though it's there's nothing there and assumes that it will all just sort of disappear when the bubble pops.
00:16:56.050 --> 00:17:02.850
But actually, I think there's much bigger questions about what, you know, if and when a bubble pops, this technology will continue on.
00:17:03.010 --> 00:17:08.289
It will continue to have transformative effects on the economy, on society, on politics.
00:17:08.450 --> 00:17:11.170
And we need to be thinking critically about it now.
00:17:12.289 --> 00:17:13.650
I think that's really crucial.
00:17:13.730 --> 00:17:19.009
I listen to Editron's show all the time, but he is basically the embodiment of what you're complaining about.
00:17:19.170 --> 00:17:20.529
And I've had similar sponsors.
00:17:21.170 --> 00:17:27.650
I I I I I love his his his sort of doggedness in finding all these you know pieces of evidence to support his case.
00:17:27.809 --> 00:17:31.250
But yeah, I I fundamentally sort of disagree with him on that.
00:17:31.809 --> 00:17:39.569
Yeah, I'm I I kind of split the baby in that, like I think he's right about the financing of AI not making any sense.
00:17:40.769 --> 00:18:22.210
And being based off of a model that we saw emerge post the Great Recession, where you had where this model that was really perfected by Amazon, where you know, where you would run an improfitable business enough to buy all your competitors and then set up a monopoly, and then you know, as Curry Doctor says in Shittify, the issue that you have with the AI as it was sold to people, and particularly as not as so much it was sold to individuals, but as it was sold to companies, is that it was like, oh, we'll take care of your labor problem by getting rid of the need for like 90% of your labor.
00:18:22.529 --> 00:18:24.129
And it doesn't do that.
00:18:24.210 --> 00:18:29.089
Like, you do need people who are skilled to check it, interact with it, train it.
00:18:30.450 --> 00:18:46.049
But I like you, you know, I started messing first with Gemini, then with my spouse is in medical machine learning, and she's trying to combat the evil of the way AI is being used in medicine.
00:18:46.210 --> 00:18:56.689
Um, but also by using smaller, efficient non-enterprise models run on independent machines to do it.
00:18:56.769 --> 00:18:58.529
So I've increasingly started even using that.
00:18:58.609 --> 00:19:04.769
So I've started you know having her help me train stuff to use for my own stuff.
00:19:04.929 --> 00:19:09.250
Still don't like a I mean, I think AI writing is is fine.
00:19:09.569 --> 00:19:18.849
It it has it has the feel of corporate ad copy, but you can you know you can manipulate it to do other things.
00:19:19.009 --> 00:19:26.929
I also find you have to check it, it still hallucinates like crazy and combines ideas and and all and all that, but I have to check people too.
00:19:27.009 --> 00:19:29.809
I mean, that's about my argument a little bit.
00:19:30.609 --> 00:19:36.849
Uh I guess my my concerns are not about the AI.
00:19:36.929 --> 00:19:48.129
Because, like I was telling someone privately the other day, if we're dealing AI on the scale in your independent machine at home or in a small server deck, you're really not seeing what you're seeing with the hyperscales.
00:19:48.289 --> 00:19:51.970
You're not using that kind of water, you're not using that kind of resources.
00:19:52.129 --> 00:20:26.129
But when you're doing these like hypermod scales and you look at the energy expenditure needs of them, I was gonna ask you, you know, I learned from the from the open source and the deep seek model that part of the problem that we have is for reasons that are political economic more than technological, there's no incentive right now in the United States to actually start trying to get these uh models to be more energy efficient.
00:20:28.289 --> 00:20:29.409
Is that possible?
00:20:29.569 --> 00:20:36.529
I mean, like I've looked at the studies, some seem to say that they could be, some seem to say that like, well, they could be, but you gotta stop at where you're at.
00:20:36.689 --> 00:20:42.529
If you continue increasing the intelligence, they're only exponentially use more power.
00:20:42.849 --> 00:21:15.009
And I have been like, is this, you know, it seems to me that what we've seen in the last few months, and this, you know, this is after you wrote the book, but uh with these public IPOs, is that these companies are realizing that they're not gonna have time to pull an Amazon before needing to be profitable, and they're going to need to charge the actual energy cost, which is going to be a discipline on the abuse downstream from companies who thought they could use it as a magic fix and pay, I don't know, a few thousand dollars a month and fire all their employees.
00:21:15.889 --> 00:21:22.369
So why don't you think there's been more focus on developing more efficient AI models?
00:21:22.450 --> 00:21:28.369
Because we saw that in China from necessity, just because they couldn't get the NVIDIA chips of the high enough quality at the time.
00:21:28.529 --> 00:21:30.289
I mean, they're now gotten past that.
00:21:30.450 --> 00:21:38.450
But like, so deep sea came out and it was competitive with all three models at the time it came out, and it was more efficient.
00:21:38.769 --> 00:21:41.329
So, so what do you think is going on there?
00:21:42.450 --> 00:21:47.089
Yeah, so I would say that the models have become much more efficient, actually.
00:21:47.409 --> 00:21:47.730
Okay.
00:21:47.970 --> 00:21:53.970
So if you look at, for instance, you know, it's a matter of looking at the level of capability.
00:21:54.129 --> 00:22:03.809
So, like a GPT-3 model used, I don't know, X amount of water, X amount of energy per inference when it was first launched.
00:22:03.970 --> 00:22:11.730
And then a year later, it used probably about a hundred times less of each of those things to get the same capabilities.
00:22:12.049 --> 00:22:18.369
The issue is that we're not just staying at those capabilities, we're increasing the capabilities.
00:22:18.450 --> 00:22:25.009
And so these models are becoming bigger, they're requiring longer training runs, they're requiring more GPUs to train them.
00:22:25.169 --> 00:22:32.289
And so you end up using more energy and water and all these resources to train and run these bigger models.
00:22:32.529 --> 00:22:38.849
But actually, at that at any given level of capability, there's huge efficiency improvements being made.
00:22:39.089 --> 00:22:47.169
And it's precisely because you know, the the until about this year, most of the costs for these companies were.
00:22:47.730 --> 00:22:49.250
In training these models.
00:22:49.490 --> 00:22:53.089
This year it's now moved to most of the costs are in running these models.
00:22:53.250 --> 00:22:58.769
And so there's a strong economic incentive actually for these companies to make it really efficient to run these models.
00:22:59.009 --> 00:23:05.009
So OpenAI a couple weeks ago came out and said that they'd, I think it was they'd cut their inference costs in half.
00:23:05.169 --> 00:23:08.049
They'd found some major new technique to do this.
00:23:08.210 --> 00:23:13.250
So there is a real strong focus on making these things more efficient to run.
00:23:13.490 --> 00:23:22.450
There's also a lot of impacts along the hardware side of things as well, because Nvidia's next generation of chips are always more efficient.
00:23:22.609 --> 00:23:25.329
That is, you know, that's in fact their big promise.
00:23:25.569 --> 00:23:34.450
Because effectively, what you have right now is a situation where there is a fixed amount of energy and you want to maximize the amount of compute that you can get out of it.
00:23:34.609 --> 00:23:40.529
So the more energy efficient your chips are, that means the more compute that you can get out of this fixed amount of energy.
00:23:40.769 --> 00:23:48.609
So that is sort of the economics of the situation right now, and it's why NVIDIA is worth$5 trillion at the moment or whatever.
00:23:49.089 --> 00:23:55.490
Because they are the ones that are producing the most energy efficient chips to produce the most compute.
00:23:55.649 --> 00:24:04.769
So there is a lot of focus on this stuff, but it's just that it's also driven by this demand, this race to produce more and more capabilities.
00:24:05.009 --> 00:24:14.369
And on top of that, the sort of agenda GAI that we were seeing this year, it requires these systems not just to sort of take in a question and give you an immediate response.
00:24:14.529 --> 00:24:18.849
Instead, it is taking a command, go and work for hours at a time.
00:24:19.009 --> 00:24:24.049
All of that is inference costs, energy costs, water costs, and then give you an answer.
00:24:24.289 --> 00:24:31.649
So even if these things are really efficient on a sort of basic inference level, we're now asking them to run for much, much longer.
00:24:31.889 --> 00:24:40.450
So that's why these sort of energy costs and things are exploding, despite the fact that we are actually making these models and chips much more efficient at the same time.
00:24:41.490 --> 00:24:41.809
Okay.
00:24:42.129 --> 00:24:52.689
So in some way, as much as I'm loath to admit this, market competition actually is beginning to force considerations of efficiency.
00:24:53.490 --> 00:24:54.689
Yeah, yeah.
00:24:54.929 --> 00:25:05.009
I mean, the other thing I'll say is that the the the the costs, the total sort of costs of running building and running a data center are mostly in the hardware.
00:25:05.409 --> 00:25:11.009
So it's the the the GPUs and the memory chips and connecting them all together.
00:25:11.250 --> 00:25:15.490
The energy costs are actually not that much relative to the rest of it.
00:25:15.649 --> 00:25:19.730
You know, they're not minimal by any means, but they are relatively small.
00:25:20.049 --> 00:25:24.529
So most of the focus is on yeah, that sort of thing, saving costs there.
00:25:27.970 --> 00:25:38.450
That said, I am gonna kind of insist that from my studying of the market dynamics of this, this has been an irrational market in a lot of ways.
00:25:38.529 --> 00:25:44.769
Like like the and I wanted to ask you because I don't know if we agree about this.
00:25:45.009 --> 00:25:56.210
Actually, I do tend to be a rather old school, almost grossmanite Marxist on looking at profit, but my definition of profit is really strict.
00:25:56.529 --> 00:26:02.289
So I bracket out rent revenues, IP revenues, and everything like that, which is very hard to do.
00:26:02.849 --> 00:26:15.730
And so I look at like raw input, output cost and go, okay, there is a decline in rented profits, but it's mitigated by all these other factors, IP rents, subscription models, et cetera, et cetera, et cetera.
00:26:15.970 --> 00:26:27.569
But one of the things that you mentioned uh about the about like NVIDIA, and I pointed out like when you look at the financing of this, the truly profitable parts so far are still the physical infrastructure.
00:26:27.809 --> 00:26:45.169
The the the mismatch between initial cost of the use of the pro of you know the the models and the the enterprise frameworks and all that are not profitable yet.
00:26:45.409 --> 00:27:02.369
I think there's you know, we've had one quarter where anthropic alone was profitable, and it's not entirely clear to me looking at the numbers if that has anything to do with their product or if other uh other parts of the equation.
00:27:02.929 --> 00:27:24.289
And as that's happened, you've also seen these tech companies try to tie their product more and more into the state, which you know gives them a steady revenue service, at least barring massive wars and losses and stuff like that.
00:27:25.569 --> 00:27:47.569
And so trying to understand the economics of the silicone empires does bring me back to your book, which is like some of this does seem to be about integrating private capital into government functions in a way that makes them almost impossible to pull apart without just nationalizing them.
00:27:48.289 --> 00:27:49.970
Why do you think that's happening?
00:27:50.129 --> 00:27:54.129
Is it driven by the economics or is it driven by power, or can we say it's driven by both?
00:27:54.210 --> 00:27:55.409
And if so, how?
00:27:56.929 --> 00:27:58.769
Yeah, it's a complicated question.
00:27:59.009 --> 00:28:06.289
So, I mean, taking taking the integration of the models and the companies within the government, I think there's a few different reasons.
00:28:06.450 --> 00:28:19.009
One, if you look at a company like OpenAI or Anthropic, these are startups who are dependent on continued funding from venture capital and sovereign wealth investors and things like that.
00:28:19.250 --> 00:28:27.569
They rely upon this constant increase in funding precisely because they don't have any sort of profit-making entity.
00:28:27.649 --> 00:28:33.809
You know, they don't have a digital advertising arm or cloud computing arm or anything like that that's making loads of money.
00:28:34.049 --> 00:28:40.210
So for them, it is a real desperation effort of like, how do we continue to secure funding into the future?
00:28:40.450 --> 00:28:47.649
And the standard response throughout capitalist history has been the government is a great resource for that sort of thing.
00:28:47.809 --> 00:28:56.289
You know, it's not subject to the same sort of competitive market elements that might eliminate unprofitable companies much quicker.
00:28:56.529 --> 00:29:04.450
So there's you know, a certain different set of incentives, different set of structures and support systems available through that sort of government financing.
00:29:04.849 --> 00:29:18.049
On the flip side, I think you also have companies like Palantir, as probably the paradigmatic example, which see themselves as wanting to integrate into the government precisely to build dependency.
00:29:18.210 --> 00:29:31.889
You know, they want to have their systems at the heart of governments around the world for effort, uh, for sorry, for control, for political control, and for economic extraction reasons, ultimately.
00:29:32.049 --> 00:29:44.529
You know, one of the sort of basic functions that Talentier does is they will come in and they will offer, for instance, one month free of their services, uh, and then they will jack prices up once you're dependent upon them.
00:29:44.769 --> 00:29:53.089
So this is sort of their business model, is that building into that dependency is different, I think, than say anthropic and open AI.
00:29:53.329 --> 00:30:05.649
And then you sort of have the hyperscalers, which, you know, I think in many ways it's it's a diversification of revenue for them because they make loads of money, absolute bucketfuls of money.
00:30:06.129 --> 00:30:10.689
They don't need government funding, but it's a nice way to diversify.
00:30:10.929 --> 00:30:18.689
And it's a nice way to build relationships with governments like the Trump administration, and including the Biden administration.
00:30:18.769 --> 00:30:29.409
You know, the Biden administration took a heavier hand against tech companies than had we'd previously seen from, you know, uh Obama or the first Trump administration.
00:30:29.649 --> 00:30:37.970
So building those relationships with government for them is a way of sort of just securing that that freedom to act that they they've had for so long.
00:30:38.210 --> 00:30:42.849
So I think there's all these sorts of different reasons why these companies are getting involved with government.
00:30:43.009 --> 00:30:46.210
And they don't, you know, they don't sort of neatly overlap.
00:30:46.289 --> 00:30:50.529
And there's there's ideological reasons, economic reasons, and political reasons.
00:30:51.329 --> 00:30:51.649
Okay.
00:30:53.409 --> 00:31:02.609
So we could say that it's a mixture of support, even and in a small way a backdoor bailout, nothing new that's not unique to the tech sector at all.
00:31:03.089 --> 00:31:06.769
It's a it is aiming for integration.
00:31:07.569 --> 00:31:14.129
It's the government is the largest market, governments are the largest markets in most polities, period.
00:31:15.009 --> 00:31:16.369
It seems like there's a variety of reasons.
00:31:16.689 --> 00:31:23.169
Yeah, I mean, one thing I'll point out though is that actually the size of the government contracts for a lot of these companies is relatively small.
00:31:23.490 --> 00:31:33.089
So, like I think I think Anthropic's deal with the Department of Defense, Department of War was$250 million or something like that, which is nothing to them.
00:31:33.329 --> 00:31:35.889
That's probably like the salary of one of their people.
00:31:36.450 --> 00:31:38.529
So, yeah, it isn't a lot of money.
00:31:38.609 --> 00:31:45.809
And same with the hyperscalers, you know, it's you know, it's not insignificant, it's down 10, 20 billion dollars.
00:31:46.129 --> 00:31:52.689
But given that they they bring that in easily every quarter, you know, it's not that much for them either.
00:31:55.569 --> 00:32:16.769
So one of the things I find interesting in is this observation you had about geopolitics in the book, that the initial impulse of Silicon Valley was to use a deluze and guitari kind of term, deterritalization, are you know the borderless Silicon Valley consensus, right?
00:32:17.730 --> 00:32:36.450
And now we have a complex that is tied into tight import and export controls that's hit both OpenAI and cloud models pretty profoundly in the last month, again, before you wrote your book, but your book does kind of indicate that this would happen.
00:32:36.769 --> 00:32:44.529
Do you think that the big tech companies are successfully remonizing national security anxieties to avoid domestic antitrust regulations?
00:32:44.769 --> 00:33:01.409
Do you or do you see this as them just realizing the change in the geopolitical economy uh to add a bunch of sufficients together?
00:33:01.649 --> 00:33:19.089
So that you know they realize that now they can't really deal with this borderless, deterritalized internet anymore, which has been the trend of the internet until probably if they probably until you wrote your book, Platform Capital, and we're gonna put a like rough time on it.
00:33:19.569 --> 00:33:24.849
So so do you think this is intentional, or is it just a recognition of reality, a little bit of both?
00:33:24.929 --> 00:33:26.450
Like what's going on here?
00:33:27.809 --> 00:33:39.490
So, one of the sort of core arguments I try and make in Silicon Empires is to say that the ruling classes in America are fractured in a way that we haven't seen for some time.
00:33:39.730 --> 00:33:47.409
And I think, as you say, platform capitalism was sort of at the peak period, you know, Obama and Silicon Valley were all on the same page.
00:33:47.569 --> 00:33:50.609
Whatever Silicon Valley wanted, Obama was like, go for it, whatever.
00:33:50.769 --> 00:33:51.329
Great.
00:33:51.649 --> 00:33:57.250
You didn't have any sort of tension between the tech elite and the political elite.
00:33:57.409 --> 00:34:01.169
And even within those two groups, there was a hegemonic consensus.
00:34:01.250 --> 00:34:12.849
You know, there was always people who weren't entirely happy with the situation, particularly around China, but there was a consensus amongst the political elite in Washington and the tech elite in Silicon Valley.
00:34:13.010 --> 00:34:15.970
And I think what's happened since is that that has really fractured.
00:34:16.050 --> 00:34:19.010
It's fractured along the lines of China.
00:34:19.250 --> 00:34:21.650
So, you know, are you a China hawk?
00:34:21.889 --> 00:34:30.530
Are you more of a neoliberal who thinks that bringing China into the liberal world economic order is the way forward?
00:34:30.849 --> 00:34:32.929
Are you somewhere in between?
00:34:33.090 --> 00:34:35.970
It's also fracturing along the question of AI.
00:34:36.050 --> 00:34:38.450
You know, is AI a transformative technology?
00:34:38.769 --> 00:34:40.369
Is it just a bunch of hype?
00:34:40.450 --> 00:34:41.329
Is it nothing?
00:34:41.570 --> 00:34:52.849
And then I think as well, within the Silicon Valley sphere, you have this sort of fracturing of the big tech companies who continue to thrive and want a neoliberal world.
00:34:53.010 --> 00:35:00.610
They want that deterritorialized world, they want global markets, they want free trade in goods, services, and data.
00:35:00.850 --> 00:35:12.930
And then you've got an emerging defense tech sector, Palantir, Andoril, all these other companies which are emerging out of that space, who want geopolitical tensions.
00:35:13.170 --> 00:35:27.570
You know, their their stocks, their value is skyrocketing because Germany is now spending loads of money on re-arming themselves, and Japan is spending money on it, and Britain wants to spend more money, and the US is spending more money.
00:35:27.809 --> 00:35:31.010
All of this geopolitical tension is great for their business.
00:35:31.170 --> 00:35:44.769
And so they're quite happy to break down the neoliberal order to say, you know, the most important thing about a supply chain is not where the cheapest labor is, but where we can secure it for our country's, you know, sake.
00:35:45.090 --> 00:35:49.490
So that is like the real sort of fracture which is going on within Silicon Valley, I think.
00:35:49.730 --> 00:36:00.210
And it's interesting because Apple, for instance, you know, Apple's been looking for uh one of the big hits for Apple this year is that a gentig AI requires a lot of memory.
00:36:00.370 --> 00:36:07.090
In order to keep all these things that it's doing in its context window to keep going, it requires lots of memory.
00:36:07.250 --> 00:36:12.610
Memory chip prices have skyrocketed, which is now having derivative effects on smartphones.
00:36:13.010 --> 00:36:17.330
You are not going to be able to buy a smartphone this year for the same price as last year.
00:36:17.490 --> 00:36:21.730
The prices are going up, and you won't be able to get as much memory in them either.
00:36:21.970 --> 00:36:31.010
And so Apple, in response, has actually started looking at a Chinese supplier of memory, which goes against all of this geopolitical sort of strife.
00:36:31.250 --> 00:36:43.890
And, you know, Apple's been moving a lot of their supply chain out of China because of that, but they are so desperate to keep prices down because it's central to their business to like keep that profit margin, they're now looking to Chinese companies.
00:36:44.130 --> 00:36:51.170
So there's all these sorts of tensions going on between like what different elements of the tech sector want and what different elements of Washington want.
00:36:51.330 --> 00:37:02.210
And there is, I don't think, any consensus right now about the role of digital technology in America's strategy and then in the rest of the world, either.
00:37:03.970 --> 00:37:08.370
Well, I why do you think this is being missed about the tensions between these sectors?
00:37:08.450 --> 00:37:15.410
Do you think it's because like the Mag 7 dominates the stock market and people thus see them as a unified whole, even though they're not?
00:37:15.650 --> 00:37:28.370
Do you think it's everyone has been trying to figure out how to position themselves to both Trump, but also dealing with the fact that you know it doesn't look like the thousand-year Trump Reich is actually going to happen.
00:37:28.690 --> 00:37:31.809
So they have to figure out how to posit out of that.
00:37:31.890 --> 00:37:43.090
Like, what are the factors leading to this, given that this has become so integrated into uh their business models and the way that we talk about them?
00:37:43.330 --> 00:38:01.250
Because I, you know, I could tell you the discourse in America on like social media, and by social media here I mean mostly uh visual social media, not you know musk text media and not Facebook, although Facebook is mostly visual media now anyway.
00:38:01.730 --> 00:38:05.970
Yeah, the the the line between Facebook and Instagram is basically age.
00:38:07.170 --> 00:38:12.289
But I about lost my my thought there.
00:38:12.450 --> 00:38:16.289
No, I mean why do why are people sort of overlooking these these fractures?
00:38:17.250 --> 00:38:21.730
Why do you think they're overlooking these fractures and and why do you think people are so paranoid about it?
00:38:21.970 --> 00:38:30.370
Because on one hand, you know, in America right now, economically, people responding to the fact one in three men have exited the job market.
00:38:30.690 --> 00:38:40.289
But the problem that I have with blaming that on AI, if you look at the long-term trends, that's been building actually like almost 15, 20 years.
00:38:41.090 --> 00:38:45.490
And so it does seem like American capital has a massive problem on its hands.
00:38:45.570 --> 00:39:11.890
And it seemed like, you know, all this venture capital, particularly this money that came in back into the United States from this one time resuring without penalty from all that Panama money, did fuel a lot of this investment because these tech companies were sitting on massive cash reserves for and not investing them and not doing much with them for a very long time until the AI boom.
00:39:12.050 --> 00:39:16.210
It does seem like that money's beginning to get strung out of the venture capitals.
00:39:16.450 --> 00:39:18.690
You the capexes are insane.
00:39:19.090 --> 00:39:22.450
So that's leveling out.
00:39:23.170 --> 00:39:37.890
But the tensions between these groups and what they're aiming for does not seem to be covered very well, either in social media responses or in the mainstream press and mainstream press, actually call it like legacy media.
00:39:37.970 --> 00:39:39.890
I don't think there is a mainstream press anymore.
00:39:40.210 --> 00:39:44.530
So I wonder like, why do you think that is the case, you know, in light of your book?
00:39:44.850 --> 00:39:45.170
Yeah.
00:39:45.410 --> 00:39:45.970
Yeah, yeah.
00:39:46.130 --> 00:39:58.530
I mean, I think one simple reason is that it it is much easier to sort of unify the American ruling class and simply say that there is one set of shared interest amongst all these different elites.
00:39:58.769 --> 00:40:03.090
And that that sort of analysis, I think, has worked quite well for a long period of time.
00:40:03.250 --> 00:40:14.530
You know, if you look at the hegemony of financial interests in Washington, then the hegemony of tech interests in Washington, there have been these periods of consensus amongst the elites that have driven things forward.
00:40:14.690 --> 00:40:26.690
And, you know, you may not be able to explain every action of the US state as a result of these shared interests, but you could explain a good chunk of it, I think, by just accepting this consensus formation.
00:40:27.010 --> 00:40:32.769
I think that's one reason why is that until recently, this sort of analysis did work.
00:40:33.090 --> 00:40:47.090
The other sort of thing which blurs the lines is that there are very public overtures, particularly under Trump, to whatever the administration in power wants and says.
00:40:47.250 --> 00:40:57.570
And you know, the classic example here is people pointing to all the big tech heads who showed up for Trump's inauguration and donated to it and things like that.
00:40:57.890 --> 00:41:04.850
And it's true, you know, they did show up there and they sort of gave public fealty to Trump.
00:41:05.490 --> 00:41:21.730
But at the same time, if you look at their the the sort of industry organizations that they fund, their lobbying organizations and their public statements, they dislike a lot of what the Trump administration is doing, and they're actually quite vocal about it.
00:41:21.890 --> 00:41:26.610
And they disliked a lot of what the Biden administration was doing too, and they were quite vocal about that.
00:41:26.850 --> 00:41:33.490
So there is a lot of tensions here between the tech elite and between the political elite, if you know where to look.
00:41:33.730 --> 00:41:39.650
And but it has been sort of confused by these public shows of sort of harmony.
00:41:39.890 --> 00:41:46.850
And I think especially under Trump, because he's such a sort of fickle narcissistic figure that wants these things.
00:41:47.650 --> 00:41:47.970
Yeah.
00:41:48.050 --> 00:42:02.130
And then the other thing I would say is that there's a sort of a lack of materialist analysis in many ways, which is, you know, one of the things I tried to do in platform capitalism and I try and do in silicon empires as well, is to look at how do these businesses make their money?
00:42:02.370 --> 00:42:04.610
What sort of interest can we derive from that?
00:42:04.769 --> 00:42:11.809
So I'll give a very simple example, which is Palantir makes almost all of its money either from the US government or US corporations.
00:42:12.130 --> 00:42:16.370
Palantir is not an export-oriented company by any means.
00:42:16.530 --> 00:42:18.690
You know, most of their money is domestic.
00:42:18.769 --> 00:42:24.450
They are effectively a domestic company with a few contracts with foreign governments and foreign companies.
00:42:24.690 --> 00:42:39.170
So their interests are very narrow and they align with the US state and its geopolitical interests, much more so than a company like Microsoft or Google, which has a global footprint, is interested in global markets, is a very export-oriented sort of company.
00:42:39.490 --> 00:42:41.410
Similar thing with Tesla.
00:42:41.490 --> 00:42:48.850
And in fact, all of SpaceX and Musk's sort of conglomerate of companies, most of them don't make money.
00:42:49.090 --> 00:42:56.690
SpaceX is not profitable, XAI is not profitable, X, I don't think, is profitable, all these things combined in together, two to one.
00:43:00.450 --> 00:43:02.130
And it's profitability dropping.
00:43:02.210 --> 00:43:12.450
So yeah, BYD is sort of killing it, and most of its stock value is reliant upon the sort of mimetic value of Elon Musk as a figure.
00:43:12.610 --> 00:43:18.850
And then if you look at what Elon Musk's actual wealth is based upon, it's not cash or anything like that.
00:43:18.930 --> 00:43:21.650
It's stock in his own companies, particularly Tesla.
00:43:21.890 --> 00:43:28.769
And so when these things go down in value, which they will at some point, his wealth will just disappear.
00:43:28.930 --> 00:43:34.130
And so, yeah, wealthiest man in the world, but it's entirely precarious paper wealth.
00:43:34.289 --> 00:43:41.570
So it's a very different set of interest again from somebody like Sataya Nadella or, you know, classic Jeff Bezos or somebody like that.
00:43:41.730 --> 00:43:51.170
So looking at these sort of materialist analysis, we get, I think, a very different sense of where these companies are driving towards and what they're interested in.
00:43:51.490 --> 00:43:57.090
Yeah, to talk about this materialist analysis a little bit, and I was thinking about this when I was reading your book.
00:43:57.970 --> 00:44:14.370
There is not, you know, the the I feel like maybe I'm a little bit gloating, but I feel like for like most of the aughts and into the middle of the aught teens, everyone told me I was crazy for being kind of a classical Marxist, even though Marxism was coming back, right?
00:44:14.610 --> 00:44:16.370
And there are problems with classical Marxism.
00:44:16.450 --> 00:44:20.930
We've talked about this 10 years ago, like surplus population is kind of a messed up idea.
00:44:21.090 --> 00:44:31.570
There, you know, there I don't I don't want to valorize all of it, but I have been like the most profitable infrastructure is usually physical.
00:44:32.130 --> 00:44:40.690
And rents have a paradox because these companies want a strong state, but they also don't want to pay for a strong state.
00:44:40.930 --> 00:44:46.690
Um, and there isn't anyone else of scale to pay for it.
00:44:46.930 --> 00:44:54.530
And so So there's all kinds of conflicting incentives there, even within a company, much less within the entire sector.
00:44:55.010 --> 00:44:56.769
And that's been underexplored.
00:44:56.850 --> 00:45:01.410
That's one of the things I've liked about your work and why I wanted to have you back on is you actually do look at some of that.
00:45:01.890 --> 00:45:08.930
But it's to tie it in a little bit to some things that are very contemporaneous right now.
00:45:18.050 --> 00:45:24.530
You're in an island and sea somewhere that has some vague historical attachment to a continent and no longer identifies with.
00:45:29.809 --> 00:45:30.210
That's right.
00:45:30.289 --> 00:45:30.450
Yeah.
00:45:30.690 --> 00:45:30.930
Yeah.
00:45:31.170 --> 00:45:41.650
So you've been in these areas that are adjacent to both the United States and continental Europe for a long time and have been historically major economic plot players.
00:45:41.809 --> 00:45:57.730
I've gotten a lot of pushback for you know for years for saying Europe seems to be getting closed out of not just the geopolitics, where you're kind of seen as a proxy for the United States, particularly the UK, but in general, but also economically.
00:46:18.930 --> 00:46:23.650
And thus its auto industry is like under massive assault.
00:46:24.690 --> 00:46:36.370
From the standpoint of these uh of these uh areas that are geographically and historically and economically still quite important, but not hegemonically important.
00:46:37.330 --> 00:46:42.530
What does this move in Silicon Valley really signify to Europeans?
00:46:42.690 --> 00:46:51.010
Because one of the things that I've thought a lot about is yes, there's Palantir, which is a domestic country, and then there's like Apple and Google, which are very much not.
00:46:51.330 --> 00:46:56.769
But Apple and Google do have been bowing to American pressures sometimes.
00:46:56.930 --> 00:47:01.650
I mean, it's uneven if you actually read the books, but they are, and that's led to European distrust.
00:47:01.730 --> 00:47:06.370
But there's not really a strong enough industry domestic to Europe or the UK.
00:47:07.090 --> 00:47:09.330
I don't even know how the UK would do it, honestly.
00:47:09.570 --> 00:47:16.289
But domestic to the U to the Europe or UK to counteract this and provide these services at scale for Europeans.
00:47:16.450 --> 00:47:18.289
Uh, what does this mean?
00:47:18.530 --> 00:47:23.809
You know, what does the development of these silicone empires mean for Europe and for Canada?
00:47:26.130 --> 00:47:30.610
Yeah, I mean, it's the big question of digital sovereignty in a nutshell.
00:47:31.010 --> 00:47:38.210
I mean, so I'll tackle digital sovereignty broadly and then maybe speak a little bit more about AI sovereignty specifically.
00:47:38.450 --> 00:47:43.809
I mean, digital sovereignty, fundamental to it, is you know, cloud computing nowadays.
00:47:44.050 --> 00:47:46.130
And the hyperscalers absolutely dominate it.
00:47:46.289 --> 00:47:51.410
Amazon, Microsoft, and Google control about two-thirds of the global cloud computing market.
00:47:53.170 --> 00:48:05.890
And the sort of story is that, well, you'll never be able to compete with them because they're just so massive and the economy's a scale, and you know, the barriers to entry are so huge that you know it's not even worth it.
00:48:06.130 --> 00:48:21.410
I don't necessarily buy that because I think one of the interesting curiosities of the sort of AI period is the rise of these so-called neo clouds, which are companies that are, as the name suggests, new cloud computing companies.
00:48:21.570 --> 00:48:22.850
And in fact, there's a British one.
00:48:22.930 --> 00:48:37.809
You know, there's a British one, there's a couple of European ones that are now making, you know, 10, 20 billion dollar deals with Microsoft and NVIDIA and Enthropic and all these other companies.
00:48:38.130 --> 00:48:47.010
So there's a sort of proof of concept here that actually European companies could build significantly sized cloud computing companies if they wanted to.
00:48:47.250 --> 00:48:59.250
You know, they could take advantage of this massive demand for compute by the hyperscalers, use that to leverage themselves into a basic cloud computing business that was, you know, European-wide and sufficient.
00:48:59.490 --> 00:49:02.210
So I think there is, you know, there is something there that could be done.
00:49:02.370 --> 00:49:05.650
I mean, we do sort of have these smaller proofs of concept.
00:49:05.970 --> 00:49:33.650
On AI sovereignty and specifically, I mean, this is an argument I've sort of been working on for a little while now, but the traditional idea of AI sovereignty is that you're supposed to follow what the US is doing and to build these massive frontier models, to spend hundreds of billions of dollars doing so, to disrupt many, many communities in order to get these data centers embodied in things like that.
00:49:34.289 --> 00:49:38.850
But actually, with the rise of agentic AI, that's changed.
00:49:38.930 --> 00:49:41.570
And I think there's a there's a curious sort of opening here.
00:49:41.970 --> 00:49:49.330
So the opening is that these models, you don't necessarily need just the massive frontier models.
00:49:49.970 --> 00:49:54.370
You can instead have, for instance, a team of smaller models working together.
00:49:54.530 --> 00:50:06.370
And there's been research which has shown that you can get, with the right sort of scaffolding around these things, you can get frontier level capabilities with teams of cheaper, smaller models.
00:50:06.769 --> 00:50:24.370
On top of that, if you look at where a lot of the capability improvements over the past six, seven months have come from, it's not really been from advances in the models themselves, with sort of, you know, the recent Fable and GPT 5.6 models accepted.
00:50:24.610 --> 00:50:29.010
Most of the capability improvements have come instead from what's called a harness.
00:50:29.330 --> 00:50:37.170
And this is basically a sort of a system which tells these models how to operate and continue operating over very long periods of time.
00:50:37.410 --> 00:50:40.930
And those sorts of improvements are very cheap to do.
00:50:41.410 --> 00:50:45.490
One guy put together a thing called open claw and open sourced it.
00:50:45.570 --> 00:50:51.250
And this went absolutely viral amongst the tech elite, but also amongst Chinese people.
00:50:51.490 --> 00:50:58.930
In China, it was a huge phenomenon where apparently even grandmothers were raising a lobster, as they called it.
00:50:59.090 --> 00:51:04.610
So using this open claw thing, which is effectively a harness, to run an AI agent.
00:51:04.850 --> 00:51:07.890
Extremely technical, but it was weirdly viral.
00:51:08.050 --> 00:51:10.450
And again, that was one person doing that.
00:51:10.690 --> 00:51:15.809
So a lot of the capability improvements are not spending tens, hundreds of billions of dollars.
00:51:15.970 --> 00:51:18.769
They're just in the scaffolding around these things.
00:51:18.930 --> 00:51:24.769
So there's a lot of room there, I think, for actually building your own an open source approach.
00:51:25.010 --> 00:51:42.370
And then the other thing, which I think is really interesting, is that with agentic AI, we're sort of at the early stages of like the internet, where we don't quite have the set of standards and protocols and technical understanding to know how all this stuff should work together.
00:51:42.610 --> 00:51:50.370
And it's not consolidated into like HTTP, TCP, IP, all these sorts of protocols which are foundational to the internet.
00:51:51.170 --> 00:52:01.650
Governments could push for certain standards and protocols to be implemented in ways which enabled, for instance, a harness to use any model whatsoever.
00:52:01.809 --> 00:52:10.289
Because what the AI labs are doing is training their harnesses in such a way and then mandating that those harnesses can only be used with their models.
00:52:10.530 --> 00:52:13.010
So it becomes a closed locked-in system.
00:52:13.170 --> 00:52:21.490
And there's a brief sort of moment of time that we have where we could keep them interoperable and open and sort of prevent that concentration of power and wealth.
00:52:21.730 --> 00:52:56.690
So these are all sort of projects that I think AI sovereignty could do really cheaply and could you know, sort of collaborate amongst the middle powers, you know, Europe and Canada and the UK and Japan and South Korea and Australia, and you know, not to mention India, Brazil, South Africa, all these other countries that also want AI sovereignty could all contribute to, you know, building interoperable standards, building open source models, building an ecosystem, which at the very least would allow countries to go to open AI or Anthropic and say, no, you're asking too much of us.
00:52:56.769 --> 00:52:58.370
We're gonna go with the open source solution.
00:52:58.450 --> 00:53:00.610
And it's a bargaining chip at that point.
00:53:03.010 --> 00:53:05.890
The comparison to the only internet is actually very interesting.
00:53:05.970 --> 00:53:07.730
I talked about that a lot from your book.
00:53:07.809 --> 00:53:14.050
Uh, one of the things I was thinking about is like how luckily we were that the world wide web was deliberately unpatented.
00:53:14.370 --> 00:53:22.210
It's like just like that one decision made had huge ramifications for the entire planet made by one British dude.
00:53:22.690 --> 00:53:28.370
But I was thinking about this in terms of like your platform capitalism book, too.
00:53:28.930 --> 00:53:40.769
That there is an attempt to turn everything into closed platforms, but uh I see that as short-sighted even from the capitalist point of view.
00:53:40.930 --> 00:53:46.050
Like, like I actually do see this like oh, you're gonna have all kinds of problems when this breaks down.
00:53:46.450 --> 00:53:48.450
And historically it always has.
00:53:48.850 --> 00:53:57.650
Like you containing these, you know, it's like it reminds me of like when we tried to contain videotapes or something, more more analog technology.
00:53:57.809 --> 00:54:05.410
And this technology is actually easier to, in some ways, open source, not at the scales that the hyperscalers are doing, you're absolutely right.
00:54:05.650 --> 00:54:11.330
But like I do take some hope that it is scalable for smaller groups of people.
00:54:12.050 --> 00:54:23.010
However, the the one of the things that I get the feeling from both the left and a right in the United States is they kind of feel like we're in the boot on human face forever mode right now.
00:54:23.170 --> 00:54:36.930
And I maybe that's something that people feel because of geopolitical tensions, maybe that's something that people feel because they're becoming aware of things that have actually been going on for a lot longer than they realize and a lot longer than AI.
00:54:37.730 --> 00:54:42.050
Like I was telling someone the other day, I was like, the only thing really new about AI tech is the neural networks.
00:54:42.210 --> 00:54:46.289
Like everything else that already existed, some of it already existed for like 20 years.
00:54:47.090 --> 00:54:49.490
And the people were very surprised about that.
00:54:49.570 --> 00:55:04.370
And I was like, well, new technologies, you know, burst on the scene like this, and they have rapid uh problems, but also they they're rarely ever as new as they appear, and they tend to convergently develop, if you know anything about the history of technology.
00:55:04.530 --> 00:55:13.330
It's like certain points we're like, well, you know, five people invented the radio, or four people invented something like a car, and like two different people invented an airplane.
00:55:14.050 --> 00:55:25.250
So I have been thinking a lot about that, but I also have been thinking about the decay of a lot of this infrastructure because this is also happening simultaneously.
00:55:26.690 --> 00:55:32.850
Particularly in the United States, like like right now, and this is actually has to do with legal cons legal decisions made.
00:55:32.930 --> 00:55:36.930
Like someone say, well, why are we building AI infrastructure at such an insane rate and no other infrastructure?
00:55:37.010 --> 00:55:44.130
I'm like, well, because AI infrastructure has a massive tax break in the law that other infrastructure doesn't.
00:55:44.850 --> 00:55:50.930
It's just you know partly a big, beautiful building, and it's gonna lead to all kinds of uh uh early form of rationalities.
00:55:51.010 --> 00:55:52.210
But your book made me think about this.
00:55:52.289 --> 00:55:54.130
This was true with the early internet too.
00:55:54.370 --> 00:56:05.090
Like, like people who go back and look at the dot-com bubble, like if they'd have told you the internet was gonna go away, and I remember I was adult by the time this happened, barely an adult, but I was.
00:56:05.250 --> 00:56:18.530
I was in college at the time, and I remember people saying, Oh, tech is dead, you know, we need to go back to like old 70s-style blue ship stuff, and that would have been the complete wrong answer, too.
00:56:19.650 --> 00:56:26.450
So, how do we think about this in light of the comparison to the early internet?
00:56:27.330 --> 00:56:40.370
And, you know, maybe to pivot this to more thinking about workers and and what we can do, but you you mentioned sovereignty, but how could we utilize the tensions between these companies better?
00:56:40.690 --> 00:56:50.450
Because to me, treating them as an unstoppable monolith is also admitting defeat and also kind of not understanding how they actually work.
00:56:50.530 --> 00:56:52.450
And I just wanted you to respond to that.
00:56:53.890 --> 00:57:00.610
Yeah, I mean, I think the data center protests are a really good example of what is sort of possible.
00:57:01.010 --> 00:57:24.769
Again, it goes back to the point I made earlier about you know, the data center protests, not necessarily a sort of you know, complete rejection of building this infrastructure in a local area, but instead doing so in ways which actually benefit the local communities, whether it be better electrical grids, more tax revenue, you know, guaranteed compute access, all these sorts of things that could be built in.
00:57:25.490 --> 00:57:33.650
That's you know, sort of one way which, you know, it isn't just a boot on the face forever sort of thing.
00:57:33.970 --> 00:57:41.410
And I think, you know, there's there's all these sorts of different choke points available within the you know the these companies.
00:57:41.970 --> 00:57:47.730
There's lots of areas where you know they need they need to get data centers built, they need memory chips built.
00:57:47.890 --> 00:57:48.930
I'll give a really good example.
00:57:49.090 --> 00:58:01.650
Samsung's workers in Korea, they got, I think it was$400,000 bonus because Samsung had just made massive profits and they threatened to go on strike for a huge amount of time.
00:58:01.890 --> 00:58:09.170
And that's such a massive choke point for the AI industry that Samsung was like, okay, we'll give you all$400,000 each.
00:58:09.730 --> 00:58:13.170
So that sort of thing, you know, that that's a good example.
00:58:13.330 --> 00:58:26.930
Find choke points, put pressure on them, make sure that the massive amounts of money being thrown around are not just going to Jensen Wong and NVIDIA, but are going to everybody, as many people as possible.
00:58:32.289 --> 00:58:36.289
We're gonna travel back to 2015 for a second.
00:58:36.530 --> 00:58:54.610
Um you know, back in Inventing the Future, you and you and Alex Williams talked a lot about the problems of of left-wing and Marxist folk politics, and that is the tendency of the modern left to favor localized, defensive, and small scale actions under uh overstructural strategy.
00:58:54.850 --> 00:59:01.170
I have seen, since you wrote that book, a rise and fall of that, you know, national orientation.
00:59:01.330 --> 00:59:06.930
You weren't the only person coming up, but I remember Ed Sherizer, who I knew from Korea, wrote a book that was similar.
00:59:07.250 --> 00:59:17.170
Today uh we live in an era of left populism where we, you know, we're talking about socialism in five boroughs to be kind of a jerk about it.
00:59:17.410 --> 00:59:28.450
And while I, of course, don't have a problem with Zora Mamdani, it we have seen a turn back to although electoral but hyper-local politics.
00:59:28.690 --> 00:59:42.130
Now that makes sense in the United States in some ways because, like, we're a hyper-federated country, you can't really do politics starting at the national level in any meaningful sense, and not be like a lame duck demagogue, basically.
00:59:43.570 --> 00:59:52.050
But I was gonna ask you though, as you show in Silicon Empires, the centralization of tech power now is pretty vast.
00:59:52.769 --> 01:00:02.930
And the bear to entry of it as a competitor, if you're trying to scale at the same way they are, is billions, if not trillions, of dollars.
01:00:04.210 --> 01:00:22.050
If we assume that you were right back in 2015 that the folk politics of localized resistance is insufficient and resistance itself is frankly insufficient, it might be necessary, but it's not gonna actually get you very far, it's just gonna stop things from getting worse, maybe.
01:00:22.530 --> 01:00:38.930
How do you build, you know, counter hegemonic AI infrastructure with like digital sovereignty and and AI sovereignty uh using using both state and grassroots movements to do it?
01:00:39.010 --> 01:00:49.010
And what kind of folk political impulses would we need to re-evaluate this time to handle the situation?
01:00:50.769 --> 01:00:52.370
Yeah, it's a difficult question.
01:00:52.530 --> 01:01:26.050
I mean, I'm not sure I have a good answer, but I will sort of say that the sort of ideas that I sketched out about AI sovereignty, so particularly around interoperability and and things like that, setting standards, building a resource of open source materials, all of that I think is simultaneously the building of an alternative, as well as undermining quite a bit of the capability of these companies to concentrate and lock in these systems for for their own benefit.
01:01:26.450 --> 01:01:35.010
So I do think that is you know one sort of thing which can be done to try and shape things into a different direction.
01:01:35.250 --> 01:01:46.930
And you know, you might look at, for instance, this sort of parallel between a shift from like mainframe computing in like the 1960s to personal computing in the 1980s.
01:01:47.170 --> 01:01:59.650
And you know, what might have seemed completely technologically impossible in the 1960s to have you know this computing power on your desktop or in your hand in a smartphone eventually became possible.
01:01:59.809 --> 01:02:01.970
And a similar thing I think will happen with AI.
01:02:02.130 --> 01:02:13.890
So, like, will we always need these massive data centers to do it, or will actually a lot of the capabilities increasingly be pushed to the edge to the smartphones, to laptops and things like that?
01:02:14.210 --> 01:02:24.450
So there are you know sort of ways of thinking about those futures and pushing for those futures as a way to try and decentralize this power a little bit in terms of the infrastructure.
01:02:25.170 --> 01:02:47.010
On top of that, I think there's also a sort of political response to all of this, which is, you know, I've advocated for it for many, many years now, but it's like a post-work future to say that yes, we will accept as workers that this technology can come in and be adopted, but in return, we get, you know, we don't have to work as much, and we get the same pay.
01:02:47.330 --> 01:02:55.970
And there's apparently some Australian workers at the moment who are pushing for this right now, pushing for a much shorter working week by virtue of AI.
01:02:56.289 --> 01:03:04.850
So that sort of thing, I think, is like the the sort of political response, building the the you know, the institutions for that, the political movements, social movements for it.
01:03:05.090 --> 01:03:24.050
And then the infrastructure side is you know the sort of technical element of trying to prevent that concentration of power and and and lock-in and all these sorts of things, along with many, many other actions, I think are are sort of necessary for building a different future, a future that actually does work for everybody.
01:03:25.809 --> 01:03:27.170
That brings me to the next question.
01:03:27.410 --> 01:03:28.690
Actually, it's a perfect segue to it.
01:03:28.850 --> 01:03:31.650
This one's a little bit of a bummer, I'm not gonna lie.
01:03:32.130 --> 01:03:40.850
In the public imagination, and actually from the statements of AI figures themselves until about two months ago, where they finally stopped this shit.
01:03:41.170 --> 01:03:46.690
AI automation was framed as a sudden near-total displacement of human labor.
01:03:46.850 --> 01:03:48.370
That isn't happening, you know.
01:03:48.450 --> 01:03:52.210
There has, you know, yes, there's a hiring shortage right now.
01:03:52.450 --> 01:03:54.289
Yes, it's harder to get jobs.
01:03:54.370 --> 01:03:55.650
That's all true.
01:03:57.730 --> 01:04:02.289
But in general, we have not seen mass unemployment.
01:04:02.370 --> 01:04:05.010
There have been hiring, there have been hires back.
01:04:05.490 --> 01:04:07.650
The jobs reports in America.
01:04:07.890 --> 01:04:14.530
I don't know how much you should trust them now anyway, but because they keep on getting revised every month and they always get revised worse.
01:04:14.690 --> 01:04:17.490
However, they also aren't apocalyptic.
01:04:17.570 --> 01:04:27.970
Everyone keeps on, you know, pushing at me because I will I will say things are gonna get worse before they get better, but also we're not in an apocalyptic situation.
01:04:28.050 --> 01:04:43.730
There's not like uh we're I don't even like the the analogies to like even the bubble bursting is a mixture of 2001 and maybe some of the 70s stagflation readjustment of the blue chips, right?
01:04:43.970 --> 01:04:47.890
Like, I hate to tell people that's not the end of the world, you know.
01:04:48.130 --> 01:04:51.970
So capitalism doesn't automatically collapse from this, blah, blah, blah, blah, blah.
01:04:53.330 --> 01:04:56.769
But let's just play this out.
01:04:56.850 --> 01:05:04.050
If we don't politically intervene in some key way, capitalism actually tends to use new technology to intensify work rather than eliminate it.
01:05:04.130 --> 01:05:08.370
We don't get Keynes's world, we get I have three jobs.
01:05:09.010 --> 01:05:14.289
And personally, I have two, so I you know, like this and a whole other job teaching.
01:05:14.850 --> 01:05:19.490
So, you know, I I I feel it too.
01:05:19.970 --> 01:05:30.050
Under your framework in Silicon Empires, do you see generative AI leading to more mass unemployment?
01:05:30.530 --> 01:05:38.450
Or are we looking at an intensification but de-skilling a professional and creative laborer?
01:05:39.170 --> 01:05:44.289
And what's that you know, what's at stake?
01:05:44.450 --> 01:05:49.970
Because clearly we could use this to reduce work hours, but no one trusts anyone right now.
01:05:50.050 --> 01:05:55.809
Well, not no one, obviously Australia, but no one in America trusts anyone to reduce work hours because they're like, oh, we're just gonna die then.
01:05:55.970 --> 01:05:59.970
So, or I'm gonna have to get four more gig jobs.
01:06:00.289 --> 01:06:05.570
So, how how do we sell this towards a post-work future?
01:06:05.730 --> 01:06:08.690
And and also I'm just gonna give you a chance to define this.
01:06:08.930 --> 01:06:12.850
What do you and your your co-author mean by postwork?
01:06:12.930 --> 01:06:16.370
Because that's also a label that's meant a bunch of different things since I've encountered it.
01:06:16.450 --> 01:06:17.970
So I'll let you answer that.
01:06:18.530 --> 01:06:22.450
Yeah, so I'll maybe start with that question first because I think it's sort of central.
01:06:22.769 --> 01:06:33.809
So, postwork, I think the the sort of naive idea is that postwork means we end all necessary labor and we just sort of sit around doing nothing.
01:06:33.890 --> 01:06:38.930
You know, we sit on the couch, we play video games, whatever, you know, it's just lazy doing nothing.
01:06:39.170 --> 01:06:42.050
That's not actually what postwork means.
01:06:42.289 --> 01:06:47.330
Postwork is, I think, maybe better defined as post wage labor, uh, and wage labor involves.
01:06:47.650 --> 01:06:58.370
Involves this economic compulsion, as people like Soren Mao have pointed to, you know, the ways in which capitalism as a social system compels us to continue working even when we don't want to.
01:06:58.450 --> 01:07:00.769
You know, there's an economic necessity for it.
01:07:01.010 --> 01:07:15.090
And I think, you know, there's sort of a grand simplistic vision of history that you might say that before capitalism, we were compelled by natural necessity to go and work and generate food to be able to eat and survive and things like that.
01:07:15.250 --> 01:07:22.610
And under capitalism, we potentially have enough of all of that food and basic means to just continue to exist and subsist.
01:07:23.090 --> 01:07:31.970
It's capitalism as a social system which compels us to continue working because in order to get these basic means of subsistence, we have to earn an income and whatnot.
01:07:32.690 --> 01:07:43.250
Post-work and post-wage labor and socialism or communism, as I would understand it, is a sort of third stage where that natural necessity is gone and that social compulsion is gone.
01:07:43.410 --> 01:07:54.370
And instead, people are doing effortful activities, but they're doing so not out of some economic need, but simply because they they're passionate about something.
01:07:54.530 --> 01:08:00.930
They have some project that they want to work on, they have some collective community they want to contribute to.
01:08:01.490 --> 01:08:20.609
And then, you know, it ultimately I think there will always remain some socially necessary labor which would be distributed in as equal a manner as possible amongst people, but leaving people with plenty, much more free time than they have now to freely choose what they want to do with their lives and how they want to build their existence.
01:08:20.850 --> 01:08:22.529
That for me is what postwork is about.
01:08:22.609 --> 01:08:24.769
It's not about lazing about and doing nothing.
01:08:24.930 --> 01:08:28.529
It is about the freedom to choose exactly what we want to do.
01:08:29.489 --> 01:08:32.689
Given that very utopian idea, you know, how do we sort of get there?
01:08:32.769 --> 01:08:35.090
And I think it is difficult to imagine.
01:08:35.250 --> 01:08:55.409
I think it is, you know, I can only I can sketch out this sort of far-fetched idea, and then I can sort of sketch out immediate steps, which I think don't get us to that stage, but at least start us on the path towards that stage, which is things like a shorter working week, you know, which is why I'm massively supportive of any worker effort to do that sort of thing.
01:08:55.649 --> 01:08:59.809
And then I think with the impacts of generative A on the labor market, I think there's a couple of interesting ones.
01:08:59.970 --> 01:09:09.809
One is, you know, the sort of job churn I think will increase, even if it's only because companies think we don't need these workers, we've got AI, we'll get rid of them.
01:09:09.969 --> 01:09:15.090
And so all these workers, you know, get kicked out of their jobs and have to go find other jobs.
01:09:15.250 --> 01:09:21.489
But even within jobs, uh, people will have to be constantly learning new elements of it.
01:09:21.649 --> 01:09:25.090
You know, it will be a sort of tumultuous period, I think.
01:09:25.250 --> 01:09:26.689
And that's the baseline.
01:09:27.489 --> 01:09:39.649
I think another interesting element is, as you sort of mentioned, the sort of de-skilling of the professional classes, which is, you know, it's speaking as a member of the professional classes, doesn't sound great.
01:09:39.969 --> 01:09:45.729
But I think there is a real political element here, which is, you know, people like Ben Tarnov have been quite good about this.
01:09:45.969 --> 01:09:59.009
The ways in which the tech sector, the union organizing and work organizing that's gone on there, is in part because programmers have sort of seen that a lot of elements of their jobs can be effectively proletarianized.
01:09:59.170 --> 01:10:12.369
And so that special status that they might have had as knowledge workers, as professional workers, when they start to see it disappearing, there's a moment there that you can actually build political solidarity amongst a much wider group of workers.
01:10:12.529 --> 01:10:17.250
And I think the people on the left should be attuned to this as it's happening.
01:10:17.809 --> 01:10:39.250
The other thing, and this is, I think, more far-fetched, but still plausible enough that we should take it seriously, there are serious people who are honest and intelligent, and they really are worried about things like recursive self-improvement from these AI systems.
01:10:39.489 --> 01:10:42.689
The idea being that AI will be able to improve itself.
01:10:42.849 --> 01:10:47.729
And at that point, you will have weird things start to happen effectively.
01:10:47.889 --> 01:10:51.809
That AI will just suddenly make leaps and bounds, improvements.
01:10:52.049 --> 01:11:03.729
It will start to be able to, you know, take over companies, take over jobs in a much easier way than we sort of imagine right now and that we see right now, and that all of this could happen really quickly.
01:11:03.969 --> 01:11:10.529
Now, I think that's unlikely, pretty unlikely, but I don't, I wouldn't ascribe a 0% chance to it.
01:11:10.609 --> 01:11:13.250
You know, I think there is a possibility of this.
01:11:13.329 --> 01:11:15.489
It's what the labs are trying to build.
01:11:15.649 --> 01:11:22.849
And if it happens, I think at least some people on the left need to be considering what do you do in that sort of situation?
01:11:23.009 --> 01:11:27.489
You know, what should the political response be to recursive self-improvement?
01:11:27.569 --> 01:11:31.009
And I haven't seen anybody really on the left do that sort of work.
01:11:31.489 --> 01:11:39.969
Uh to me, it makes sure we chrain that recursive self-improvement and not the military seems to be the first thing that really matters.
01:11:40.529 --> 01:11:48.289
I'm always like talking about yes, most of the unfortunately a lot of our military AI is weird because it isn't actually from military investment.
01:11:48.449 --> 01:11:50.049
It doesn't seem to be anyway.
01:11:50.529 --> 01:11:58.210
I've looked at you know how much of the uh modern tech infrastructure comes from DARPA and it's uh it's a crazy amount.
01:11:58.369 --> 01:11:59.809
AI doesn't.
01:12:00.210 --> 01:12:03.889
And which is why you have the tensions between Department of Defense and Anthropic.
01:12:04.130 --> 01:12:10.769
Yeah, oh yeah, and then it's been private-led innovation, and that's been an ongoing pattern since the end of the Cold War.
01:12:12.449 --> 01:12:31.649
Well, you know, I have taken up saying by friend of the show, Yasha Levine, that whatever the problems of the Soviet Union were, America was a lot more innovative and a lot kinder to workers when it existed.
01:12:31.889 --> 01:12:54.289
Um now that it is gone, oh you know, the you know the and I think there are people who predict it, it's uh sadly not enough Marxists, but like the pragmatist liberal Louis Minon actually said, like, look, the the university in about 20, 30 years is in a lot of trouble because a colour large S has been over for a while, and yet we have not adjusted to that reality.
01:12:54.849 --> 01:12:58.049
That is happening now, and it's kind of ugly.
01:12:59.409 --> 01:13:06.369
And you know, you do have, you know, in our country, you have a lot of outright looting of public resources.
01:13:06.449 --> 01:13:11.969
Um, and it's not actually by the tech people so much as by actual politicians themselves.
01:13:12.769 --> 01:13:21.090
But I've also said this this is not sustainable and they know it, so something else is gonna come after it.
01:13:21.170 --> 01:13:22.449
So you know, be prepared for it.
01:13:22.529 --> 01:13:27.729
And I I think what I see right now in the US is very different than the UK.
01:13:27.889 --> 01:13:33.569
So in the US, we're gonna throw the bastards out moment politically again.
01:13:33.809 --> 01:13:42.049
And things like the DSA are the most organized factors in general society that are not connected to legacy parties.
01:13:42.210 --> 01:13:59.329
Are they, but they're not, although they are adjacent to things that are like these massive packs that exist even in the progressive movement, and so they're able to take advantage of this and and get into office.
01:13:59.489 --> 01:14:09.009
Like, yes, you know, Zoomers like socialism, millennials like socialism, uh, no one's scared of communism anymore, unless you're a right-winger.
01:14:09.729 --> 01:14:33.489
But to tie all this together into what you're talking about, it does seem like that we're gonna need like coordinated leverage on like all cylinders to start thinking about this, and that this is a moment where we do have some capacity to intervene because of the dismal public atmosphere and something else that you mentioned earlier: the division amongst capital.
01:14:33.569 --> 01:14:39.569
I mean, we can talk about division amongst tech, but like if you get outside of tech, the division amongst capital is even more profound.
01:14:40.609 --> 01:14:46.289
The attempts to reshore have failed over and over and over again, you know, all kinds of things.
01:14:46.449 --> 01:14:58.369
The the US actually has hampered a lot of its domestic production and trying to increase its domestic production because apparently our politicians don't understand understand supply chains, etc.
01:14:58.609 --> 01:14:58.849
etc.
01:14:59.090 --> 01:14:59.409
etc.
01:14:59.889 --> 01:15:25.170
And Europe also has been reeling from this, trying to play it, you know, all you it's weird to talk about Europe like Nasser's Egypt, but in some ways that is the situation that finds itself, like trying to pivot between you know, traditional allegiance with the US, the need for NATO to exist, fear Russia, etc., with making peons into China, trying to get more efficient development with local sovereignty, etc.
01:15:25.329 --> 01:15:27.649
And those are coming directly into a head.
01:15:27.969 --> 01:15:42.689
Although ironically, I actually think the US even benefits from decoupling because we will help provide, even if indirectly, a lot of the resources for rearmaments and stuff like that, like you see in Germany.
01:15:42.929 --> 01:15:49.809
I bring this up though, because it does seem like we have in the United States some capacity to intervene right now.
01:15:52.049 --> 01:16:04.529
But as we saw with the Biden administration, and bring something back that you brought out, the Biden administration was both heavy-handed and supporting some of these tech communities, but also heavy-handed in like coming down on them.
01:16:04.689 --> 01:16:10.769
And that did seem to a lot of these tech oligarchs seeming to the public to change sides.
01:16:11.489 --> 01:16:18.689
And then you have a lot of media in our country being bought up by right-wing billionaires, just outright.
01:16:19.009 --> 01:16:23.250
And so it does seem like it's you have a public trust problem.
01:16:23.729 --> 01:16:27.090
How do you think a left should operate in that environment?
01:16:27.170 --> 01:16:34.689
And then maybe can you contrast a situation in the UK which seems similarly dire, but very different?
01:16:36.449 --> 01:16:37.649
Yeah, yeah.
01:16:37.809 --> 01:16:49.569
I mean, it's hard to give a sort of general pronouncement, but I think you know there's massive discontentment with the establishment, as you say, throw the bastards out, sort of thing.
01:16:50.049 --> 01:16:54.609
But that doesn't just extend to the Republicans, it also extends to loads of the Democrats as well.
01:16:54.929 --> 01:17:11.889
And I think there is that sort of opening for people like Mamdani to come in and uh, you know, push the the parties in in different ways, in in much more socialist ways, which is unique.
01:17:12.130 --> 01:17:26.849
You know, the the the Democratic Party machine has been so dominant and solid and established for so long that you know to to have an upstart come in and uh change things in the ways that some of these people are, yeah.
01:17:27.090 --> 01:17:38.769
I think it is a real moment to dramatically change the party and and you know, given m the the nature of American politics, the centrality of the Democratic Party to everything.
01:17:40.210 --> 01:17:42.609
In the UK, I mean it's it's different.
01:17:45.729 --> 01:18:17.250
It's it's I mean, I'll say that I this was predicted by you know, not just myself, but pretty much everybody on the left here in the UK, that when Starmer came in power and there was loads of gloating amongst liberals and centrists about you know this overwhelming victory he had in the election, and you know, this sort of, you know, particularly after Corbyn saying, you know, this is what the Labour Party is all about.
01:18:17.409 --> 01:18:19.969
This shows that the centrist path is the right way.
01:18:20.130 --> 01:18:38.929
And everybody on the left basically saying if Starmer does not address the very fundamental issues that Corbyn was trying to address, and that the far right are a symptom of as well, that Starmer will end up you know plummeting in the polls and will lead to the rise of the far right.
01:18:39.009 --> 01:18:40.849
And that's exactly what has happened.
01:18:41.009 --> 01:18:50.049
You know, we've got the reform party, which is demolishing the Tory Party, we've got you know the Labour Party uh in the absolute pits in terms of in terms of polling.
01:18:50.529 --> 01:19:04.929
And Andy Burnham, which is coming in now, is you know the sort of you know, a maneuver sort of pioneered by the Conservative Party, which is like if your party's doing shit in the polls, don't go for an election, just change your leader.
01:19:05.170 --> 01:19:08.689
And the conservatives have done that for many, many years and very successfully done that.
01:19:08.769 --> 01:19:13.409
And the Labour Party is now doing that, hoping that Burnham will be their savior.
01:19:14.210 --> 01:19:15.489
And I'm I'm skeptical.
01:19:15.569 --> 01:19:25.889
I'm I'm quite skeptical that he will do anything significant to deviate from the path that Starmer and you know New Labor basically set out.
01:19:26.049 --> 01:19:31.250
And actually, no, New Labor was even more ambitious than what has been going on lately.
01:19:31.889 --> 01:19:42.689
So, in that sense, you know, the the party system here is different because you've got the Greens, you've got reform who are overturning that established party system.
01:19:43.729 --> 01:19:45.409
So, yeah, it's it's it's different here.
01:19:45.489 --> 01:19:53.809
And I think the risk for the left here is more about splitting attention between trying to save the Labour Party and supporting the Green Party.
01:19:54.769 --> 01:20:05.969
Whereas in the US, you know, it's it's there's one option and trying to take over that machine in terms of national power is is a very significant battle.
01:20:06.369 --> 01:20:12.689
Well, in the US, all change has been parasitic on an already existing party, and that's like true for the right, too.
01:20:12.849 --> 01:20:26.609
Like, I mean, I've seen the right the American liberals have not recognized that the white has been at a low-key hidden civil war with itself all of my life, actually, since since since the end of Reagan's second term.
01:20:27.009 --> 01:20:34.289
And well, not all my life, I guess it wasn't during the Reagan period, but like now you have something very different.
01:20:34.449 --> 01:20:37.649
I was gonna ask you a little bit, you know, you and I both have ties to Canada.
01:20:37.809 --> 01:21:09.409
I uh briefly, as a child, was a dual citizen, complications with my father, but we've seen a similar disappointment with uh Mark Carney, who was able to really ride and save the Liberal Party from what probably would have been a decimation until Trump started, for reasons that even we don't entirely understand in America, antagonizing Canada and the way that I would make off-color trollish jokes in my twenties.
01:21:10.929 --> 01:21:16.210
And that seemed to but you know by the Liberal Party sometime and led to the rise of Mark Carney.
01:21:16.289 --> 01:21:35.889
But Mark Carney's had a not a severe, but a similar problem to Starmer, in that he, you know, came in, he actually made some pretty bold statements about you know American hegemony and and all that and the need to start dealing with climate change.
01:21:36.210 --> 01:21:41.329
But in power, he's actually been pretty much an extension of the Trudeau policy.
01:21:43.489 --> 01:21:55.729
It does seem to me kind of interesting because in theory, in theory, it would be easier to change the situation in the UK because of the viability of uh of party coalitions.
01:21:55.969 --> 01:21:59.729
We don't have party coalitions legally at this point.
01:21:59.889 --> 01:22:09.889
You'd have to rewrite probably 35 of the 50 state constitutions to be able to even begin to consider something like that.
01:22:10.049 --> 01:22:15.170
Our parties are not in our national constitution, but they are in the state ones, blah, blah, blah, blah, blah.
01:22:16.929 --> 01:22:22.849
And there are legal constraints for us even acting like a like the Labor Party does in the UK, to be honest.
01:22:23.170 --> 01:22:30.369
Like, there's like party lines and ballot lines are really hard to enforce in the US, legally speaking.
01:22:30.609 --> 01:22:53.409
And so it does seem like weirdly, though, despite all those differences, we're in a slightly better place than you guys right now, despite the fact that it seems also that that Trump too, while not quite the apocalypse predicted by like some of the most concerned, has been far worse than even a lot of the a lot of the left thought it was going to be.
01:22:53.729 --> 01:23:07.569
We were we were mixed between those who thought this was gonna be a fascist takeover and a like a thousand-year Trump Reich, and those who thought it was going to be, you know, the last two years of the first Trump administration.
01:23:07.649 --> 01:23:08.849
And it has been neither.
01:23:09.009 --> 01:23:10.689
It has been quite bad though.
01:23:11.090 --> 01:23:18.049
And it does seem like the left here is like lost as to what to do, despite the fact we're being electorally successful.
01:23:18.210 --> 01:23:23.969
Like, beyond that, we don't really have we don't have an answer to climate change.
01:23:24.049 --> 01:23:28.210
We're we're like, we're not even in agreement with each other about how to handle AI.
01:23:29.649 --> 01:23:41.809
We need, you know, up until the AI thing, it was our great debate was between growth and degrowth socialism, which I I had gone on the record to say both are wrong.
01:23:41.889 --> 01:23:47.170
Um like like there's a substantive difference between socialism and capitalism.
01:23:47.329 --> 01:23:59.809
So growth uh growth models are are are not helpful, but also degrowth is actually dependent on you know measuring this by GDP or throughput, which is also not something we should be doing.
01:24:00.369 --> 01:24:03.090
And now even that debate's kind of gone away.
01:24:03.569 --> 01:24:15.409
Your your book was an interesting thing because it's got me thinking about, as I mentioned before, the divisions in these you know, these like private empires, basically, because they're fighting amongst themselves.
01:24:15.809 --> 01:24:27.090
And I guess my last question, and then this is a kind of follow-up on a question before, but um, I want to tie it into your to your work with Helen Hester.
01:24:28.689 --> 01:24:40.049
If we look at even the way capitalism is developing right now, like if you look at what the most job productive areas of capitalism are that are stable, there's they're all care work.
01:24:40.609 --> 01:24:49.409
Like, I'm not, even though we have a declining population, I'm not particularly worried about losing my job because I'm a teacher who also does ed tech.
01:24:50.049 --> 01:24:52.130
Like, I'm fine.
01:24:52.769 --> 01:25:02.289
But a lot, you know, and if you're in healthcare, despite seemingly a total assault by the Trump administration, uh, that seems like self-defeating almost.
01:25:02.449 --> 01:25:08.130
Like, like, you're why are you destroying your means of social reproduction, you fucking fools?
01:25:08.449 --> 01:25:09.889
We have that too.
01:25:10.210 --> 01:25:12.689
Could we tie together these two themes?
01:25:12.849 --> 01:25:22.210
Like, if we look at the the way AI has been deployed in everyday life, it's integrated our communications, you have a massive surveillance day that we didn't have before.
01:25:22.369 --> 01:25:26.849
Actually, as I pointed out, we have actually had it the whole time, it just they couldn't correlate it to any useful way.
01:25:26.929 --> 01:25:29.170
They've been trying to do this since the 90s.
01:25:32.130 --> 01:25:36.769
It doesn't seem like, however, this can truly touch reproductive labor.
01:25:37.090 --> 01:25:48.609
Like, it doesn't like it doesn't seem like it's gonna be able to replace doctors, it might make doctors way more efficient, which would be good actually, but it doesn't seem like it's gonna be able to replace doctors.
01:25:49.569 --> 01:26:04.369
Could we tie these two things together and look at this sphere of social reproduction as a place of actual leverage to also tie it to attacking these divisions within these tech oligarchs and playing them off each other?
01:26:04.689 --> 01:26:12.130
And I know this is really hypothetical and speculative, and feel free to be like, you don't know, because I don't know either.
01:26:12.449 --> 01:26:23.569
But is there a way that we can start tying these things together to think about okay, we have a we're already seeing that social reproduction is actually the prime is is the primary focus of capitalism, even right now.
01:26:24.529 --> 01:26:27.969
And we have these fragments over the welfare state.
01:26:28.049 --> 01:26:47.569
I was talking to the uh Mike McNair and he pointed out that like the the capitalists are in a weird place with the welfare state because if they get rid of it, they have a mass die-off, and then then they don't have a labor supply, and their labor supply has a lot of power, so they don't want that, but they also don't really want to pay for it either.
01:26:47.969 --> 01:26:54.210
And this seems like an area where we really could leverage things if we understood them better.
01:26:55.569 --> 01:27:14.769
How could we utilize this to get past, you know, to use your term from the past, this folk political habits that we have, which tend to be mana-chain, all or nothing, you know, either AI is the best thing ever and it's gonna lead to, you know, post-scarcity, super communism, but how I don't know, but it will.
01:27:15.009 --> 01:27:20.849
And then the other way is well, we need to like, I don't know, go back to cutting our grass with scissors or something.
01:27:21.170 --> 01:27:29.649
And it seems to me like you're actually pointing to throughout your work a way that we can think about this, but it isn't entirely clear yet.
01:27:29.729 --> 01:27:58.849
And so I wanted to give you a chance to like how can we tie the divisions between these tech oligarchs and their in and their internal private empires with looking at the way social production is right now and tying these together as a point of leverage for both normal political action like Mamdani and not normal political action, like developing co-ops, doing things that could leverage this, helping people set up legal frameworks to do their own and get past the hyperscalers, etc.
01:27:59.009 --> 01:27:59.170
etc.
01:27:59.329 --> 01:27:59.489
etc.
01:27:59.729 --> 01:28:01.649
Like, like, can we tie it together?
01:28:01.729 --> 01:28:03.729
And if we can, how would we do it?
01:28:05.409 --> 01:28:08.449
Yeah, I mean the honest answer is I don't know.
01:28:08.689 --> 01:28:09.009
Fair.
01:28:09.329 --> 01:28:16.130
I think it's uh it's it's a provocative sort of idea, but I'm not sure what the linkages exactly would be.
01:28:17.250 --> 01:28:38.449
I think I mean it depends how sci-fi you sort of want to get with futures of AI in terms of like for instance, how much of humanity is necessary to continue existing if there's just a handful of elites, you know, sort of running the show.
01:28:39.090 --> 01:28:40.769
Peter Deal does that all the time.
01:28:41.649 --> 01:28:42.769
Well, yeah, exactly.
01:28:42.849 --> 01:28:51.090
You know, the sort of question of social reproduction takes on a much more much less pressing question at that point in time.
01:28:51.250 --> 01:29:02.049
And it's just a matter of an economy and uh you know, an entire planet oriented towards the the utility of a handful of people, really.
01:29:02.369 --> 01:29:07.729
And whatever happens to everybody else is sort of inconsequential as far as they're concerned.
01:29:08.369 --> 01:29:42.369
I mean, my my so my my wife Helen Hester is doing interesting work on this sort of question in terms of the ways in which the tech elite are thinking about reproduction of their own, you know, of you know, as Elon Musk would put it, of the white race and the sort of compulsions which are driving them to focus on you know having so many children, as Elon Musk is, or investing in reproductive technologies and innovations in that sector in ways that are curious.
01:29:42.609 --> 01:29:55.329
You know, I think there is sort of interesting intersections between that question of social reproduction and the question of the tech oligarchs, but in terms of political leverage, yeah, I find it hard to sort of extract anything from that.
01:29:55.969 --> 01:29:58.369
I will just say I need to go as well.
01:29:58.769 --> 01:30:01.250
Okay, yeah, it's it's quite late here now.
01:30:01.729 --> 01:30:03.009
Thank you for your time.
01:30:03.170 --> 01:30:06.929
And I hope people are interested in picking up your work.
01:30:07.170 --> 01:30:11.329
I find it very illuminating to think about this, particularly in light of your other work.
01:30:11.489 --> 01:30:13.329
And I would endorse all your books.
01:30:13.489 --> 01:30:16.929
Where can people find your work if they want to find what you're doing that's new?
01:30:17.329 --> 01:30:17.889
I don't know.
01:30:17.969 --> 01:30:19.250
Where do people pirate books nowadays?
01:30:19.489 --> 01:30:21.729
LibGen, that sort of thing.
01:30:22.049 --> 01:30:22.849
Got it.
01:30:23.250 --> 01:30:23.809
All right.
01:30:23.969 --> 01:30:25.889
Thank you so much and have a good rest of your day.
01:30:27.090 --> 01:30:27.889
Good evening.
01:30:28.210 --> 01:30:29.009
Thank you much.
01:30:29.569 --> 01:30:30.289
Bye bye.