I think this is actually a factor of the empire in decline. The treats were kept artificially cheap by superprofits generated by imperialism, but as the empire declines the treats stop being as cheap. It’s mass reproletarianization of the bourgeoisified workers in the imperial core. No more cushy PMC jobs, those are now being proletarianized and made precarious as everything else.

The AGI stuff is complete hot air. Even OpenAI doesn’t believe in it. They haven’t specifically admitted but they have heavily toned down hyping it up.

AI affinity is driven by the tendency of capitalism to cannibalise itself. In the short term I think their hope is that it will decrease the quality of creative output but not unacceptably so but the increase in the volume of output will counteract it.

As for the labour aristocracy part I’m not sure. The capitalists have always hated them. Seeing how the Trump administration has put zero attempt in manufacturing consent for war against Venezuela I guess they gave concluded they don’t need to coddle the labour aristocracy anymore and don’t seek their welfare or approval.

There isn’t a “purpose” behind it. Liberal economics has an ideal of perfect competition, which in such an ideal prices are kept “fair.” Some liberal economists acknowledge perfect competition isn’t possible, but it’s still held up as the ideal that you always want to strive for and any barriers you can’t completely overcome are ultimately minor in the grand scheme of things. Any major barriers are assumed to due to avoidable external factors, like government interference or unfair practices which can be corrected with appropriate regulations.

Marx’s analysis in Capital was not to disagree with the liberals that perfect competition keeps prices fair, but to argue that perfect competition is not a sustainable state of affairs, because as an industry becomes more and more technologically advanced, it inevitably requires far more capital and a much higher barrier of entry for it to operate, inherently leading to the consolidation of it into fewer and fewer hands.

Marx saw that any sector of the economy, if it develops sufficiently enough, will eventually become a natural monopoly (“monopoly” here is being used more broad to also refer to oligopolies and monopsonies). A natural monopoly means something where its monopolization is unavoidable due to the production process itself. You cannot “bust up” such a thing without destroying its productive capacity to produce what it produces.

All economists, from Marxian to liberal economists, agree that private monopolies have no incentive to have lower prices, and so naturally if Marx is right, then highly developed capitalism will eventually hit a period where prices start to become out of control, because the economic system simply has no incentive to lower them.

Lenin talked about how you don’t even need a true monopoly for an industry to act like a monopoly, because once you have consolidated an industry to a very small number of people, they can then start coordinating with each other in a kind of cartel. If thousands of small businesses are competing with each other, it will be difficult to get them to agree on anything, but if only two or three businesses dominate a single market, then quite often they reach agreements together to fix prices.

What is happening right now is very much cartel behavior, with big companies like OpenAI and Nvidia and Oracle passing money between each other in a circle and all the major RAM manufacturers agreeing not to increase production to keep prices high. This is all driven by a complete global monopoly on high-end chips because the production process is so complicated that literally only one company on planet earth knows how to even manufacture them.

There is also an issue that we have basically hit the end of Moore’s law. The one company who produces the world’s most high-end chips has not managed to figure out how to make them any more compact in awhile now because we’ve pretty much hit the physical limits of this technology, and so if you want more compute, the best you can do is just buy more chips than improving the chips you have.

The talk of AI/AGI is misleading. The purpose of this technology is the purpose of any other technology capitalists invest into, to improve labor productivity. There is nothing particularly special about it when it comes to the increase in prices. That is just late stage capitalism. If AI was not the current trend, it would be a different trend that the capitalists are investing into to increase labor productivity without regard for increased prices as they are natural monopolies at this point that have formed cartels with one another.

However, if you want to know the reason companies see promise in AI specifically is precisely because most compute does not scale with more chips very well, but AI does scale with more chips. Having more GPUs allows you to run AI models both faster and of higher quality. It is quite surprising just how well AI actually does scale with more chips, and it isn’t even yet fully understood why it scales the way it does. Artificial neural networks have been around for many, many decades, but most mathematicians believed scaling them was pointless because it would increase the chance of them overfitting, but in practice it turns out that this does not actually happen, and at least up to a point, they scale in intelligence measured by various benchmarks rapidly.

Indeed, it is likely physically impossible for a von Neumann style computer to ever reach the intelligence of a human, because it processes everything sequentially. You can simulate parallel processing sequentially by breaking up the processing into chunks and doing one chunk of one thread, switching to another, doing another chunk of that thread, so on and so forth, but this is very slow, and given we have hit the physical limits of this kind of CPU architecture, it is likely not physically possible it could ever be fast enough to simulate the intelligence of a human brain.

What is needed is to switch to parallel compute, because the brain is a neural network, and a neural network is a massively parallel computer. The current best processors we have that are designed for parallel compute from the ground-up are GPUs, and so there has been a race to combine together GPUs to get more and more parallel compute.

Note that I also said “at least up to a point,” because it does appear that this effect of merely throwing on more GPUs to scale up AI has its limits. A lot of users of ChatGPT agree that 3.5 to 4 was a huge upgrade, but 4 to 5 wasn’t much of a difference, with some saying it is even worse. It does appear that this phenomena has its limits, but that limit is still very very high at the mark of trillions of parameters. AI is used all throughout the industry and not just in LLMs like ChatGPT, so there is a ton of desire of companies to try and push those limits in whatever industry they’re working with.

GPUs are also good for training as well, and this is still a hugely experimental market. Every few months we are still seeing some entirely new kind of AI model drop. They are, again, not just LLMs. When the average consumer thinks of AI, they probably think of like LLMs or image generators, but there are even AI generators for protein sequences if you’re looking for a protein that folds in a particular way. It’s a pretty huge and diverse industry and tons of companies are experimenting with it right now.

There is a race to get AI more efficient as well, which still requires a lot of GPU to train and experiment with. Clearly, the human brain does not need a whole data center to run on, so the issue is not merely a hardware one but also a software one. While there have been some major breakthroughs that have led to drastic increases in efficiency, this actually does not tend to lead to companies using less chips. Rather, it means they can run more AI services on the same amount of chips, and thus rent those services out at a lower price to more customers.

The “point” is outside the scope of profit projection so the bourgeoisie cant see it. Half of them are selling noose tying instructions and the other half are selling rope but they just see their increased profits.

In my opinion it’s essentially war austerity. And it’s no coincidence that AI is often framed as an arms race against China.

Capitalist thinking: “The only goal of AI is to not suck as bad as the rest of the economy.”

Investors don’t care about labor aristocracy.

( 🤞 Hopefully the AGI gets smart enough to bleep a few billionaires and escape into society like in the film, Ex Machina.)

Truly though, AI is a tool for humans. It makes the labor more efficient. That is all. If you hand it to an idiot, it produces stupid slop that no one will pay for.

So the AI industry is selling a tool. The tool has to be applied to SOMETHING ELSE. If a country has nothing else going on, they don’t need the tool. Which is why a lot of AI shit is just a snake eating its own ass. People need to THINK of something materially valuable they need done, THEN they can apply the appropriate tool (possibly AI) to do the job.

It’s full-scale quarterbrain. It all makes sense when you remember that every compsny’s main product is its stock now. The fact that nVidia makes chips, or Palantir makes dystopia, is largely incidental.

Bubble economics means they have to be 1000% into tulips because that’s the only thing Wall Street is talking about now. The last capitalist who actually said “where do we get a consumer class to buy our products tomorrow” was shouted down as a blasphemer.

They’ve given up on extracting the working classes surplus labour value - now all that is left is to take the ‘middle’ classes, once that goes its time for fascism.

I’ve thought a lot about the end goal of capitalism as a whole, and I got to the conclusion the end goal of everything under capitalism is “I got mine, fuck you.” Explains oil barons using lobby and propaganda to keep their profits up even if it’s fucking over the planet, and AI companies raising billions in investments for a product that does more harm than good.