im going to double post oh my God I hate AI so much. LISTEN. I bet it's really nice to have certain intellectual tasks offloaded to the dead-eyed thing, but no dude. The marginally less destructive, subjectively better origin doesn't make this technology safe.

there are real dangers in letting the US govt have this tech because it will gleefully rot brains to control people and enact their so-called perfect panopticon

but that doesn't mean that by changing the social and intellectual role of this technology, it becomes more acceptable. The problem with the technology doesnt change just because we can envision a more ethical owner and distributor of said.

I'm slapping LLM devices out of the hands of Chinese people in China.

the technology as a consumer product is poison. It does not care if it is rotting the brain of a citizen of your nation or another. it does not care.

They're both selling a solvent based paint

the chinese solvents are notably less dangerous to the environment, but it still gets you wicked sick nasty high, like tripping balls

but it's still paint. You're still huffing paint fumes, you, yourself; emotionally, can allow yourself to enjoy your huffed paint, if you want, as more ethically sourced?

BUT YOU are STILL HUFFING PAINT

Honestly, my stance is it is ok if it is open source.

I dint think any AI slop is good

It sounds like you don't actually know the meaning and origin of "luddite". All of your concerns can be applied to the Spinning Jenny.

I'm less critical of Chinese AI because it exposes the full ass+hair of western AI companies. Their valuation is tied to making frontier models. The US economy is super concentrated around overproducing in the hopes that US frontier models will become the global standard. If Chinese companies can make a better model more efficiently, it undermines the whole business model of OpenAI and Anthropic. If those models are open source or run locally, it throws a wrench in American companies' attempts to price gouge customers locked into their ecosystems.

In the long-term I want AI as a marketing term to be as stupid as NFT, with iterative improvements to machine learning like we had for 20 years prior to it. In the short-term I want the US companies to implode because it will destroy the national economy and severely weaken the empire.

Similarly I don't think society should be car-centric but as long as it is I want those cars to be Chinese EVs because American automakers are the ones enforcing car dependency. BYD doesn't concentrate capital in the hands of people who can directly impact my life. It reduces the potential scale of American manufacturing to workshops making ebikes and velomobiles.

environmental costs

China's models are afaik much less environmentally intensive and cheaper to operate, which means they are actually undermining the capitalist models and providing an alternative for users which is much less destructive. Unfortunately people are going to use it whether we like it or not, so having a Chinese alternative which isn't destroying the planet nearly as much is the only alternative to the maximally destructive alternatives.

Related to that, China's models getting better quicker could help to burst the AI bubble sooner which would be the best thing to happen to slow the capitalist march towards destruction. The people running the capitalist models are openly evil freaks and the only way to stop them aside from a revolution is for Chinese models to outperform the capitalist models, take market share from them, and show that the capitalist models are just worse in every way.

China is also regulating AI on the social side in a very progressive and pro social way, so they are also building a legislative and social framework of how to ensure AI is not having such negative social impacts.

It feels similar to nuclear deterrence. I'd rather no one has nukes, but the worst scenario is the US being the only one to have it

I think we can all agree that China has refused to take the role of the Soviet Union in exporting socialism to the world

I don't think that's a given. Exporting socialism is a lost cause when you don't even have a single nuclear-powered aircraft carrier, or a fleet of strategic airlifting planes. It can only meaningfully happen after the inevitable climactic geopolitical confrontation with the US.

Let's not forget that China is still contained within the "First Island Chain."

China hasn't just refused to take the role of exporting socialism, that phrasing is actually from long-running explicit policy of the CPC and its leadership. It is an intentional strategy for their own nation-building project, though one should just focus on the material history.

As for LLMs, it's just another industry among many. What is uniquely harmful today is the intense tech-finance gambling on it, the so-called "hyperscalers", which just means huge amounts of resource use using finance capital to make bets on having monopolies on it later. Lose billions, trillions for 5 years and then ramp up prices when CEOs consider you "essential", or so goes the gambit. This is incredibly wasteful and most of it has no legitimate use case.

Chinese industry is being more coordinated and is far less impacted by finance capital and monopoly seeking. Both exist, including the latter, but the party is much more involved and can make corrections. In combination with their academic institutions, Chinese models tend to focus on attenuating the excesses of the "hyperscalers" by focusing on models that can be trained with lower power requirements and releasing open weights so they can be used widely on consumer grade hardware. The latter is fundamentally in opposition to monopolization at the point of consumption, though training is not and won't be without a distributed open computing approach. And while there is still a lot of pointless or obscene use of LLMs, they tend to be used in a way that is more pragmatic.

The situation is basically the same as for every industry in China. Its purpose is to grow productive forces in a way deemed strategic, a balancing act that requires constantly weighing different forces. They decided that sitting out LLMs was less strategic than using them and growing domestic expertise, but in their own strategic direction. This means the warts will also be there and a new round of contradictions. Actual bad things will happen, like wasting resources on a data center. But this also already happens in their general strategy of building productive forces and being enmeshed in export markets. They build most of the pointless consumer products you can think of because it builds their productive forces, keeps exports up, centralizes production so they are "too big to fail" geopolitically. And those productive forces have a domestic boon. Being the best die mold production center on the planet means they can produce almost anything at scale in almost any material. As a result, something like new batches of trains come together at a speed and scale that's mind-blowing. It can be built out to support the people, to support solar energy, to electrify, and so on. The boons of industrial capital with their worst excesses significantly diminished.

This is all true with their LLMs as well. Not a pure good, but a much better method and one chosen strategically to position themselves against the imperialists.

I think the major material point missing from this analysis is that every major model from China is "open-weight". This means that in principle you could run these models on your own hardware or the hardware of a third party.

Even though most people still use these models hosted by the provider anyways, the threat of high quality open-weight models has been a persistent drag on the valuations, prestige and revenue of the US tech/AI firms. In fact, they started a massive price war in token pricing that has reached all the way to the top of the market

Chinese firms are also a pressure relief valve for monopoly pricing in high bandwidth memory and soon even advanced chips.

I therefore argue that even without "exporting revolution", in practice Chinese industrial policy is hazardous to imperial tech conglomerates on both sides of the Pacific, and I couldn't be happier about it

I think we can all agree that China has refused to take the role of the Soviet Union

I believe China sees 2 storms coming and has chosen to bunker down and shelter it. I believe China has done the math and seen that nothing more they could do today would make anything better long term if they don't survive the storms. Once the US empire has fallen and Climate Change has been provably survived, all bets are off.

On some podcast or whatever someone said the west was using ai to cheat on school assigments, china to improve crop yields

I essentially fully agree with you, here. To be honest I'll forgive China for AI same way I'll overlook some real concerns about their censorship - I understand they have a need to be on top of it to avoid being totally screwed over by other countries. It's a bit of a necessary evil.

I still hate AI in its entirety, I've been to China since the proliferation of AI, and I don't think AI is doing them any favours either. At least from my limited experience using native services, I found slop and AI chatbots inserted into every last app and website as much as in the west.

Why should one be completely against AI?

Would it not be good to socialists have AI as a tool? The issue is capitalism being inefficient due to it focusing on profit too much, right?

the usage of ”exporting socialism” is funny here because the cold war didnt constitute of the USSR telling states to be socialist if they like. people died in droves in the name of socialism. socialism is not ”exported” it is fought for

I like seeing the US tech giants disgraced but the Chinese AI still has most of the same harms.

China is the world leader in renewable energy. This means they have significant competitive advantage for any productivity that utilizes electric power. They are also generally utilizing "AI" much more contextually then any of their competitors. There is miniscule belief in AGI within China compared with the capitalist AGI hype that continues to inflate the bubble.

There are many implications that come out of these differences, but the most important one in my perspective is that the CPC is not beholden to the interests of AI firms, unlike the entirety of the financial capital machine at this point.

Unfortunately China still allows businesses like this to exist, thankfully highly regulated. Hopefully they see that this is a short lived thing that will disappear once the bubble bursts. Regulating the fuck out of AI and keeping their wealthy on a right leash might save them from the fallout.

Also there is something to be said about allowing tightly controlled AI to be developed so that you can better protect yourself from its use by countries like the US.

the purpose of china as it exists is to reduce treatlerism levels of imperial core, open weights go in the same bin as cheap electric cars. e.g. buying a singular nvl72 and running glm is cheaper for a business than paying microslop at 50+ users, not to mention significantly safer.

I don't have too much to add about china specifically but it might be useful to explore if the "social and environmental costs" would be so high if it wasn't developing in a capitalist framework?

As far as I understand, a lot of the environmental/community harms are multiplied in a rush to hoard resources and build up capacity for which there isn't yet a need.

Because of the rush to bring capacity online (whether or will be used or not) and because infrastructure has been so neglected throughout the western world, the data centres are monopolizing power generation and transmission, leaving others to pick up the cost of expansion and maintenance.

Lack of power in places where land is cheap or where there are significant perks for businesses means on site power generation with things like diesel generators is seen as a valid solution.

The same situation arises for cooling. Water cooling is the cheapest in places where water is cheap and regulations are limited, but this is a detriment to infrastructure and the environment. Air cooling or closed loop liquid cooling has a much lower impact, but is more expensive and limits where you can build.

If there was no incentive to chase higher and higher valuations, would the same decisions be made?

To be clear, I have no idea how china is handling these problems. I would believe there is a sort of AI arms race going on as well, so perhaps removing capitalism wouldn't mean better, more sustainable choices are made.