What’s worse than AI? Ai that is owned and completely controlled by oligarchs that are funding the worst things the world has ever seen and we right now have an opportunity to develop an effective and actually productive strategy to work against that future. So here are some of my starting notes on how to do that.
While I think his intentions may be somewhat pure, he’s starting from flawed premises.
Nobody reproduced Photoshop or Microsoft Office or whatever with LLMs. Those slop-coded monstrosities are mockups and/or tech demoes at best. ArtCraft, the company behind a lot of these so-called “clones”, calls all their software “early alpha” or “in development” on their own website (top right in the header). Turning them into something anywhere near as feature-rich and reliable as the Adobe/Microsoft/whatever applications they claim to “copy” would take more work than it would starting from scratch.
The problems with LLMs isn’t (just) that they funnel money to oligarchs. It’s that despite what the tech bros say constantly, LLMs can’t produce functional software.
He argues that they are useful and can be used to positive effect. His rationalle? “Would China be investing in it if it was really useless?” Even if you think “China’s using it” implies it’s geinuinely useful, has China actually used LLMs for something worthwhile? Or have they mostly just built LLMs to undermine western companies? There was an article posted to this community just today called “In China, the AI bubble has already burst” (no-paywall version) which makes it sound like China’s faith in “AI” has largely run out.
(Mind you, I can’t say I’m disappointed that Chinese companies have created very cheap models you can run locally. I don’t say that because I think LLMs are genuinely useful. I don’t think they are. I say that because the existence of cheaply-created models that run locally greatly undermines the credibility of the big AI companies like OpenAI and Anthropic even to people who think LLMs are useful. “Local LLMs are basically just as good as ChatGPT or Claude, so OpenAI and Anthropic don’t really have any future and will probably end up going out of business.” So goes the logic. And that’s part of why the bubble is faltering even among LLM boosters. And the bubble popping sooner is a good thing.)
It also bothers me that he keeps calling locally-runnable LLMs “Open Source”. It’s not really entirely his fault. It’s the way everyone is speaking anyway. But it’s just LLM companies approrpiating friendly-sounding language.
Basically I think anyone who has a problem and decides to use LLMs to solve it now has two problems. LLMs, local or otherwise, won’t be any boon for the left (or the right or anyone else, except maybe the tech bros who manage to find eager bag holders to use as exit liquidity before their companies crash and burn) any more than NFTs were.