China built one under water, and it uses wind and solar, but I totally agree with you. AND, US companies like OpenAI, should not be able to build them in Canada, like they want to.

If it was an AI that actually understood what it's doing (I.e. Something based on advanced neuromorphic architectures) it would be a boon easily since it would be like a genuinely intelligent web. Ironically more like the OG dream of the world wide web. That isn't this. This is "AI Slop as Browser"

Should be stressed Aurore didn't wake up on the woke side of the bed and tweet "hey peasants, #eattherich" This was the end result of an extreme character arc that one page couldn't cover, and the actual strike/revolution still required an organized working class and assistance from some clankers

I mean I know WHY. It's the DNC that's the harshest roadblock.

Funny thing about that too, we're so whipped up against socialism that actual socialists should exploit that and start running. For half the country, there's no difference between Chuck Schumer, Zohran Mamdani, Subcommandante Marcos, and Mao Zedong. So why bother running anymore

Yeah, I'm not thrilled about it. It's a death star for fascism.
bsky.app/profile/symb...

Sad part is, China's still researching the kinds of AI that could actually lead to that. Yet if they get to it first, the result might be automated socialism— exactly what we DON'T want over here in USica. While we keep doubling down on this venture capitalist money cheat code.

Saddest part is how AI bros act like Zionists. They DELIBERATELY associate generative AI = ALL AI, then get angry when people push back against generative AI slop saying "Don't you want to cure cancer and end poverty and have fusion?"

Yes, I do.
What does gen AI have to do with that???

And sad how they have ruined the promise of it. Like here's China's Darwin Monkey www.livescience.com/technology/c...

Uses 2 *kilowatts* and is as capable as it is without plagiarism. Probably they'll have automated socialism soon. Meanwhile we're building terawatt-chatbots for...

That last one is easy to explain -- they're imagining (or being told by Altman) that if they can just make their overhyped Clippy "good enough" they'll be able to replace three quarters of their workforce with an AI that doesn't sleep, doesn't take time off and doesn't complain about wages.

It's been known for a while that if we ever get AGI or anything like it, it'd be through neurosymbolic/neuromorphic AI, not generative AI.
But part of AI bro culture is literally, unironically, dedicated to mocking this.

What if I told you the answer is even dumber than that.
They CAN conceive of it. But they regularly shit on it, and whipped their followers into a cult like hatred of anything that isn't transformer scaling. They DELIBERATELY seek this bullshit because it makes more money.

Let me reword this. There is a path to an AI architecture that can do everything LLMs *purport* to do, but with actual understanding and ability to say "I don’t know." This just isn't what the American AI Bros are building. But they pretend they are.

It's just that everything here is so fucked up. The explanation is technical, and it doesn't change the cold fact generative AI is a psycho-capitalism at its most unhinged

I do believe there is a path to AGI and we've known what it is for a while. Hence why I say the Chinese will do it. This path we're on now is not it. I can even explain in detail why transformers aren't the way but also why they whipped USican investor capitalists into a Jim Jones-esque fuckfrenzy

In fact China's doing just this. It seems inevitable to me China will achieve something genuinely interesting while the West keeps trying to pump generative AI up for investors.

To say nothing of the Big Data approach likely being flawed from the outset. It seems increasingly likely to me that Big Data Scaling to AGI is like trying to achieve nuclear fusion with TNT. There's a reason a good number of people say neurosymbolic and neuromorphic architecture is the way

And the answer is no, you can't. Transformers by themselves inherently can't not hallucinate because they have no grounded internal state and operate through accurate token prediction, which mimics but is not the same as actual concept-based natural language understanding. It's a fake version of it

They're an inevitability because of the architecture they use and refuse to abandon (autoregressive attention-based transformers). The reasons why are technical, and well understood. But they think they can overcome it with scale and RL.