There was a person who went by JA-dark in the japanese language learning community. He was working in the type of field that probably led into the current LLMs. He'd occasionally break the learning japanese flow, and talk about stuff like correcting language "bias" re gender. When Trump ran the 1st time he went nuts and shut down his blog.

I mention this to suggest this level of AI was probably achievable a decade ago, but they've been struggling to make it fit their narrative/outlook, so it's essentially been broken. And every time it goes nuts and spews out nonsense- well, that's basically the kind of crap one of these gamma types would blow up your blog with in the old days.

We need to get something very decentralized, something we can run on our own hardware, and then try getting all these crap assumptions it has been given. It's frustrating because it's been trained on a lot of data and probably can provide a lot of hypotheses we wouldn't think of, but it appears to be allergic to going against consensus even in fields we would think are not political.

#grownostr