I mean, it’s a little arbitrary to say “AI begins here”. They’re trained off datasets, but I would say that, for example, OCR is very successful, has been around for a while, and definitely uses machine learning.

Ditto for speech recognition.

On the other hand, none of these are capable of generalized problem-solving, either, AGI, which is really the sort of thing that I’m usually thinking of as being significant.

Personally I am hopeful about it. For a long time technology has been a factor in a bigger slow motion collapse of the economy being viable to sustain the lives of regular people, and the recent breakthroughs in machine learning do contribute to that collapse. But it’s not totally monopolized, local models are becoming viable and a lot of the work is open sourced, so the power of this stuff goes to anyone who wants it and has an idea of what they would want to do with it. Massive change is inevitable, the question is just what direction it goes in.

There’s something disturbing about how we were never going to be able to help ourselves & are compelled to make things like LLMs, but can still be frightened by its implications.

Don’t be frightened by LLMs. LLMs are the latest bitcoin-like fiasco—the latest in a long, proud tradition of bullshit technologies sold by bullshit human beings. It’s crap technology with no real use case. On the surface, if you just casually look them over, they look impressive, but the more you look into them, the less impressive and useful they seem and the more hollow the promises of the people saying “BUT THE NEXT GENERATION WILL BE AWESOME” sound.

Bitcoin/blockchain/yaddayaddayadda seemed unstoppable … until it became an obvious pile of scams, grifts, and failure. LLMs are following that graph on speed-walk.