Comment on Welp ...
brucethemoose@lemmy.world 4 months agoIts far worse because LLMs are so data hungry. Getting quality data for image diffusion models is not nearly as much of an issue, though still a problem.
Comment on Welp ...
brucethemoose@lemmy.world 4 months agoIts far worse because LLMs are so data hungry. Getting quality data for image diffusion models is not nearly as much of an issue, though still a problem.
YourPrivatHater@ani.social 4 months ago
You can probably limit the data intake of the image ones pretty good, with LLMs its worse yes. But since the entire shit is a bouble and a environmental catastrophe im happy when they all fail for the better, at least the big ones. The open source guys can still tinker on their hobby of course, maybe some day we get a actual AI that is actually useful.
brucethemoose@lemmy.world 4 months ago
Nah. I hate to sound bitter/salty, but all the AI haters are just going to fuel OpenAI’s crusade/lobbying against open source, and we will be stuck with corporate API models trained on copyrighted material in secret because the corporations literally don’t care. And it will do nothing to solve the environmental problems.
There’s tons of research on making training and especially inference more power efficient, on making data cleaner and fairer, and it’s getting squandered from the lobbying against open source that the “AI is all bad” crowd is fueling.