>>513406738
>It's the same as how they didn't care about manufacturing getting offshored.
exactly
>The whole process will take longer than most think though.
well, we're obviously guesstimating
IMO there's a some clear stages ahead:
1. better more efficient hardware: GPU clusters are ghetto as fuck, and decent local basically requires almost triple digit vRAM
2. better LLM structures: LLMs are lagging compared to image/video, but it will catch up
3. specialization: massive LLMs that are good at everything seems retarded as fuck, like trying to create a multi-tool for professional construction work, the obvious path should be tailored specialization
4. once 1,2,3, are solved, then and only then, big tech will finally be able to package AI into products companies can actually adopt and use
we should be here around 2035, I think, pulling that date out of my autistic asshole