Comment on ELI5. Limit of current gen AI/LLMs

brucethemoose@lemmy.world ⁨1⁩ ⁨week⁩ ago

Others have explained it well; splitting calls up and programatic prompt engineering.

And what is the limit of AI non theortical limit of AI?

Who knows?

But practically, transformer models are kinda hitting an “innovation” wall. Big companies aren’t taking risks to try and fix (say) the necessity of temperature to literally randomize outputs, or splitting instructions/context/output, or self correction (like an undo token), self learning, anything. All this has been explored in papers, but they aren’t even trying it.

Their development is way more conservative than you’d think, and that’s the wall LLMs are smacking into.

source
Sort:hotnewtop