You have to verify everything this thing proposes. It will be like having an apprentice or student that's always guessing the answer.
Unpredictable abilities emerging from large AI models
321–326 of 326 posts
Re: Unpredictable abilities emerging from large AI models
#322Earlier quoted context omitted.
My memory could be improved by connecting my brain to an external hard drive. Wiring them together, alas, is not just hard; we have absolutely no idea how.
We do have some idea how, most people just don't really want to deal with the nightmare of being augmented and the life changing consequences that come with it, on top of the risk.
Really?! Have we done it in mice?
Re: Unpredictable abilities emerging from large AI models
#323Earlier quoted context omitted.
We can't trust it. That's a pretty hard bound.
We can't trust the police, politicians, or businesses either, but we're still on their hands...
Fair. It's always a question of how much. If someday we can trust AI more than those groups (to do the kinds of things they do), that will be amazing.
Re: Unpredictable abilities emerging from large AI models
#324Earlier quoted context omitted.
It's not clear why this would be a fundamental limit rather than a design flaw that will eventually be solved.
It might get solved but we have no idea how. There's no (readable) db of facts and no (visible) logical processing to improve on.
To me "we have no idea how" != "this is not fixable" (and even "we have no idea how" even seems like a strong statement.)
Perhaps it's because I'm ignorant about the inner workings, but calling the problem "unfixable" so early in the evolution of LLMs seems like foolish certainty.
Re: Unpredictable abilities emerging from large AI models
#325Earlier quoted context omitted.
It might get solved but we have no idea how. There's no (readable) db of facts and no (visible) logical processing to improve on.
To me "we have no idea how" != "this is not fixable" (and even "we have no idea how" even seems like a strong statement.) Perhaps it's because I'm ignorant about the inner workings, but calling the problem "unfixable" so early in the evolution of LLMs seems like foolish certainty.
I pretty much agree. That wasn't me who called it unfixable. (This is a UI issue with HN that keeps coming up for me -- people being mistaken for the OP.)