Re:Artificial intelligence is losing hype
by MaybiusStrip on Tuesday August 20, 2024 @09:28PM
(#41305878)
I'm confused, you said earlier that you use it every day.
Re:Artificial intelligence is losing hype
by codexon on Tuesday August 20, 2024 @10:45PM
(#41306292)
Where did I say I didn't?
Re:Artificial intelligence is losing hype
by fragmede on Wednesday August 21, 2024 @06:19AM
(#41308740)
I mean, call it dot producting if you don't want to use the word reason, but it takes a bit of multiplying matricies and converting tokens to words to be able to say that if b follows a, it doesn't mean that a follows b, but that not b follows not a (the contrapositive). sure, that's in an unknown amount of textbooks in the training corpus, but at what point does putting down a bunch of axioms and applying them ok each other lead to what you'd actually call reasoning? at some point when a human does it, the human's gone past its training corpus and entered the hallowed realm of reasoning, unless we're able to deliniate clearly where that line is and isn't, and then all agree on it, we're going to keep going in circles about if it can be called reasoning or not.
Many people here argue that LLMs are able to reason, you can check my post history 3 months ago to see an example of this.
And if you have to double check everything, how much time are you really saving? And is this thing really on track to replace programmers any time soon? Absolutely not.