U tweetove putem weba ili aplikacija drugih proizvođača možete dodati podatke o lokaciji, kao što su grad ili točna lokacija. Povijest lokacija tweetova uvijek možete izbrisati. Saznajte više
this is absolutely nuts. The AI GPT-2 has learned to play chess moderately well (able to give bad human amateurs a game) – despite only being a text AI, learning from a corpus of chess notation text, and not having any concept of what a chessboard ishttps://slatestarcodex.com/2020/01/06/a-very-unlikely-chess-game/ …
re people saying this isn't so exciting because it's just repeating old games & makes a lot of illegal moves: I get that, but 1) it's not JUST repeating old games and 2) it's like the Dr Johnson line about the dog on its hind legs, it's not that it's good but that it works at all
and apparently there is a better version! I dunno, it seems odd to me not to be quite surprised that an AI that literally just predicts text can get reasonably far towards playing a game of chess. I don't possess the world-weariness required to find that boring or unimpressivehttps://twitter.com/gwern/status/1214603658144485377 …
This is bonkers
did you see his earlier piece on GPT-2 as a step towards general AI? https://slatestarcodex.com/2019/02/19/gpt-2-as-step-toward-general-intelligence/ … I think it's really interesting as an idea that you could get this general pattern-recogniser and build quite complex intelligences on top of it
People talk about the difference between today's ML and "real intelligence". I often wonder if we have "real intelligence" in the way these people imagine...
that's the conversation I keep having! "It doesn't *truly* understand". But I don't know what that means, or how we'd know, or whether we even *truly* understand anything or just move symbols around in ways that turn out useful results. If it wins chess games it's doing SOMETHING
In my eyes this only shows that there is a pretty good, extensive corpus of chess games.
If you feed it cooking recipes it will create OK-ish recipes in some cases. Which doesn't mean it learned how to cook...
It's just reproducing most likely corpus elements (see @GaryMarcus)
exactly! don't you find it interesting that a totally un-chess-oriented AI can get from text to moderately OK chess playing like this? Yes it's just simple tricks! But those simple tricks build up complex behaviour. Which must be what happens with us, too. I find it fascinating
Apparently, we aren't even the best GPT-2-chess model, someone already made a better one: https://www.reddit.com/r/slatestarcodex/comments/el87vo/a_very_unlikely_chess_game/fdh0vqd/ … Who knew?
Absolutely bonkers
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
