2 ms·
Since I'm not a Go player but am a NetHack player, I got more personal appreciation out of BotHack's achievement from last year than from AlphaGo's victory. Tha
by schoen 11y ago
Since I'm not a Go player but am a NetHack player, I got more personal appreciation out of BotHack's achievement from last year than from AlphaGo's victory. That was also something that I hadn't expected to see for years to come; you can see from the bot comparison page
https://taeb.github.io/bots.html https://taeb.github.io/bots.html
that no other bot had really come anywhere close to ascending before BotHack pulled it off.
It's interesting to see that the winning approach was GOFAI
https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence https://en.wikipedia.org/wiki/Symbolic_artificial_intelligen...
-- that is, the author taught the bot human-level rules and concepts based on human knowledge about the game. It wasn't a deep learning or neural net system; there is specific code about what altars do, how to fight mind flayers, or whatever.
For instance, here's a hard-coded list of corpses to avoid eating.
https://github.com/krajj7/BotHack/blob/master/src/bothack/player.clj#L359 https://github.com/krajj7/BotHack/blob/master/src/bothack/pl...