6 ms·

i dunno... i just read about alphago zero. Only needed to teach it the rules of the game, and it thrashed the current level of cumulative human expertise (alph

i dunno... i just read about alphago zero. Only needed to teach it the rules of the game, and it thrashed the current level of cumulative human expertise (alpha go) 100-0 matches, which had just thrashed a single human the year before. This seems like its happening at some hedge funds as well for stock picking. I guess the closed system of the game rules is more manageable than all of medicine, but... I can't agree with "stuck as the current level of expertise".

knicholes 9y ago
I wasn't able to determine from what I read yesterday if AlphaGo Zero played vs AlphaGo or vs itself (AlphaGo Zero). This probably isn't the place to ask, but if it is, do you know?
- nschucher 9y ago
  It was trained with self-play and evaluated on AlphaGo Lee (the version that beat Lee Sedol) and AlphaGo Master (the version that beat Ke Jie and I believe a number of online matches against top-ranked professional players).