人家已经明说有 18 个 AlphaGo 版本了,和李对战的是第 18 个版本,并且是单机模式。 此前 Google 一直在用 18 个 AlphaGo 做相互之间的对战训练的。。 AlphaGo 虽然是 AI ,但是是无形的,就像机器宝宝,最终那个小盒子也不过是它临时的容身之所罢了,本质上就是个应用。
hahastudio
2016-03-14 11:16:10 +08:00
https://deepmind.com/alpha-go.html "..., and by reinforcement learning from games of self-play." "This program was based on general-purpose AI methods, using deep neural networks to mimic expert players, and further improving the program by learning from games played against itself."