AlphaGo Zero

[flonnebonne]

AlphaGo lost 100-0 against the newest version of itself, AlphaGo Zero. The crazy thing is that Zero had no human input whatsoever--it just learned by playing itself. I am excited but also non-plussed. I can only imagine how pro go players are feeling. This machine leapt over a thousand years of human learning in a few days.

https://deepmind.com/blog/alphago-zero-learning-scratch/

Comments

[troisroyaumes]

Whoa I hadn't seen this news! Thanks for linking. I'm also impressed that the algorithm also manages to be much more efficient than the previous model.

[flonnebonne]

I can't wait to see the applications! Let's use this to solve problems! I am a little sad for the go world but chess still goes on, so eh.