You are either not logged in or not registered with our community. Click here to register.
 
December 04, 2016, 04:19:40 PM

Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length

Click here if you are having problems.
Default Wide Screen Beige Lilac Rainbow Black & Blue October Send us your theme!

Hark!  The Herald!
Holiday Issue 2016

Wiki Blogs Dicebot

Author Topic: AI can play Go at a professional level  (Read 436 times)

0 Members and 1 Guest are viewing this topic.

Offline TheGlyphstoneTopic starter

AI can play Go at a professional level
« on: January 29, 2016, 01:15:30 AM »
http://gizmodo.com/google-just-beat-facebook-in-race-to-artificial-intelli-1755435478

Interesting milestone in AI development. Go being ridiculously more complicated than chess in its number of possible solution trees (10^700, over 10^60), and they've finally worked out a learning algorithm that can process that level of complexity.

Offline Blank

Re: AI can play Go at a professional level
« Reply #1 on: February 04, 2016, 10:36:57 PM »
Guh, Google is disgusting with how much computing power they have.  This aside, their capability to farm out computational problems is insane.  They have a scholarly article out there about their mapreduce farms that just boggles.

Offline Blank

Re: AI can play Go at a professional level
« Reply #2 on: February 20, 2016, 09:13:53 PM »
So I read through the paper that Google published to Nature.  It's really cool.  They basically used two Neural Nets (Policy and Value NNs) and used the Monte Carlo Tree Search (used for solving other supposedly intractable games like chess) to beat top tier players and go players.  AlphaGo beat a high tier player in europe five out of five games.