Stronger than Lee Setol 3 days after having learned the rules !
Who could do this ? AlphaGo of course !
The Deep Mind team has just released a new version of AlphaGo, AlphaGo Zero.
This new version is stronger than the previous versions (AG Zéro won...100-0 against the AG LeeSetol version !).
But now comes the most amazing info...
Why did they named it "Zero" ?
Because, contrarily to the previous versions that has been "fed" by thousands of games played by the humans since centuries, AG Zero version has learned everything from scratch, from just learning the rules.
Just after 3 (yes, three !) days of playing against itself and learning from its own mistakes, without any pollution by human play (!), AG Zero has already reached the level of the LeeSetol version; and after 40 days, s/he - it surpassed the KeJie version, to become the best ever playing entity in this world.
More than even, IA will be modelling our world, and not only in go !
Read more here: (I recommend you to watch the graphic and the video !): https://deepmind.com/blog/alphago-zero-learning-scratch/
PS: the Deep Mind team opened a forum
https://www.reddit.com/r/MachineLearning/comments/76xjb5/ama_we_are_davi...