News for hackers. Stuff that matters.
Re:Algorithms for Reinforcement Learning by onorton on Tuesday February 19, 2019 @05:07AM (#19197909)
Maybe best foreigner (non-Korean) Protoss player. The match that AlphaStar lost had actions restricted to a single screen. Before AlphaStar could see and control everything that was not in the fog of war simultaneously. Still quite impressive though.
Re:Algorithms for Reinforcement Learning by AlexTelon on Tuesday February 19, 2019 @06:29AM (#19198210)
I agree with you and would like to add that he is probably not the best non-korean Protoss either. The blog presenting AlphaStar itself only claims he is one of the worlds strongest players and link to this post: https://liquipedia.net/starcraft2/2018_StarCraft_II_World_Ch... showing Mana at 5th best non-korean Protoss players in the World Championship series.

AlphaStar has gotten critisism for it having unfair advantages. It was 5-0 against Mana when it could see and control the whole map at once for example. But after a camera-restriction was given (so it sees the map like humans) it lost 0-1.

With all this said, it is still impressive. Best bot we have by far in sc2!

Re:Algorithms for Reinforcement Learning by yonkshi on Tuesday February 19, 2019 @08:13AM (#19198691)
Less than four years ago our state-of-the-art RL system (DQN) could only beat some Atari games. Now we can almost beat best human player in SC. That to me is very impressive.