The next game tackled by the Deep Blue machine learning team is Starcraft 2.
They did a live stream showing replays of a few of the matches that you can see over on Youtube, with caster play by play commentary.
The first set of 5 games, against LiquidTLO, were won 5-0 by alphaStar. The AI can currently only play Protoss vs Protoss matchups. And TLO is not a Protoss player.
They then brought in a pro player who is a Protoss player, but he also was swept 5-0 by alphaStar. That was LiquidMana, who in 2018 finished ranked #13 among Protoss players in WCS points.
It is very interesting to see what limitations, and advantages are programmed into alphaStar.
At the end of the video, there is a 1 game show match with LiquidMana, and he wins against a modified version of alphaStar that was given one additional limitation to better imitate how humans play.
The Deep Mind team plans to go forward with having the AI learn how to play any matchup involving all 3 races in the game: Protoss, Terran and Zerg.
They did a live stream showing replays of a few of the matches that you can see over on Youtube, with caster play by play commentary.
The first set of 5 games, against LiquidTLO, were won 5-0 by alphaStar. The AI can currently only play Protoss vs Protoss matchups. And TLO is not a Protoss player.
They then brought in a pro player who is a Protoss player, but he also was swept 5-0 by alphaStar. That was LiquidMana, who in 2018 finished ranked #13 among Protoss players in WCS points.
It is very interesting to see what limitations, and advantages are programmed into alphaStar.
At the end of the video, there is a 1 game show match with LiquidMana, and he wins against a modified version of alphaStar that was given one additional limitation to better imitate how humans play.
The Deep Mind team plans to go forward with having the AI learn how to play any matchup involving all 3 races in the game: Protoss, Terran and Zerg.
via International Skeptics Forum http://bit.ly/2S9W7eY
Aucun commentaire:
Enregistrer un commentaire