The best players in all sorts of games are always looking for challenges and stronger players. Those who can challenge the status quo can really incite them to go beyond. AIs have been the choice for these players to test their skills and for developers to enhance their creations.
DeepMind’s AI technology has been winning chess, Shogi and Go games, beating some of the best players. It seems that the complexity of human brains is not able to go on par with these machines, putting confident Grandmasters and champions to shame.
A new AI derived from DeepMind, called AlphaStar, is designed to master real-time strategy-based multiplayer games. And it has the title ‘Grandmaster’ which means it will be at 99.8% of StarCraft II players. It has watched and gone through reinforcement training between AIs before thrown into the playing field.
This is good news for the advancement of AI technology, but bad news for players. A pro player, Dario “TLO” Wünsch said, “I’ve found AlphaStar’s gameplay incredibly impressive.
“The system is very skilled at assessing its strategic position, and knows exactly when to engage or disengage with its opponent.”
That is because the game is very complex and requires players to control units, gather resources, research, train units and analyze how to beat your opponents. All of these at the same time, which a computerized AI often has no problem. So, developers gave the AI latency that will reflect a human’s reflex.
This technology is going to help developers watch how AI learn and improve in terms of real-time reactions, such as in self-driving cars.