The superiority of the artificial intelligence of Google to humans in multi-player games, so after that I started a company DeepMind, the British – owned Google – specialized in the field of artificial intelligence in dealing with a different class of games represented games multiple players – that would normally require collective action.
Became artificial intelligence systems able to defeat humans in individual games, such as chess, after that the beat System Deep Blue Deep Blue from IBM to Gary Kasparov in 1997, cover system Google artificial intelligence in 2017 the first player in the world in the game of Go.
The multi-player games constitute a task of hard on the artificial intelligence planning because of the need for collective action and interaction in complex environments.
The separation of a group of researchers in a company DeepMind – through Paper Research Journal published by Science – how to design a system of artificial intelligence capable of mastering the stage of capture the flag Capture the Flag in a game of Quake III, where the use of the system agents and artificial intelligence – they taught themselves how to play the game – to represent players in the game, and play with or against adversaries and humans.
Were trained agents of artificial intelligence through 450 thousand rounds of capture the flag, and in spite of the need of a human element to approximately four years of play to complete the 450 thousand rounds of capture the flag, except that the artificial intelligence to complete the tour in a few weeks.
Related topics what you read now:
And the team of researchers led by Max are Max Jaderberg on a modified version of the game Quake III Arena, a shooting game released for the first time in 1999, but continued to flourish in the world of e-sports.
The phase in which they chose to stage a capture the flag Capture the Flag, which includes teamwork with the rest of the team to get the flag of the opponent team with the protection of your knowledge, forcing players to develop complex strategies that combine aggression and defense.
Wrote the team of researchers: the panel was unable to the human race – after 12 hours of training – to win by only 25 percent of matches against a team of artificial intelligence, has remained the performance of the agents artificial intelligence superior even when slowed reaction times them artificially to the levels of human.
Clients had artificial intelligence they speak the random movements at first, but after playing over and over again, they eventually discover strategies and techniques, and use team games – such as the stage of capture-the-Flag – been difficult to explore general concepts such as planning, strategy, memory, which thinks that it is necessary to develop algorithms that can be used to help solve the problem of the real world.
Artificial intelligence beats human in a game of Quake III