Monday , June 21 2021

DeepMind interrupts StarCraft II professionals

DeepMind's artificial intelligence, which Google began developing in 2016, was an extremely successful show against professional players in the results of the games posted on YouTube and Twitch.

We do not know if artificial intelligence can take over the world one day, but it can definitely take StarCraft II. Against two professional players in Blizzard's StarCraft, DeepMind managed to win 10 out of 11 games.

The artificial intelligence, called kaybet AlphaStar and the DeepMind team, won 5 games without losing all the games in two separate series against Team Liquid Grzegorz and MaNa of Komincz and Dario TLO, Wünsch. Genoese "Manna of Kominz could win the victory of mankind.

The matches were played in December 2018, and the results were announced recently. Wynsh said in a statement that he had seen several images of AlphaStar's game and that he had started the game with a lot of confidence despite the fact that he had received the race of the Protoss without the main choice. However, each player encountered various obstacles and unusual strategies, he had a lot of rhythmic skills and rivalry.


Google DeepMind's artificial intellect wins the challenge of protein folding

Although Cominz's matches were similar, Artificial Intelligence was similar to a match between people because he had already acquired new information in their matches with Wynn.

The DeepMind team has announced that there are actually 5 different agents at AlphaStar. In other words, five different models of artificial intelligence, each with their own learning points, are collected and AlphaStar begins to fight. Players always struggle with another agent.

AlphaStar also sees "a game different from humans. While we should share our interest in meat bags like a mini-card and in the game map, AlphaStar should not do that. However, AlphaStar changes its focus 30 times per minute, similar to its rivals.


Google's AI can now learn without human help

He is quite surprised by the performance of AlphaStar in what is supposed to be an asset. AlphaStar, which is expected to be profitable in terms of click time, response time, has reached 280 in APM statistics. There is a delay of 350 milliseconds between the observation times and actions. From the point of view of statistics, artificial intelligence remains behind professional players.

The DeepMind team says AlphaStar's success is a basser because of its high ability to make macro and micro-strategic decisions, not because of its superior pushing speed, faster response time, or the deep Deep interface. Alphabet league matches created by artificial intelligence versions with different learning points correspond to a total of 200 years.

In the show of yesterday's show, Komincz managed to beat the new version of AlphaStar, based on how people use the camera during the game. The DeepMind team explained that this artificial intelligence is as good as the artificial intellect that almost uses the harsh Deep interface. Cominz says the game against artificial intelligence provides a new perspective on the game. Komincz, AlphaStar in different strategies in every game and advanced moves, says he finds it very impressive. I have seen how much my gameplay is based on pushing people into error and taking advantage of human reactions, which gave me a new idea, "Kominz says.

Kaynak : Google’ın DeepMind Yapay Zekası, StarCraft II’de Profesyonel Oyuncuları Ağır Tokatladı

Source link