DeepMind developed Algorithm Artificial Intelligence called MuZero is able to master Go, chess, and video games without giving it rules, which represents another major breakthrough for the AI department in a company Google.
The MuZero algorithm follows the AlphaGo algorithm, which was the first computer program to beat a human hero in the complex board game Go.
And in 2016 DeepMind pioneered the AlphaGo AI algorithm that outperforms the best human players in Go.
DeepMind continued to improve its work and introduced one year later the AlphaGo Zero algorithm, which learned from the AlphaGo algorithm to play Go by observing amateur and professional matches.
AlphaGo Zero perfected the old game by simply playing against itself. DeepMind then introduced its AlphaZero algorithm, which was able to play Go and chess using a single algorithm.
The defining feature of these algorithms was that they know the rules of games that they must master while training.
And there was no need to tell MuZero’s algorithm about the rules of go, chess, shogi, and a host of Atari games to master, as it learned it all on its own and without any human data or grammar knowledge.
The researchers described the new algorithm as an important step forward in the pursuit of general-purpose algorithms. And theWas published The MuZero algorithm details in a paper in the scientific journal Nature.
Demis Hassabis, CEO and founder of DeepMind, said: “This brings us closer to having algorithms that are able to deal with chaos and complexity in the real world.
Prominent AI experts praised the London-based startup, which Google acquired in 2014 for $ 500 million for its leadership in developing AI at the human level, after it had achieved a number of key achievements in this field.
The company has not yet made a profit, and earlier this month it announced annual losses of 477 million pounds in 2019, up from 470 million pounds in 2018, but Google is able to fill the liquidity drain.