 |
| MuZero by Google is an expert in the game and doesn't know the rules |
DeepMind has developed an artificial intelligence algorithm called MuZero that can master Go, chess, and video games without entering any rules. This is another major achievement for Google's AI division.
MuZero's algorithm follows AlphaGo's algorithm, and is the first computer program to defeat a human hero in the complex Go board game.
In 2016, DeepMind took the lead in introducing the AlphaGo AI algorithm, which outperforms the best Go players.
DeepMind continued to improve its work and after a year it launched the AlphaGo Zero algorithm, which learned from AlphaGo how to play Go by watching both amateur and professional games.
AlphaGo Zero perfected the old game simply by playing against itself. Then DeepMind introduced AlphaZero algorithm, which can play chess and chess with a single algorithm.
The defining feature of these algorithms is that they know the rules of the game that must be mastered during training.
Additionally, since there is no need to learn human data or grammatical knowledge, MuZero can learn everything on its own so that Go is not aware of Go's algorithm, Chess' MuZero. Shogi and many rules of the Atari game.
Researchers describe the new algorithm as an important step in pursuing a general algorithm. Details of the MuZero algorithm have been published in Nature.
“It brings us closer to algorithms that are able to deal with chaos and complexity in the real world,” said Demis Hasabis, CEO and founder of DeepMind.
Top AI experts have welcomed this startup in London. After achieving many important milestones in this field, Google acquired the company in 2014 for $ 500 million in recognition of the human development of artificial intelligence. Smart driving.
The company is not profitable yet. Earlier this month, it reported an annual loss of £ 477 million in 2019, which was higher than the loss of £ 470 million in 2018, but Google has managed to recoup the cash loss.