AlphaGo software no longer plays tournaments
After a smooth 3-0 win against the current number one in the Asian board game Go, Google’s AlphaGo software will no longer play tournaments. The team behind it should now concentrate on more general algorithms that should help humanity, said the co-founder and head of the developer company DeepMind, Demis Hassabis, after the match at the weekend. This included medicines for illnesses, lowering energy consumption and inventing new materials.
During the tournament, the Chinese Ke Jie tried to throw AlphaGo off course with unusual moves. According to experts, at times it looked quite good for him with this tactic – but in the end the software wrestled him down every time. “We have always believed in the potential of artificial intelligence to help society discover and benefit from knowledge – and AlphaGo gives us an early look at what is possible,” said Hassabis.
In March of last year AlphaGo defeated the South Korean champion Lee Sedol 4-1 and caused a stir. Because Go was long considered too complex for computers. The game, in which white and black pieces are alternately placed on the board, has many more possible combinations than chess – too many to calculate completely. AlphaGo therefore calculates the moves a human would likely play and only focuses on them. DeepMind first fed the program tens of millions of moves from games by human masters. Then the software played millions of games against itself – and analyzed which moves brought them to their destination. 50 of these games are now to be published so that the Go player community can evaluate them. Google bought DeepMind a good three years ago for £ 400 million. Hassabis had always emphasized that he wanted to use the company’s artificial intel