Google acquired the British AI company DeepMind back in 2014 to stay ahead of this flourishing tech field. Their biggest achievement so far is the AlphaGo Zero, a digital Go champion.
The bot is already famous due to its matches against two world champions in Go which it won in 2016. However, Google now claims that the AI has achieved so much in the last year, that it devised innovative strategies by itself that humans have never thought of before.
AlphaGo Zero Needed only 40 Days to Become a Go Champion with No Preloaded Information about Chinese Chess
Go is a Chinese type of chess that has a 2,500-year old history. Two players compete against each other by surrounding more territory than the other can. The board game uses abstract strategies, and it is said that there are more possibilities of playing it than the number of atoms that exist in the visible universe.
AlphaGo Zero started to play the game from scratch. It wasn’t aware of all the strategies that were created along the 2,500-year old history of this type of chess.
Nonetheless, after only three days of studying it, DeepMind’s bot defeated AlphaGo Lee. This is an AI that previously beat world champion Lee Sedol last year. After 40 days of learning, AlphaGo brought all AI programs versed in Go to their knees.
DeepMind Team Intends to Assign the Bot Other Challenges to Improve Human Knowledge
Lead researcher of the AlphaGo project, David Silver, announced the latest updates on this artificial intelligence last week. He reiterated that the bot has never received tips on how humans play Go from creators. Instead, it learned them by itself.
However, as of lately the bot stopped using these ancient techniques that once mastered can bring prestige to a person. However, the bot replaced them with its own improved versions. No human has ever used these strategies in any recorded game.
“For us, the idea of AlphaGo is not to go out and defeat humans, but… for a program to be able to learn for itself what knowledge is.”
Now that the bot proved that it could become a world Go champion in one year and be able to come up with its own unbeatable solutions, DeepMind has other plans for it. Researchers intend to put AlphaGo Zero to other uses that can better serve humanity. Google wants the bot to contribute to human knowledge with its own input.
Image source: 1