Last October, behind closed doors, a 2,500 year old contest of strategy and intellect took place. The contest was the ancient Chinese game of Go. The contestants were a Go Grand Master versus a machine. A smart machine. European Go champion Fan Hui was up against AlphaGo a computer programme developed by Google DeepMind, the Artificial Intelligence (AI) arm of Google based in London.
AlphaGo won. Not once, but five times in a row. The significance of this win, for the AI community, has been likened to the chess defeat of Garry Kasparov by IBM’s Deep Blue computer in 1997 which became a milestone in AI research. In reality, this win is far more significant. In chess there are approximately 35 possible moves for a player during a single turn. In Go there are around 250. For each of the 250 moves there are a further 250 and so on. This effectively means there are more moves than atoms in the universe. Even the fastest supercomputer cannot calculate all of the conceivable moves. So a different approach was required, this is why using AI to win at Go is so significant.
AlphaGo uses basic computing power to build a ‘neural network’ of information about the game and probable moves. But over this it layers ‘deep reinforcement learning’ techniques by pitting one machine against another to play thousands of games and ‘learn’ the best ways to win. Clearly, the programme does not become conscious, but it can refer to a massive catalogue of probability in a fraction of a second.
Jon Diamond, President of the British Go Association said “Before this match, I was still expecting that it would be at least five or ten years before a program would be able to beat the top human players.”
Building a successful Artificial Intelligence has become the holy grail for all of the tech giants. Google, who are in the press today, chose their route through the acquisition of DeepMind for $400M in 2014. There is a smaller in-house team at Facebook that founder Mark Zuckerberg brags about, “The researcher who works on this, Yuandong Tian, sits about 20 feet from my desk.” he says. Microsoft and Twitter are both already in the game.
We are a long way from the horrors of Skynet’s Terminator, but we are well on the road to very smart applications and devices that will cause pause for thought in the not to distant future. For more information on the system and implications of this technology, there is a great article on Wired.com
