In March 2016 AlphaGo computer program defeated the top human player Lee Sedol. During their match AlphaGo was much more powerful than during the games it had played against Fen Hui a few months prior. This was because AlphaGo had been playing a lot of practice games over that period. About two million a day to be precise. And it played them against itself.
AlphaGo utilises a programming feature mimicking the way human brain functions known as neural networks. Without going into too much detail it has two modules assessing the current board position which it used to work out the best move. But more importantly, AlphaGo can learn from the outcome of the games it plays against itself. This way it evolves much like humans do – it gets better at playing Go every single day.
When AlphaGo defeated Fan Hui in 2015 the experts predicted that a world beating Go playing computer program was still a decade away. In fact it only took under a year which shows that the progress in the development of artificial intelligence surpasses all expectations. The next challenge for computers is the Turing test – being able to pass as a human in a casual conversation. I doubt if it takes another decade.