An advanced AI developed by Google DeepMind defeated the European champion of Go, an ancient Chinese strategy game. For those who follow the development of artificial intelligence, this achievement marks a highpoint in AI learning not expected to occur for at least another ten years.
Go is an extremely intricate strategy game where players place markers on a board, trying to enclose the largest possible territory on the board while simultaneously capturing opponents markers and protecting their own. Go is widely considered to be a far more complex game than chess. According to Phys.org, "But as simple as the rules are, Go is a game of profound complexity. There are 1,000,000,000,000, 000,000,000,000,000, 000,000,000,000,000, 000,000,000,000,000,000,000, 000,000,000,000,000,000,000,000, 000,000,000,000,000,000, 000,000,000,000,000,000,000,000,000,000, 000,000,000,000,000,000, 000,000,000, 000,000,000 possible positions."
That's a huge number, far too large for a computer to simply crunch numbers and determine the best possible outcome, which is why the learning and predictive nature of this AI is impressive.
AlphaGo, the name of this particular AI, was designed using an imitation of the human brain. "AlphaGo is programmed using so-called deep-neural networks, which are inspired by biological brains. The networks have millions of neuron-like connections that AlphaGo can rearrange as it plays. In essence, the program reprograms itself in order to "learn" the optimum strategy," reports NPR. The AI is actually constructed of two different types of these neural networks, one of which focuses on the best current move while the other determines which move will be most likely beneficial for later in the game.
The computer also demonstrated a learning capacity. While original programmed with many moves from games already played, the developers had AlphaGo play millions of games against itself, and it learned the best movements and strategies based on its experience.
The developers are extremely hopeful that the techniques used to develop AlphaGo can be transitioned into other fields, including climate change and medicine.
