Google DeepMind’s AlphaGo program refers to the Artificial Intelligence program built with an advanced system and deep neural networks with machine learning. The program was pitted against human intelligence in the ancient Chinese board game, Go, which has beforehand been deemed impossible for a computer to play at a world-class level due to the high level of intuition required to master its intricate strategies. The program has however proved skeptics wrong as it attained a stunning victory over legendary Go player Lee Se-dol in what was meant to be a 5 time match that currently stands at a score of 3-1, with 3 victories for AlphaGo and 1 for Lee Se-dol, but why is this a big deal?
Go which is known as Weigi in China, Igo in Japan and Baduk in Korea, creates a challenge of extraordinary depth and nuance; it is in fact a game of strategy played across a 19 x 19 grid which involves players taking turns in placing black and white stones to surround points on the grid and capture their opponent’s territory. The game is thought to be far more complex than chess with DeepMind’s Artificial Intelligence researchers estimating that whereas chess offers 20 possible choices in every move, Go offers about 200. Due to that fact, machines could not be taught to play Go by mere straightforward algorithms, they needed to be taught to create their own strategies.
DeepMind currently uses two deep-learning algorithms in AlphaGo. There is the “policy network,” which is trained to imitate human play by watching millions of games, and the “value network,” which goes on to evaluate how strong each move is. In this way, the program is constantly playing against itself, learning new ways to evaluate its opponents moves and developing its own methods. With the shocking victories against 18 time world champion Lee Se-dol which were successive wins, AlphaGo shows the amazing progress on the fronts of Artificial Intelligence.
Even Lee Se-dol’s eventual victory, presents yet another side to the fascinating technological journey as it was attributed to mistakes made by the program. The concept of mistakes is a very human one and shows that AlphaGo is indeed a little human-like. DeepMind which believes that the principles applied to AlphaGo could be useful later in things like; robotics and smartphone assistant systems will certainly present more fascinating work on Artificial intelligence down the line.
So is anyone else thinking machines taking over the world? That would obviously be a little melodramatic but with this victories, DeepMind has indeed proved that Artificial intelligence is growing at an amazing pace and soon enough, all those science fiction movies may not seem so out of this world again.