In October 2015, a program developed by Google DeepMind, called AlphaGo, succeeded in defeating Fan Hui, the European champion, with five wins to zero. This was the first time that a machine had beaten a professional go player, an outcome that many experts did not expect to see for a few decades yet. At the same time, Google won an important battle in the war of communication that opposes another web giant, Facebook, which is also working on a program that can beat the best players in the world at Go. But this was not enough for Google Deepmind, who decided to go further. Their artificial intelligence program is now battling it out with South Korea’s Lee Sedol, who is considered the best Go player in the world. The duel will end on 15 March, but four games out of five have already been played.
An historic victory for Alpha Go
The first four encounters have created a scoreline of 3-1 in favour of Alphago, the program having defeated Lee Se-dol in the first three matches. Lee Se-dol finally secured a win this Sunday, taking the tally to 3-1 This now inevitable victory in a best of five match, marks an important moment in the history of artificial intelligence, like that of the IBM supercomputer against Chess champion, Garry Kasparov in 1997. “Yesterday I was surprised, but today it’s more than that—I am quite speechless.” said Lee Sedol at the post-game press conference after losing his 2nd match. “I have to say, if you look at how the game was played, I admit that it was a very clear loss on my part. From the very beginning of the game … there was not a moment in time where I felt that that I was leading.“
Lee Sedol beats AlphaGo! – Credit Google Deep Mind
The holy grail of artificial intelligence
Despite appearances, the game Go is much more complex for a machine to try and decipher, compared to other games, particularly because of the large number of possible combinations. Strategy games are, since the earliest beginnings of artificial intelligence, a significant challenge for researchers, and are emblematic of what an intelligent person can accomplish. Almost all of the pioneers of computing have tried to make gaming systems. The challenge of beating a human at Go has been impossible to overcome until the arrival of deep neural networks (or deep learning,) which have in recent years, upset the field of artificial intelligence. These neural networks are actually a “virtual machine” with adjustable parameters to provide the best possible response. It works through thousands of interconnected units (the “neurons”) that each perform a small, simple calculation, and give excellent results when combined in the process of learning. Deepmind is a company that specialises in deep learning, with AlphaGo being the pinnacle of it’s developments. The program has fed 30 million moves of professional players, and plays against itself to learn from it’s own experience. This is the unique process that needs to be credited for AlphaGo’s victory.
Playing Go in Shanghai (Photo Brian Jeffery Beggerly)
Poker and Starcraft: The Two next challenges for artificial intelligence?
Go was the last board game where humans still had an edge over machines. With this historical victory – this time is now over. The question now is, what are the next challenges for artificial intelligence? Rumours suggest that the Google team may show an interest in the online strategy game Starcraft. Poker is also a game where humans are showing a lot of resistance. Limit Hold’em in it’s heads up format is considered effectively solved. But, humans are still beating computers when it comes to No Limit Hold’em. From April 24 to May 8, four of the best high stakes heads up players (Doug Polk, Bjorn Li, Dong Kim, and Jason Les) in the world competed against“Claudico,” a poker bot powered by artificial intelligence (AI.) Claudico was developed by a computer science team at Carnegie Melon University (CMU). The goal was to see whether or not man is still better than machine at No Limit Texas Hold’em. The results are in, and man still proved superior to machine, even though “Claudico” put up a good fight.
Machines will eventually be better than humans at poker – it’s just a question of when, and how.