The millennia-old board game of Go had long been thought beyond the reach of Artificial Intelligence because of its notorious complexity. As recently as 2014
, it was thought impossible that an AI could match a top human Go player, let alone beat one.
At Codebots, we've been eagerly anticipating the result of the "Future of Go Summit", hosted in Wuzhen last week (May 23-27), where the Google DeepMind developed AI beat the world's top Go player 3-0.
Although not an unexpected result, there had been high hopes that Ke Jie, who has held onto the coveted world #1 ranking since 2014, could do what no other human could: beat AlphaGo.
But inroads into machine learning, deep learning and elsewhere are expanding AI's once narrow range of capabilities. Machines can now write music
, draw landscapes
, play Go, and in cases like Codebots, even code!
Go is played on a 19 by 19 boardâ€”in comparison, Chess is played on an 8 by 8 board. Furthermore, whereas in Chess there are only 20 possible opening moves, Go has 361. To make a long story short: Go is much more complex than Chess. This is why it took so much longer for an AI to win at Go than at Chess.
Between the 9th and 14th of March 2016, it versed 18-time world champion Lee Sedol in a best of 5 match, winning 4 games.
So complete was AlphaGo's dominance that when Lee won game 4, a dead rubber as he had already lost the previous 3 games, he was cheered by the press and commented that "this win is so valuable that I wouldn't exchange it for anything in the world".
AlphaGo would go on to close out the series 4-1.
To some, AlphaGo is intimidating, but to others it's exciting. Go pro Gu Li thinks that "together, humans and AI will soon uncover the deeper mysteries of Go."
AlphaGo has surpassed the top human players, and this is an opportunity. AlphaGo's playstyle is distinct. Sometimes the timing and placement of its stones perplex its human opponents and spectatorsâ€”only to, upon further analysis, prove to be "divine moves
". A term used to describe moves that changed the entire course of the game.
So, while there's almost no chance that humans will ever take Go back, human players can learn more about Go and uncover its deeper mysteries by watching the AI play. And this is already happening, with Ke Jie adapting his own play style during the three-game match.
Deepmind has announced
that it will publish an academic article detailing the extensive set of improvements they made to AlphaGo's algorithms; develop a teaching tool; and release 50 AlphaGo vs. AlphaGo games. These steps will undoubtedly help human Go players come up with many new and interesting ideas and strategies, and, in the words of Gu Li: allow humans and bots to "uncover the deeper mysteries of Go" together.
To some, AlphaGo and Codebots are the vanguard of a new generation of bots that will take over the world. This simply isn't true.
Just as AlphaGo has empowered human Go players with new and innovative approaches to playing the ancient board game, Codebots will empower developers with a collaborative platform where they and code-writing robots work together and build awesome applications.
With bots helping with tedious tasks like creating and linking databases with code, human creativity has more space to thrive. Using Codebots, developers will be able to create awesome new software-lead solutions (and in less time!).
The bots are like sidekicks; creative humans are still the heroes!
So to anyone who thinks that with AlphaGo, Codebots and other AI on the horizon that humans are now obsolete: you can stop worrying. While it can be intimidating to watch a computer do something we had come to think of as being distinctly human, Bots are not the new alphas. Human-bot teams that harness human creativity and bot logic are.
Still a little intimidated? Keep your eyes peeled as my next article will geek out on the things humans beat bots at.