Facebook says that its artificial intelligence (AI) division embarked on an ambitious quest: to build an AI so advanced that it can beat a human player at Go.
The two and a half thousand-year-old strategy board game is one of the toughest nuts to crack for programmers and AI experts. The rules are dead simple - each player places white or black stones at intersecting lines, on a 19x19 grid, and the final purpose is being able to control the board.
The minimalistic game has significantly more permutations than chess, and this is one of the reasons why humans are still better at it than machines. The tremendous number of variants is a challenge for any software engineer who develops a system to scrutinize them all before acting.
"After the first two moves in a chess game, for example, there are 400 possible next moves. In Go, there are close to 130,000" Facebook Chief Technology Officer Mike Schroepfer explains.
One reason why humans are so good at Go is that they use visual patterns to recognize the best course of action. Schroepfer points out that the AI Go-playing-system deviced by his company integrated this into the code.
"We've taken some of the basics of game-playing AI and attached a visual system to it, so that we're using the patterns on the board-a visual recognition system," he points out.
The visual recognition technology already works on Facebook's social media site, where the software can easily identify human faces.
The AI program is part of a larger research in the field of deep learning. A sector of artificial intelligence, deep learning aims to mimic human brain processes in order to tap into probabilistic and statistical capabilities, where contemporary computers are unprepared to go.
Schroepfer detailed that deep learning is developed with a few global perspectives in mind. One purpose is to make Facebook user-friendly to the visually impaired so that voice-recognition and live feed-back to the user become commonplace.
Another area where deep learning proves useful is Facebook's new AI assistant, dubbed M. The AI assistant is complex enough to take matters on its own, digital hands. It has the ability to make business trip arrangements, order personalized gifts for your loved ones, book reservations and more. It does so by observing and learning from the feedback that human operators offer it.
Even if world-class Go players are not yet threatened by Facebook's technology, we might see Kasparov's demise to Deep Blue happening again, this time on a 19x19 grid.
