Facebook to develop A.I. system that plays game no other computer program can master

Brain-Magnetic-Field-Computer-Thoughts

Computers trump people at various games, from chess to trivia. Nevertheless, there remains one pastime where humans still reign supreme: the game of Go. This supremacy may be in its twilight, however, as Facebook officials study Go in hopes of building an artificial brain and, in turn, a virtual assistant.

Computers are able to win various games by analyzing the possible outcomes of every possible move. On the other hand, Go, an Eastern variation of chess, is much more difficult to predict. The number of possible moves is far greater, and knowing the benefits of a specific move is next to impossible. Top players of Go claim that the game almost has to be played at a subconscious level.

Consequently, Facebook officials are now studying Go using an artificial intelligent system known as deep learning.

Social media companies compete for A.I. systems

Social media companies have recognized, in recent years, that deep learning is on par with recognizing photos, identifying words and translating one language into another. To recognize a dog, for example, a deep learning system reviews thousands of dog pictures, pumping them into a network of machines that estimate the neural networks of the human brain.

Thanks to these networks, Facebook has developed a software than can analyze a photo and answer questions about what it sees, such as judging whether a picture of a teetering tower of blocks will fall over.

Of course, humans are just as equipped, perhaps more so, at answering questions about photos on social media than an artificial brain. Nevertheless, the program is more cost-effective than assembling a team of people fit for the same task.

“The beauty of AI systems is that we can get enough scale to deploy this to everyone on the planet,” Mike Schroepfer, Facebook’s chief technology officer, said in a briefing for reporters.

Facebook isn’t the only social media company working on these problems. Google, Microsoft and Apple, along with several smaller tech firms, are also studying A.I. to provide digital services to consumers. These companies are competing to hire talented scientists from prestigious universities in order to reap the benefits of deep learning.

Meet the online personal assistant, M

The social media networking company recently started testing an online personal assistant known as M, which uses both humans and software to answer questions and carry out various tasks. Facebook hopes M can learn from these responses, so that it can one day operate without human assistance.

Facebook has also sponsored efforts to build drones and laser networks in order to reach billions of people who do not have access to the Internet. Schroepfer claims Facebook can use A.I. to predict the information those users intend to seek.

“We need systems that can help us understand the world and help us filter it better,” he stated.

Schroepfer plans to unveil some of Facebook’s work in a presentation at an international Web Summit conference in Dublin.

Software capable of playing a game of Go will be among the works presented at the conference. As noted earlier, there are thousands of options for arranging stones on the board in a game of Go. The best players make moves based upon visual patterns, rather than strategic calculations. Schroepfer claims Facebook is using a visual recognition algorithm to examine every possible move in a similar way people determine how to make a move.

It may be years before this technology is fully developed. Nevertheless, Schroepfer claims it’s definitely on the horizon.

Sources:

Wired.com

Hosted.Ap.Org