Skip to main content

Facebook AI Research teams share new milestones and developments

While the Internet has been abuzz with various artificial intelligence (AI) research developments, Mike Schroepfer, Facebook’s Chief Technology Officer, recently provided an in-depth look at the company’s latest efforts on this front: The social network, which claims 1.5 billion users, has reached a number of noteworthy milestones in its long-term research.

Among these milestones that the Facebook AI Research (FAIR) team is prepared to present are such things as object detection, which can segment images and distinguish between objects in a photo — up to 30 percent faster than before. This feature, assisted by the work on natural language understanding through FAIR’s Memory Networks system, will likely contribute to the ongoing campaign to produce technology that allows blind people to “see” images.

At Web Summit in Dublin, Ireland, Schroepfer commented on the importance of this new technology. “Imagine that you are one of the hundreds of millions of people with some sort of vision disability and you have trouble participating in the visual part of social networks. And one of your friends, who just had a baby, posts a photo and captions it. There’s technology already out there to read all the text on the screen to you … but you wanted to learn more about what this photo is. We built a system that allows you to ask questions about a photo that it’s never seen before.”

Among its newest developments is the nearly successful attempt at teaching an AI bot to play the board game, Go. While bots in the past have been able to outplay and win games such as chess, Jeopardy!, and Scrabble, according to Wired, only humans have able to consistently win at the game Go. The Facebook team’s newest challenge is to use intelligence known as deep learning to teach systems planning by combining the “traditional search-based approach” with a “pattern-matching system.” And after just a few months, Facebook says its bot has become just as good at the game as a human player.

Editors' Recommendations