‘Deep Learning’ Professor to Head Facebook AI Lab
December 16th, 2013Via: New Yorker:
Earlier today, Facebook announced that one of the most prominent artificial-intelligence researchers in the world, Yann LeCun, will be joining the company to direct a massive new A.I. effort spread across offices in California, New York, and London. While it might sound like just another Facebook hire, the move, which LeCun describes as having “the long-term goal of bringing about major advances in artificial intelligence,” points to a new layer in Facebook’s ambitions, as well as a shift in the research and development of artificial intelligence.
…
Even before he approaches these larger challenges, LeCun can immediately get to work by mining Facebook’s enormous library of user-submitted photographs. Already the largest such photo library in the world, users add roughly three hundred and fifty million photos to it every day. Earlier this year, when Facebook announced Graph Search, it was remarkable for changing the search paradigm from page-based, like Google results from 2003, to something that is knowledge-based and capable of aggregating information from across different sources. For example, I can ask Facebook to show me friends from college who are married and live in Boston, and it generates the data instantly, rather than showing me a pre-existing Web page. However, as I noted when Graph Search was announced, it’s not very good at extracting meaning from photos. I can’t ask Facebook to show me pictures from my college friends in New York City that have babies and cats in them. LeCun’s work could change that, making it possible, for example, to even ask Facebook “Which of my friends had babies recently?” and get a reasonable answer, complete with photos, even if the new parents had never explicitly said that a baby was born.
