For better or worse, we’re all acquainted with Facebook. You can use it to share your reaction to last night’s Breaking Bad, provide real-time updates of your attempt to make lemon bars for the first time, or chronicle your new baby’s adventures in feeding, sleeping, nursing, and burping. It’s so commonplace to mention these details that we rarely think about the consequences, aside from whom we might be annoying or who might be annoying us with various inanities. In addition to the slew of ads users find on Facebook, and social networking site’s questionable privacy polices, Facebook is poised to harvest even more information and insight into their users. And, undoubtedly, to leverage what they learn as much as they can.
Facebook’s AI team is hard at work developing something called “deep learning,” an artificial intelligence that processes data much like a human brain would, using simulations of networks and brain cells. What does that mean, exactly? While computer programs are already adept at identifying key terms, likes, and other information to get an understanding of where and how users spend their time and money, deep learning AI can determine users’ emotions, even if a post isn’t obvious about them. The AI can also identify objects in users’ photos (it might notice, for example, that you have an iPad and then target you with new apps, or it might notice that you have a dog and start feeding you ads about allergies and heartworm medication). The AI can also, based on the content of posts, the associated emotions, and other information gleaned from the site, make predictions about what people are likely to do.
The idea of a psychic Facebook is terrifying, no?
The specific details about Facebook’s use of AI to harvest this information are being kept under wraps, but Facebook’s chief technology offer says that the new programming will help the site improve the personalization of the news feed. The CTO admits that, as the volume of updates grows, the site needs to improve its ability to pick out the most relevant ones for other users. So if I post that I, for example, hate sports (this isn’t true, by the way — at least, not all sports), would Facebook then give me a sports-free newsfeed? Or a baby-free or politics-free feed? Hmm…
Google and Microsoft have already incorporated deep learning. Google’s program is particularly good at recognizing objects from YouTube videos (must be all those Nyan cat videos) and has helped Google improve its voice recognition. Microsoft uses deep learning for real-time translations of spoken English into spoken Mandarin Chinese.
Facebook will eventually make their AI goals more public, but for now they’re looking to further develop deep learning, incorporating it into both hardware and software. Facebook actually hired one of Google’s deep learning experts, so we’ll see how long it takes before this programs can predict which organizations might be more likely to poach another’s employees.