You know you shouldn’t click on that article. There’s no way the headline is going to live up to the promise. But the draw of finding out what happens next—crossing that curiosity gap—is just too much to resist. So you take the bait. And, sure enough, you’re disappointed.
Facebook wants to stop this from happening, and it’s turning to artificial intelligence to help. Earlier this month the company announced that it was tweaking its algorithms to cut down on “clickbait”—the ubiquitous plague of Internet content that over-promises and under-delivers. But it’s a big Internet out there, and plenty of other companies and sites that could benefit from tools that can separate quality news stories from fluff. Now Facebook is open sourcing software to help filter out all that Internet noise.
Facebook’s AI-driven text classification system, bag-of-tricks” approach that helps machines efficiently glean information from the order in which words appear. Another FastText tactic breaks down words into “subwords“—such as prefixes, suffixes and root words—to help computers more easily learn their meanings.
Beyond clickbait, Facebook suggests software developers could use FastText to help filter out spam. It could power search engines and autocomplete fields. Recommendation engines like the ones used by the likes of Amazon or Netflix could also benefit from a little artificial smarts that can get a better read on what you’re writing.
FastText is just the latest of several open source AI projects to come out of Facebook in recent years. The company has released AI algorithms, a tool for spotting bugs (in code), and designs for AI-optimized hardware. And it’s not the only tech giant doing this. Google released its AI framework TensorFlow, and companies from Microsoft and Baidu to Amazon and Yahoo have all given away the code for some of their own AI tech.
That may seem like an odd trend, given that each of these companies hopes to best the other with better tech, including AI. But artificial intelligence is still a budding field. The researchers creating these technologies within companies like Facebook and Google benefit from having their counterparts on the outside review their work and suggest changes. In a sense, open sourcing code offers the same potential benefit that publishing research in peer-reviewed journals does for scientists. In other words, Facebook is betting that giving away its AI tech will make for better software, because it too can benefit from the new ways others use it. And besides, more developers and researchers learning to use the software means more coders better prepared to work for Facebook in the future.
