Facebook’s facial recognition capabilities are about to get a little creepier

If you’re the type of person who ducks out of photos to avoid being tagged, Facebook is on to you. The company’s artificial intelligence lab is developing an experimental algorithm that detects users’ identities based on body shapes, hair and clothes instead of faces.

The idea mimics humans ability to recognize people in the same way we use clues without seeing faces. In an interview with the New Scientist, Yann LeCun, Facebook’s head of A.I. gave the example of Mark Zuckerberg who is identified “very easily, because he always wears a gray T-shirt.”

Facebook uploaded 40,000 photos from Flickr, some of them showing people’s faces and others partially obscured, and ran them through a “sophisticated neural network.” The algorithm was able to detect people’s identities 83 percent of the time.

The intent isn’t to be creepy, claims LeCun. The tool could be used to alert people that pictures of them have been uploaded to the Internet without their knowledge. More practically, the tool will be used to ease the burden of tagging multiple photos at once.

Still, Facebook might have trouble convincing users that this yet-to-be-released tool isn’t invading people’s privacy. It’s already embroiled in controversy in Europe over its facial recognition services: European regulators have blocked the release of Facebook’s new photo-sharing app Moments over its facial recognition abilities, which the company has nailed down to human-like accuracy.

So, if you’re trying to avoid Facebook, perhaps the best solution for now is to never leave the house.

Image courtesy of YouTube.

https://digiday.com/?p=123095

More in Media

How creators are using generative AI in podcasts, videos and newsletters — and what advertisers think about it

Here’s a look at how some creators are leveraging generative AI to create video, audio and written content — and whether or not that’s a turn-off for advertisers.

Illustration of a performer balancing money weights on a tightrope, symbolizing how brand safety tools help marketers maintain performance and control.

Buzzfeed, News Corp and New York Times push back on tariff fears in earnings calls

Publishing execs pushed back on tariff and macroeconomic climate fears in Q1 2025 earnings calls, expressing confidence that their businesses would grow this year.

Digiday+ Research: Publishers’ subscription revenue is up this year, and they’ll focus on growing it even further

Subscriptions is one area where publishers are seeing more revenue, and, in turn, ramping up their plans to strengthen that part of their business in the coming months.