Hey Facebook users, the makers of Adblock Plus want your screenshots to train its AI

The makers of Adblock Plus want you to submit screenshots of the ads in your Facebook feed so a new artificial intelligence system called Sentinel can learn to recognize them.

A new project from Eyeo, the German company behind the popular ad blocker, will likely spend several months training Sentinel’s neural network system to recognize ads, so the AI system isn’t yet available to Adblock Plus users.

“It’s not for ad blocking now,” says Ben Williams, Eyeo’s director of communications. “It’s for ad blocking some time in the future.”

But anyone can see how well it’s doing in the meantime by sharing screenshots of their Facebook newsfeeds that include ads. To upload the images, interested users can take a screenshot, then message Sentinel’s chatbot, which is available at the Sentinel website and through Facebook’s Messenger. The bot will quickly respond with sections of the uploaded image highlighted in red or green. Perceived advertisements are marked in red and each section is labeled with the bot’s estimated probability that it’s right.

Users are able to ask Eyeo to delete screenshots immediately or later on, and Williams suggests participants black out any sensitive information in their screenshots before uploading them to the service.

Facebook didn’t immediately respond to an inquiry from Fast Company about the bot, and Williams says Eyeo hasn’t had any communication from the company other than Facebook approving the bot for use on Messenger.

The Future Of Ad Recognition

In the future, Eyeo might crowdsource ads from beyond Facebook in order to make the bot useful on other sites as well, Williams says. The company began with Facebook because it’s taken steps to make its ads hard for ad-blocking software to recognize.

“We’ve always been able to develop workarounds to these circumvention techniques, but this cat-and-mouse game will likely continue for the foreseeable future,” CEO and cofounder Till Faida said in a statement. “While only at its earliest stage right now, our Sentinel offers the framework of a long-term solution to the rise of anti-user ‘circumventionists’; and could have implications for viewability and ad detection within the ads industry as well.”

That is, ad networks and buyers of ads could also use ad detection technology to make sure their ads are actually being shown on sites where they’re paying to run them. And, says Williams, government regulators could use ad detection technology to find online ads and make sure they’re in line with legal standards.

Eyeo isn’t the only organization looking into AI for spotting ads. The ad blocking company Brave announced in late May that in collaboration with researchers from the University of Iowa and the University of California at Riverside, it had developed a machine learning tool that could spot ads and tracking code with 97.7% accuracy. That system, called AdGraph, looked at relationships between website code segments rather than screen images to spot likely ads.

Researchers at Princeton University and Stanford University argued in a recent paper that since ads are generally required to be identified on a website in a way that humans can understand, visual cues should be enough for ad-blocking software to recognize them.

“Our claim is that as long as the disclosure standards are unambiguous and adhered to, a perceptual ad blocker will have a 100% recall at identifying ads governed by that standard,” they wrote. “Indeed, our techniques could be seen as a way to highlight ads instead of blocking them, serving to enhance disclosure.”

For now, Williams says it’s hard to predict exactly when visual filtering will find its way into Adblock Plus or another Eyeo product.