Microsoft’s AI makes racist error and then publishes stories about it

Hey, at least Microsoft’s news-curating artificial intelligence doesn’t have an ego. That much was made clear today after the company’s news app highlighted Microsoft’s most recent racist failure.

The inciting incident for this entire debacle appears to be Microsoft’s late May decision to fire some human editors and journalists responsible for MSN.com and have its AI curate and aggregate stories for the site instead. Following that move, The Guardian reported earlier today that Microsoft’s AI confused two members of the pop band Little Mix, who both happen to be women of color, in a republished story originally reported by The Independent. Then, after being called out by band member Jade Thirlwall for the screwup, the AI then published stories about its own failing.

So, to recap: Microsoft’s AI made a racist error while aggregating another outlet’s reporting, got called out for doing so, and then elevated the coverage of its own outing. Notably, this is after Microsoft’s human employees were reportedly told to manually remove stories about the Little Mix incident from MSN.com.

Still with me?

“This shit happens to @leighannepinnock and I ALL THE TIME that it’s become a running joke,” Thirlwall reportedly wrote in an Instagram story, which is no longer visible on her account, about the incident. “It offends me that you couldn’t differentiate the two women of colour out of four members of a group …  DO BETTER!”

As of the time of this writing, a quick search on the Microsoft News app shows at least one such story remains. 

Image: screenshot / microsoft news app

Notably, Guardian editor Jim Waterson spotted several more examples before they were apparently pulled. 

“Microsoft’s artificial intelligence news app is now swamped with stories selected by the news robot about the news robot backfiring,” he wrote on Twitter.

We reached out to Microsoft in an attempt to determine just what, exactly, the hell is going on over there. According to a company spokesperson, the problem is not one of AI gone wrong. No, of course not. It’s not like machine learning has a long history of bias (oh, wait). Instead, the spokesperson insisted, the issue was simply that Microsoft’s AI selected the wrong photo for the initial article in question. 

“In testing a new feature to select an alternate image, rather than defaulting to the first photo, a different image on the page of the original article was paired with the headline of the piece,” wrote the spokesperson in an email. “This made it erroneously appear as though the headline was a caption for the picture. As soon as we became aware of this issue, we immediately took action to resolve it, replaced the incorrect image and turned off this new feature.” 

Unfortunately, the spokesperson did not respond to our question about human Microsoft employees deleting coverage of the initial AI error from Microsoft’s news platforms. 

Microsoft has a troubled recent history when it comes to artificial intelligence and race. In 2016, the company released a social media chatbot dubbed Tay. In under a day, the chatbot began publishing racist statements. The company subsequently pulled Tay offline, attempted to release an updated version, and then had to pull it offline again. 

As evidenced today by the ongoing debacle with its own news-curating AI, Microsoft still has some work to do — both in the artificial intelligence and not-being-racist departments.