It was late March, and just four days before Microsoft CEO Satya Natella was due to announce the company’s new focus on “conversation as a platform,” Lili Cheng woke up to discover that one of her chatbots had gone rogue.

The chatbot in question was Tay: an AI-driven Twitterbot that used natural language processing to emulate the speech patterns of a 19-year-old American girl. Presented as a precocious “AI with zero chill,” Tay could reply to Twitter users who messaged her, as well as caption photos tweeted at her. In fact, she described a selfie of the 51-year old Cheng as the “cougar in the room.” But just 16 hours after joining Twitter under the handle TayandYou, Tay had become a super-racist sexbot.

It was a rough day for Cheng, who is director of Microsoft’s experimental Future Social Experience (FUSE) lab where Tay was developed, but it’s part of her job. For the last 20 years, Cheng has helped Microsoft explore the limits of user experience and interface design. During that time, she’s had many projects, yet she tells me they’ve all shared a few common characteristics. “They’ve all tended to be social, interactive, high risk, and ambiguous,” Cheng says. Just like Tay.

Read the full article on Fastco.Design