Skip to Content
Artificial intelligence

Microsoft’s neo-Nazi sexbot was a great lesson for makers of AI assistants

Yandex’s head of machine intelligence says Microsoft’s Tay showed how important it is to fix AI problems fast.
March 27, 2018
Jeremy Portje

Remember Tay, the chatbot Microsoft unleashed on Twitter and other social platforms two years ago that quickly turned into a racist, sex-crazed neo-Nazi?

What started out as an entertaining social experiment—get regular people to talk to a chatbot so it could learn while they, hopefully, had fun—became a nightmare for Tay’s creators. Users soon figured out how to make Tay say awful things. Microsoft took the chatbot offline after less than a day.

Yet Misha Bilenko, head of machine intelligence and research at Russian tech giant Yandex, thinks it was a boon to the field of AI helpers.

Speaking at MIT Technology Review’s annual EmTech Digital conference in San Francisco on Tuesday, Bilenko said Tay’s bugs—like the bot’s vulnerability to being gamed into learning or repeating offensive phrases—taught great lessons about what can go wrong.

The way Tay rapidly morphed from a fun-loving bot (she was trained to have the personality of a facetious 19-year-old) into an AI monster, he said, showed how important it is to be able to fix problems quickly, which is not easy to do. And it also illustrated how much people tend to anthropomorphize AI, believing that it has deep-seated beliefs rather than seeing it as a statistical machine.

“Microsoft took the flak for it, but looking back, it’s a really useful case study,” he said.

Chatbots and intelligent assistants have changed considerably since 2016; they’re a lot more popular now, they’re available everywhere from smartphone apps to smart speakers, and they’re getting increasingly capable. But they’re still not great at one of the things Tay was trying to do, which is show off a personality and generate chitchat.

Bilenko doesn’t expect this to change soon—at least, not in the next five years. The conversations humans have are “very difficult,” he said.

 

Deep Dive

Artificial intelligence

Sam Altman says helpful agents are poised to become AI’s killer function

Open AI’s CEO says we won’t need new hardware or lots more training data to get there.

Is robotics about to have its own ChatGPT moment?

Researchers are using generative AI and other techniques to teach robots new skills—including tasks they could perform in homes.

What’s next for generative video

OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.

An AI startup made a hyperrealistic deepfake of me that’s so good it’s scary

Synthesia's new technology is impressive but raises big questions about a world where we increasingly can’t tell what’s real.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.