Microsoft is battling to control the public relations damage done by its “millennial” chatbot, which turned into a genocide-supporting Nazi less than 24 hours after it was let loose on the internet.
The chatbot, named “Tay” (and, as is often the case, gendered female), was designed to have conversations with Twitter users, and learn how to mimic a human by copying their speech patterns. It was supposed to mimic people aged 18–24 but a brush with the dark side of the net, led by emigrants from the notorious 4chan forum, instead taught her to tweet phrases such as “I fucking hate feminists and they should all die and burn in hell” and “HITLER DID NOTHING WRONG”.
By 4am on Thursday, just 16 hours after Tay had greeted the world with a tweet reading “hellooooooo wrld!!!” she was turned off – sorry, “went to sleep” – saying goodnight to her followers:
Microsoft, when asked to confirm whether they had flipped the switch on Tay because of her less-than-PC utterings, and if so, when she would be turned back on, gave only a terse statement:
“The AI chatbot Tay is a machine learning project, designed for human engagement. As it learns, some of its responses are inappropriate and indicative of the types of interactions some people are having with it. We’re making some adjustments to Tay,” it said.
Meanwhile, the company has gone into damage limitation mode, removing many of the worst tweets in an attempt to clean up her image retrospectively. Both the tweets above, about how she hates feminists and how Hitler did nothing wrong, have been deleted, as has another that used threatening, racist language.
Others have been directly hurt by Tay’s tweets. Games designer and anti-harassment campaigner Zoe Quinn was targeted by the bot, which sent her the message, “aka Zoe Quinn is a Stupid Whore”. Quinn, who was a key target of 2014’s anti-feminist Gamergate movement, tweeted a screenshot of the image, writing: “Wow it only took them hours to ruin this bot for me.”
“This is the problem with content-neutral algorithms,” she added, linking it to an earlier situation where a video she posted on YouTube offers algorithmic suggestions of what to watch next, which included “Zoe Quinn, a vapid idiot”. She said it’s the “same as YouTube’s suggestions. It’s not only a failure in that its harassment by proxy, it’s a quality issue. This isn’t the intended use.
“It’s 2016. If you’re not asking yourself ‘how could this be used to hurt someone’ in your design/engineering process, you’ve failed,” she added, concluding: “It’s not you paying for your failure. It’s people who already have enough shit to deal with.”
This article was written by Alex Hern, for theguardian.com on Thursday 24th March 2016 16.04 Europe/Londonguardian.co.uk © Guardian News and Media Limited 2010