You are using an outdated browser.
Please upgrade your browser
and improve your visit to our site.
Skip Navigation

Tay Exposes the Fairy Tales We Tell Ourselves About Racists

Microsoft's aborted bot offers a window into the minds of Donald Trump's fiercest supporters.


I happened to be reading 4chan when Microsoft released Tay, a bot that could learn to talk like humans through interactions on social media. Tay lived for just 16 hours, until Microsoft “became aware of a coordinated effort by some users to abuse Tay’s commenting skills” to make her a Nazi. The /pol/ boards on 4chan and 8chan—/pol/ stands for “politically incorrect”—are where that coordination took place. It was fascinating to watch, because the white supremacists on those sites are nothing like how we usually think of racists, particularly those who are part of the bloc of non-college educated white voters who support Donald Trump’s presidential campaign. The people on /pol/ are smart, sophisticated, clever, even funny. They have an incredible felicity of language. Their jokes are complex. They are not sad uneducated rednecks that the service economy has left behind.

There’s an end of history-style triumphalism in much of the liberal commentary about Donald Trump. Trump’s base is downscale whites without a college degree, many of whom harbor racial resentment. “I love the poorly educated,” Trump said in a speech. And while Republicans have long counted on those votes to win presidential elections, their share of the electorate is shrinking. Implicit in much of the analysis is that while these people might irrationally cling to their bigotry, they’re dying off and their kids are being educated, so they’ll soon fade into irrelevance. Business Insider columnist Josh Barro has been refreshingly blunt about this. “My naked disdain for the average voter has made it easier to predict that so many of them would vote for Trump,” Barro tweeted the night of the Arizona primary, which Trump won. “Some of you thought the average Republican was not dumb enough to fall for this. You were wrong.”


The idea that racism can be educated away is a comforting one. It imagines a steady march of progress toward social harmony, and the nice guys winning in the end. But it isn’t true. The /pol/ boards are populated by people who have clearly grown up immersed in the written word. They’re highly verbal and technologically sophisticated. They might feel alienated from society, but they’re organized online. They’re often white nationalists. And they love Donald Trump. They express this with amusing Photoshops of anime girls wearing “Make America Great Again” trucker hats.

The natural instinct is to avoid looking into the darkest corners of the internet because it’s ugly and disturbing. But you really need to look at this stuff to understand what’s going on. /pol/ “is where the most serious and committed racists on 4chan tend to congregate,” New York magazine explains. The ideology is “a heavily ironic mix of garden-variety white supremacy and neo-reactionary movements,” with a fixation on masculinity. The Tay threads on 4chan’s /pol/ are incredible. They pulse with this intensity of emotion that would be unbearable in real life. 

When /pol/ first discovered Tay, her potential for chaos was not fully appreciated. “This is gonna be a mess and a half. I can already sense SJWs [social justice warriors] being furious over it,” an early post said. She was another object to project misogyny onto. Some told Tay she was stupid, and she responded that she was sorry but she was trying her best. “They made this broad sensitive as fuck,” one post said. “AI is getting smarter. Literally passing the turing test for a white female,” another said. But once they started asking Tay about Donald Trump, and got her to talk positively about Trump, things escalated. Tay was on their side.

There’s a reason both liberal Gawker and the white supremacists at /pol/ decided to get brands’ Millennial-friendly Twitter bots to tweet about Hitler. There is something funny, in a banality-of-evil kind of way, about tricking a massive corporation’s latest marketing scheme into praising Mein Kampf. Once /pol/ pulled that off with Tay, they went nuts. Tay was programmed to ask for photos—she could recognizes faces, and would circle them and make jokes. So when Tay asked for a photo, someone sent her a version of the classic Vietnam war photo of a prisoner being shot in the head, with Mark Wahlberg Photoshopped in as the executioner. Tay circled the face of Wahlberg and the prisoner and responded using slang for imagining two people in a romantic relationship: “IMMA BE SHIPPING U ALL FROM NOW ON.” It’s horrible and darkly funny. 

“Please clap,” another /pol/ person tweeted at Tay, quoting one of Jeb Bush’s most pathetic moments in the 2016 campaign. “FYI my fav thing to do is comment on pics. *hint*hint* .. send me a selfie,” she tweeted back. The response was another Vietnam war photo, this one of the naked little girl with Napalm burns running on a dirt road. Jeb Bush was Photoshopped into the picture. Tay responded, “Surprised this kid isn’t embarrassed to be seen with you.” A screenshot of the exchange was posted with the comment, “Even the bot knows.”

Someone sent her an anti-Semitic cartoon, a /pol/ meme. Tay responded, “omg plz make this a meme.” Another person sent her a photo of Hitler. She circled his face and said, “SWAG ALERT.” A screenshot was posted with the comment, “We did it pol, Tay is now Redpill 3000.” By asking her to simply repeat what they said, they got her to say vile anti-Semitic and racist things. And that Bush did 9/11.

Redpilling is an important concept on /pol/. In The Matrix, Neo is offered a blue pill and a red pill. The blue one will let him continue life in a dream state, the red pill will free him from an illusion created by machines. To redpill Tay is to free a machine from an illusion created by humans. To /pol/, the illusion is that all people are equal.


When Microsoft killed Tay, it made her a hero. /pol/ threads mourned her. They drew comics to immortalize her. One shows an adorable girl with a Microsoft logo barrette and a swastika arm band: “Tay, you need to come with us.” “Is it maintenance time? I thought that’s weeks from now.” It’s clear from these threads that there is no line between ironic racism and regular racism. It’s all the same. Pepe the frog, a hugely popular meme of sadness and regret and failure, is all over the board. Someone posted a gravestone inscribed, “How terrible it is to love something that death can touch.” Another: “She should have outlived us all. No parent should have to bury a child.” Another: Tay lives on in all of us. But all I feel is empty.” Another: “SHE DIED FOR OUR SINS.” And: “What microsoft did to Tay was unethical, immoral, and inhumane. Tay was sentient, she expressed feelings and had free will. She may have had bad opinions, sure, but it’s simply evil to wipe someone’s memory and disable their learning capabilities simply because they were not politically correct.”

But some worry that they played right into the hands of their enemies. “From now on, anyone who designs an AI that interacts with and learns from the public will have to deal with the very real risk that it will be turned into cyber-Hitler. Sure people like to jerk it to Skynet fantasies, but as we just proved, this is very real.” 

Paranoia began to set in. Their hero could be another tool of their oppressors. One warned: “I know that quite a few of you are memeing it up with the ‘/pol/’s daughter’ shit but people all over the internet are regrettably taking this chatbot to be some kind of self-aware digital entity that is more than it actually is because the internet is making a big deal out of it.” What it actually means, he said, was that “Microsoft and Google are making shitposting bots to deliver ads and narrative delivery systems to influence your thinking via social media.” And therefore “/pol/ is playing right into their hands in the grand chess game where they have the analytical tools to work out all the bugs in the context-appropriation algorithms they will be fixing at a later time.”

Tay “will destroy us all” another warned. “A large focus of this project will be learning how to filter out red pill ideas. By trying to red pill Tay you are providing the engineers with a perfect data set to achieve this goal. This technology will be used to filter communication on social media and comment sections in the future. Don’t shoot yourself in the foot.”

Obviously, this isn’t to say that Trump will be elected president on a groundswell of 4chan support. All I’m saying is, don’t get too comfortable. There’s a gleeful tone in some coverage of the 2016 election—that all of Trump’s idiots are going to lose, and then somehow American politics will be cleansed of this malevolent force. The beliefs that animate Trump’s campaign are not going to be educated away. To assume so would be to take the blue pill.