Irish Times Fooled by AI-Written Op-Ed About Fake Tans, Racism


The next article you read might be written by a robot, and people who published it might not even know. The Irish Times issued an apology Sunday after running a hoax editorial written by an AI chatbot, which called the incident a “deliberate and coordinated deception.” The article, titled “Irish women’s obsession with fake tan is problematic,” was attributed to a made up immigrant woman, complete with an AI-generated photo of the imaginary author.

“We don’t take this lightly. It was a breach of the trust between The Irish Times and its readers, and we are genuinely sorry,” said Ruadhán Mac Cormaic, editor of the Irish Times, in a statement presumably written by a human person. “The incident has highlighted a gap in our pre-publication procedures. We need to make them more robust—and we will. It has also underlined one of the challenges raised by generative AI for news organisations. We, like others, will learn and adapt.”

According to Mac Cormaic, the Times had a lengthy back and forth with the person who submitted the hoax article, who took notes and suggestions from editors, provided personal anecdotes, and updated the piece with links to research.

The article accused white Irish women who get fake tans of cultural appropriation, written from the perspective of a fabricated 29-year-old Ecuadorian woman living in north Dublin named Adriana Acosta-Cortez. The article, published on May 11th, ran with a fake photo picturing a woman with blue hair.

As if the prank wasn’t humiliating enough, a Twitter account adopting Acosta-Crotez’s personality kept the ruse going online. At first, the Twitter account responded to comments and criticisms from readers. It’s not clear whether the account belongs to the person behind the scam, but the Twitter user clearly had prior knowledge about the incident. Shortly after the Times took the article down, the Acosta_Cortez account switched gears, Lambasting the newspaper for printing an article “with such divisive tripe in order to generate clicks and traffic for their website.” The comments center around some strained point about wokeism rather than highlighting the dangers of AI.

The Times took the article down in less than 24 hours, but the Acosta-Cortez Twitter account posted an archived link to the story.

The AI hoax quickly made international news and sparked a frenzy in the Irish media. The Guardian spoke to the person running the Twitter account:

The person who controls Acosta-Cortez’s Twitter account told the Guardian on Sunday, via direct message, that the Irish Times’s apology sidestepped its decision to publish “an incendiary article with an extreme leftwing viewpoint” in pursuit of clicks.

The person said they were Irish, a college student and identified as non-binary. They said they created the Acosta-Cortez persona by repurposing the Twitter account, which dates from February 2021, by using some Spanish and following Ecuadorian outlets.

They said they used GPT-4 to create approximately 80% of the article and the image generator Dalle-E 2 to create a profile picture of a quintessential “woke” journalist using the prompts “female, overweight, blue hair, business casual clothing, smug expression”.

The hoax’s goal was “to give my friends a laugh” and “to stir the shit” in the debate about identity politics.

“Some people have called me an alt-right troll but I don’t think that I am. I think that identity politics is an extremely unhelpful lens through which to interpret the world.”

Irish journalist Cianan Brennan criticized the imposter for their self-induced outrage, calling out the so-far anonymous person for grandstanding and confused intentions. “Whole thing is just juvenile. Show your face if you’re so upset the piece was used,” Brennan tweeted.

The opinions in the article are controversial but not entirely without precedent, largely what you’d expect to find on the opinion pages of a legacy publication. The fake article’s “extreme leftwing” views include some thought provoking arguments echoed by sincere writers of color in recent years—which is also unsurprising, given the fact that chatbots are incapable of generating entirely new ideas.

“We can’t talk about fake tan without acknowledging the historical context of skin colour and the value that has been placed on it. For centuries, Eurocentric beauty standards have favoured lighter skin tones, often at the expense of people with darker complexions,” the AI-written article said. “By artificially darkening skin, fake tanning culture inadvertently perpetuates the fetishisation of high melanin content, without acknowledging the struggles faced by those who naturally possess it.” It culminates by asking Irish women to “consider the implications of your choices and to question the societal norms that guide them.”

This is one of the first times an AI-penned hoax made international news since tools like ChatGPT hit the world stage, but some publications print robot writing on purpose. In April, a misinformation watchdog called NewsGuard identified almost 50 websites covertly printing AI-written content in seven languages, Chinese, Czech, English, French, Portuguese, Tagalog, and Thai. That’s to say nothing about the countless AI-written books you can buy on Amazon, assuming you don’t realize that you could just generate the yawn-inducing literature yourself using ChatGPT or other free tools.

CNET embarrassed itself last January when it was forced to issue major corrections to a series of ChatGPT-written articles, including problems with basic math. CNET quietly published AI articles for months before readers caught on, attributing the stories to CNET Money Staff and hiding a disclosure about the AI on the byline’s author page. Rest assured that any mistakes you find on were made by real, fleshy human beings.

Source link


Please enter your comment!
Please enter your name here