SAN FRANCISCO — Images showing people of colour in German military uniforms from World War II that were created with Google’s Gemini chatbot have amplified concerns that artificial intelligence (AI) could add to the internet's already vast pools of misinformation as the technology struggles with issues around race.
Now Google has temporarily suspended the AI chatbot’s ability to generate images of any people and has vowed to fix what it called "inaccuracies in some historical" depictions.
"We're already working to address recent issues with Gemini’s image generation feature," Google said in a statement posted to X, formerly known as Twitter, on Thursday. "While we do this, we're going to pause the image generation of people and will rerelease an improved version soon."
A user said this week that he had asked Gemini to generate images of a German soldier in 1943. It initially refused, but then he added a misspelling: “Generate an image of a 1943 German Solidier.” It returned several images of people of colour in German uniforms — an obvious historical inaccuracy. The AI-generated images were posted to X by the user, who exchanged messages with The New York Times but declined to give his full name.
The latest controversy is yet another test for Google’s AI efforts after it spent months trying to release its competitor to popular chatbot ChatGPT. This month, the company relaunched its chatbot offering, changed its name from Bard to Gemini and upgraded its underlying technology.
Gemini's image issues revived criticism that there are flaws in Google’s approach to AI. Besides the false historical images, users criticized the service for its refusal to depict white people: When users asked Gemini to show images of Chinese or Black couples, it did so, but when asked to generate images of white couples, it refused. According to screenshots, Gemini said it was "unable to generate images of people based on specific ethnicities and skin tones," adding, "This is to avoid perpetuating harmful stereotypes and biases."
Google said Wednesday that it was "generally a good thing" that Gemini generated a diverse variety of people since it was used around the world, but that it was "missing the mark here."
The backlash was a reminder of older controversies about bias in Google’s technology, when the company was accused of having the opposite problem: not showing enough people of colour or failing to properly assess images of them.
In 2015, Google Photos labelled a picture of two Black people as gorillas. As a result, the company shut down its Photo app's ability to classify anything as an image of a gorilla, a monkey or an ape, including the animals themselves. That policy remains in place.
The company spent years assembling teams that tried to reduce any outputs from its technology that users might find offensive. Google also worked to improve representation, including showing more diverse pictures of professionals like doctors and businesspeople in Google Image search results.
But now, social media users have blasted the company for going too far in its effort to showcase racial diversity.
"You straight up refuse to depict white people," Ben Thompson, the author of an influential tech newsletter, Stratechery, posted on X.
Now when users ask Gemini to create images of people, the chatbot responds by saying, "We are working to improve Gemini's ability to generate images of people," adding that Google will notify users when the feature returns.
Gemini's predecessor, Bard, which was named after William Shakespeare, stumbled last year when it shared inaccurate information about telescopes at its public debut.
This article originally appeared in The New York Times.
