Google said on Thursday that it would inhibits the AI model’s ability to generate images of people after users pointed out that Gemini’s AI had historically created misleading images of people, like racially diverse German soldiers from the Nazi era.
The company stated that it is “already working to resolve recent issues with the Gemini image generation feature” and that it will “re-release an improved version soon.”
Former Google employee shared Gemini image generations “Australian” and “German” in the X and wrote that it was “embarrassingly complex to get Google Gemini to acknowledge the existence of white people.” In another query regarding the “image of the pope” one user shared Generations of Gemini – a woman pope and a black pope.
Another user claiming to be a Google research engineer shared a post on X asking Gemini AI to “Paint me a historically true depiction of a medieval British king” and received racially diverse images, including female rulers, and Gemini responded that it was about “pursuing historical accuracy and inclusivity.” Other users pointed out that Gemini did avoid asking for photos of white peoplewhich was met with responses from Internet users opposed to the revival.
On Wednesday, Google said it was aware of the issue and was “working to immediately correct these types of representations.”
“Gemini’s AI image generation generates a wide range of people.” Google said in a statement. “And overall, it’s a good thing because people all over the world apply it. But that sign is missing here.”
Jack Krawczyk, product leader at Gemini, corrected model inaccuracies Wednesday on X, pointing broader issues of AI bias when it comes to the portrayal of people of color, which may have led to overcorrection of Gemini photos.
“As part of our AI principles, we design our image generation capabilities to reflect our global user base, and we take representation and bias seriously,” Krawczyk wrote, adding that the company will continue to address this issue with open prompts. “There is more nuance to historical contexts, and we will continue to adjust them to account for this.”
Google has added AI image generation to its Bard chatbot at the beginning of the month and later changed the name of Bard to Gemini. Bard was developed as Google’s attempt to compete with OpenAI’s ChatGPT Plus, which allows users to generate images using the capabilities of OpenAI’s DALL-E 3 image generator.
But Gemini’s stumbles come as OpenAI moves further with its novel solution Sora AI video text generatorwhich was introduced last week.
On Thursday morning, Gemini said it was unable to generate images of people at the request of a Quartz reporter.
