Google halts Gemini AI image generator after ‘inaccuracies’

Google on Thursday announced it was pausing its Gemini AI imaging feature after it said it offered “inaccuracies” in historical photos.

Social media users complained that the AI ​​tool generated images of historical figures — such as the founding fathers of the United States — as people of color, calling it inaccurate.

Google said in an X post on Wednesday that the AI ​​feature could “generate a wide range of people. And that’s generally a good thing because people around the world are using it.” But the software feature is said to “miss the mark here”, adding that the tech giant is “working to immediately improve this kind of imaging”.

Google released an updated statement Thursday, saying it will pause Gemini’s feature for generating images of people and will soon relaunch an “enhanced” version.

The imaging tool was released in early February by Gemini, formerly Bard. It faces challenges as Google tries to catch up with Microsoft-backed OpenAI.

Gemini refused to generate any images Thursday morning when it was tested by a CNBC reporter.

As Google struggles to create Gemini images, Open AI last week launched Sora, its new generative AI model capable of producing video from user text prompts.

Jack Krawczyk, senior product director for Gemini at Google, said Wednesday that the company’s imaging capabilities reflect the tech giant’s “global user base” and that it takes “representation and bias seriously.”

“We’ll continue to do this for open-ended prompts (images of a person walking a dog are universal!),” Krawczyk said in an X post.

“The historical context is more nuanced, and we’re going to adjust further to accommodate that.”

Google launched Gemini, its most advanced AI model, in late 2023. This month it also rebranded Bard, its ChatGPT competitor, to Gemini. Additionally, Google launched a subscription service for a more powerful version of the AI ​​model.