Google has announced a temporary halt to the image generation of people by its AI chatbot, Gemini, following an apology for the “inaccuracies” in the historical images it produced. Google, in a post on the social media platform X, stated that they are already addressing the recent issues with Gemini’s image generation feature
“We are working to improve Gemini’s ability to generate images of people.We expect this feature to return soon and will notify you in release updates when it does,” the company said in a statement. The message also comes back in this same format on each prompt after a few seconds of loading.
This week, Gemini users shared screenshots on social media of scenes, historically dominated by white characters, now depicted with a diverse cast. This led to criticism and questions about whether Google is overcompensating for potential racial bias in its AI model.
Google apologises
Past research has indicated that AI image-generators can intensify racial and gender biases present in their training data. Without filters, these systems are more prone to generate images of lighter-skinned males when asked to create a person in various scenarios.
Earlier this week, Google acknowledged that Gemini has been producing inaccurate depictions in some historical image generations. They are actively working to rectify these depictions.
Gemini has the capability to generate a broad spectrum of people. Google considers this a positive aspect as the system is used globally, but admits that it is not hitting the mark.
When asked by AP to generate images of people, Gemini responded that it is enhancing this feature. The chatbot assured that this feature will be reinstated soon and users will be notified in the release updates when it happens.