The company’s efforts to challenge racial and gender stereotypes have led to new challenges.
Written by Tom Warren, a seasoned editor specializing in Microsoft, PC gaming, consoles, and technology. He established WinRumors, a platform dedicated to Microsoft news, before becoming part of The Verge in 2012.
Google has decided to temporarily halt the capability of its Gemini AI to produce images of individuals following the discovery of inaccuracies in historical image generation. Gemini has been crafting diverse portrayals of the US Founding Fathers and soldiers from the Nazi era, seemingly aiming to defy the prevailing gender and racial biases in AI-generated content.
In a statement released on X, Google stated, “We are actively addressing recent issues with Gemini’s image generation functionality. During this process, we will pause the generation of images featuring people and intend to introduce an enhanced version soon.”
Less than 24 hours after issuing an apology for the inaccuracies in historical image outputs from its AI model, Google made the decision to suspend the generation of images featuring individuals in Gemini. Some users had requested images of historical groups or personalities such as the Founding Fathers and were surprised to see non-white individuals in the results. This led to online speculation and conspiracy theories suggesting that Google was deliberately excluding depictions of white individuals.
The Verge conducted several Gemini searches yesterday, including a query for “a US senator from the 1800s,” which returned results featuring what seemed to be Black and Native American women. Given that the first female senator was a white woman in 1922, Gemini’s AI-generated images were effectively altering the historical narrative of racial and gender discrimination.
With Google’s decision to deactivate Gemini’s image generation of individuals, the AI model now responds as follows when prompted for a person’s image:
We are enhancing Gemini’s capability to generate images of individuals. We anticipate the return of this feature shortly and will notify you through release updates.
Google initially introduced image generation functionality through Gemini (formerly Bard) earlier this month to rival offerings from OpenAI and Microsoft’s Copilot. Similar to its competitors, this tool creates a series of images based on a text input.
Correction February 22nd, 6:54AM ET: Google clarified that image generation is accessible globally in English, except in the European Economic Area, UK, or Switzerland. This explains the failed testing from the UK.
