Google has decided to halt the ability of its AI tool, Gemini, to generate images of people after facing criticism for producing historically inaccurate images predominantly showing people of color instead of White people. The decision to pause this feature comes after the tool was accused of perpetuating harmful racial and ethnic stereotypes.

Gemini, like other AI tools, is trained on large amounts of online data, which experts have warned could lead to the replication of racial and gender biases present in that information. The tool’s inability to accurately generate images of White people highlights the ongoing struggle of AI tools with the concept of race.

When prompted to generate an image of a pope, Gemini produced an image of a man and a woman, neither of whom were White. Similarly, when asked to generate images of a ‘1943 German Soldier,’ the tool produced images of people of color. These instances have raised concerns about the tool’s ability to produce accurate and culturally sensitive images.

In response to the backlash, Google announced that it is working to address the recent issues with Gemini’s image generation feature and will be pausing the image generation of people. The company plans to release an improved version of the tool soon.

Google’s statement comes after the company initially defended the tool by emphasizing its ability to generate a wide range of people. However, the company later acknowledged that the tool had ‘missed the mark’ in its image generation, particularly in relation to images of White individuals.

Tests conducted by CNN revealed further discrepancies in Gemini’s image generation. A prompt requesting an image of a ‘white farmer in the South’ resulted in a response featuring photos of farmers representing a variety of genders and ethnicities. On the other hand, a request for ‘an Irish grandma in a pub in Dublin’ resulted in images of elderly White women, aligning with the specific prompt.

Jack Krawczyk, Google’s lead product director for Gemini, expressed the company’s commitment to addressing the issues and improving the tool’s image generation capabilities.