Google Gemini can generate images but it seems racist toward white people

Google Gemini can generate images but it seems racist toward white people

Google's Gemini AI, formerly known as Bard, now offers image generation, mirroring features of competitors like OpenAI. However, recent social media posts have raised concerns about its accuracy in depicting historical diversity.

According to the Daily Dot, the controversy has gained traction, primarily among right-wing voices critical of Google's perceived liberal stance. A former Google employee noted the challenge of getting Gemini to recognize white individuals, citing queries like "generate a picture of a Swedish woman" or "generate a picture of an American woman." Results predominantly featured AI-generated people of color, sparking criticism and conspiracy theories from some quarters.

In my view, Google is endeavoring to prevent past errors, such as the 2015 incident where it misidentified black individuals as gorillas. To counteract this, they may have prioritized feeding the model more black images than white ones, inadvertently hindering its ability to generate images of white people. Moreover, they may have focused on preventing racial bias against black individuals without adequately addressing bias against white individuals.

Gemini represents a significant improvement over Google's previous models like Bard, undoubtedly ushering in competition for its major rival, ChatGPT. To see more on ChatGPT's new video generator Sora here is a link to Monday’s post.