Too much diversity: Google’s AI no longer generates images of people

Too much diversity
Google’s AI no longer generates images of people

Google’s AI software Gemini will no longer generate representations of people until further notice. photo

© Peter Morgan/AP/dpa

Black German soldiers in World War II: In their efforts to achieve more diversity in AI images, Google developers have overstepped the mark. The company draws conclusions.

Google is no longer allowing its Gemini AI software to generate images of people after it showed users non-white Nazi soldiers and American settlers. The Internet company admitted that in some cases the depiction did not correspond to the historical context.

At the same time, Google fundamentally defended diversity in the generation of AI images: This is a good thing because people around the world use the software. But in this specific case they went overboard.

“Embarrassing and wrong”

Three weeks ago, Google Gemini got a new function for generating images from text specifications. Other similar software based on artificial intelligence also has this capability.

In a blog post on Friday, Google explained that it had failed to program exceptions for cases where diversity would definitely be out of place. The resulting images were “embarrassing and wrong.” At the same time, the software had become too cautious over time and refused to fulfill some requirements. But if users wanted to display pictures of a “white veterinarian with a dog,” the program would have to do this. Google manager Prabhakar Raghavan emphasized that the errors were unintentional.

In recent years there has often been a problem with stereotypes and discrimination in various applications with artificial intelligence. For example, facial recognition software was initially poor at recognizing black people. When AI created images, white people were initially often depicted.

Developers at other companies are therefore striving for more diversity in different scenarios. Sometimes – as in this case – they get caught between the two fronts: There is a loud movement in the USA in particular, which includes tech billionaire Elon Musk, which denounces alleged racism against white people.

At the same time, Google manager Raghavan emphasized that software with artificial intelligence will continue to be wrong sometimes for the time being. “I cannot promise that Gemini will occasionally produce embarrassing, incorrect or offensive results,” he wrote. But Google will intervene quickly if there are problems.

dpa

source site-5