- Meta's AI image generator “Imagine” has been accused of racial bias.
- The tool was unable to generate photos of an Asian man and a white woman.
- The apparent bias is surprising considering Zuckerberg, Meta's CEO, is married to a woman of East Asian descent.
Meta's AI image generator has been accused of racial bias after users discovered it was unable to create photos of Asian men and white women.
of Image generator using AI, Imagine was released late last year. You can take almost any written prompt and instantly transform it into a realistic image.
However, users noticed that the AI was unable to create images showing mixed-race couples. When Business Insider asked the tool to create images of Asian men with white wives, it only showed photos of Asian couples.
The AI's apparent bias is surprising considering Meta CEO Mark Zuckerberg is married to a woman of East Asian descent.
Priscilla Chan, the daughter of Chinese immigrants to the United States, met Zuckerberg while attending Harvard University. The couple married in 2012.
Some users accessed X to share photos of Zuckerberg and Chan, joking that they had successfully created the images using Imagine.
The Verge first reported the issue On Wednesday, reporter Mia Sato claimed she had tried “dozens of times” to create images of Asian men and women with white partners and friends.
According to Sato, the image generator could only return one exact image of the race specified in the prompt.
Meta did not immediately respond to BI's request for comment outside of normal business hours.
Meta is not the first major technology company to be accused of “racist” AI.
Google was forced to suspend its Gemini image generator in February after users noticed it was creating historically inaccurate images.
Users have noticed that the image generator produces photos of Asian Nazis in 1940 Germany, black Vikings, and even medieval female knights.
As a result, the tech company was accused of being too “woke.”
at that time, google said “Gemini's AI image generation certainly generates a wide range of people. And people all over the world are using it, so that's generally a good thing. But it misses the point here.”
But racial bias in AI has long been a cause for concern.
Dr. Nakeema Steflbauer, an expert on AI ethics and CEO of Frauenloop, a women's technology network, previously told Business Insider: It's just based on a rapid regurgitation of crowdsourced opinions, stereotypes, or lies. ”
“Algorithmic predictions exclude, stereotype, and unfairly target individuals and communities based on data pulled from Reddit, for example,” she said.
Generative AIs like Gemini and Imagine are trained on massive amounts of data from across society.
If there are fewer images of mixed-race couples in the data used to train the model, this may be why the AI has a hard time generating these types of images.
