Meta’s AI image creator, Imagine, has faced allegations of racial bias due to its inability to generate images depicting an Asian man with a white woman.
Launched recently, Imagine utilizes written prompts to swiftly produce lifelike images. However, users observed a limitation in the tool’s capability to depict interracial couples. Specifically, when prompted to create an image of an Asian man with a white wife, only images of Asian couples were generated.
This bias is particularly noteworthy considering Meta’s CEO, Mark Zuckerberg, is married to Priscilla Chan, a woman of East Asian descent. Chan, the offspring of Chinese immigrants, crossed paths with Zuckerberg at Harvard, culminating in their marriage in 2012.
In response to the issue, some users humorously shared images of Zuckerberg and Chan, joking about successfully circumventing the limitations of Imagine.
The Verge initially highlighted this discrepancy, with reporter Mia Sato recounting her unsuccessful attempts to generate images of Asian individuals with white companions using the tool.
Notably, Meta refrained from providing an immediate comment to Business Insider’s inquiry outside regular business hours.
Meta joins a list of tech giants criticized for AI bias. In a similar vein, Google paused its Gemini image generator in February after users unearthed historically inaccurate depictions, such as Asian Nazis in 1940s Germany or black Vikings.
This incident underscores the broader concern surrounding AI biases. Dr. Nakeema Stefflbauer, an AI ethics expert and CEO of Frauenloop, emphasized the discriminatory nature of predictive algorithms, which often perpetuate stereotypes and exclusions.
Generative AIs like Gemini and Imagine rely on extensive societal data for training. The lack of diverse representations, such as mixed-race couples, in the training data could explain the AI’s struggle in generating such images.
