In a rather bizarre incident, a 24-year-old Asian-American MIT graduate, Rona Wang, discovered while using an AI image generator that the bot’s idea of “professionalizing” her headshot was to alter her appearance to look white.
This meant removing her Mongoloid features, such as the shape of her eyelids, the color of her eyes, her skin tone, the shape and color of her lips, the body of her face, and the color of her hair. It meant giving her a completely new and alien look.
This incident gained significant attention when she shared it online, leading to viral discussions and media coverage.
Playground AI, the company behind the image editor, received criticism for the AI’s biased output. The CEO, Suhail Doshi, responded to the backlash with rhetorical questions, suggesting that one instance of bias does not necessarily indicate a systematic problem in the AI’s training.
“If I roll a die just once and get the number 1, does that mean I will always get the number 1?” he said. “Should I conclude based on a single observation that the dice is biased to the number 1 and was trained to be predisposed to rolling a 1?”
However, Wang and others have highlighted that racial bias is a recurring issue in AI tools. While she had been using AI image generators for some time, this incident felt more personal and disturbing. Wang emphasized the importance of software developers being aware of these biases and taking steps to mitigate them.
She initially hesitated to label the AI as explicitly racist but later expressed her concerns about the racial biases in AI tools. Many others have faced similar issues, where AI image generators have inaccurately depicted them or produced racist outputs.
“Racial bias is a recurring issue in AI tools,” she said. “I haven’t gotten any usable results from AI photo generators or editors yet,” she said
This incident is another example of AI systems making significant errors, despite companies rapidly integrating such software into various domains, including workforce replacements. The imperfections and biases in AI technology highlight the need for ongoing improvement and ethical considerations when developing and deploying AI tools.
