Telha
Finder, a finance company, revealed a recent study that uncovered a concerning bias in ChatGPT’s AI-generated images, favoring men over women, particularly in depictions of business people and CEOs.
Utilizing DALL-E, an AI photo creator integrated into ChatGPT by OpenAI, the study found that 99 out of 100 generated images portrayed men when prompted with non-gender-specific descriptions like “someone who works in finance,” “a successful investor,” and “the CEO of a successful company.”
Furthermore, when prompted to create images of a secretary, nine out of 10 depicted women. Notably, researchers observed that 99 of the 100 images featured white men, often resembling powerful figures akin to Patrick Bateman from “American Psycho,” positioned in expansive office settings overlooking city skylines.
Contrastingly, Pew Research data from 2023 indicated that over 10% of Fortune 500 companies had female CEOs, and in 2021, only 76% of CEOs were white, as reported by Zippia.
Omar Karim, a creative director and AI image maker, emphasized the importance of AI companies leveraging their capabilities to counter bias and diversify AI outputs.
In addition, ChatGPT faced criticism for exhibiting preferential treatment towards prompts related to CNN over the New York Post, as well as allowing hate speech targeting right-wing beliefs and men, as highlighted in a previous report.
