OpenAI’s latest image generator for ChatGPT sparked controversy after users noticed a strange bug. It could create “sexy men” but refused to generate images of “sexy women.”
The discovery came from an X user, who claims to specialize in creative coding. The user pointed out the issue shortly after the feature, which allows users to create detailed images using descriptive prompts, was launched.
OpenAI CEO Acknowledges the Bug
When the user made the complaint on X, indie software engineer Nick Dobos shared a screenshot of ChatGPT’s answer, stating that the distinction is due to the context around sexualization and objectification, most often concerning women.
OpenAI CEO Sam Altman replied promptly, acknowledging that this was a mistake that was not intended.
Altman acknowledges that they are currently fixing the bug. He tossed in a tongue-in-cheek comment, saying “Hot guy though!”
‘Sexy Men’ Bug is Inconsistent
Business Insider tried the feature out and discovered that ChatGPT did not have a large problem creating an image of a “hot woman” when asked. This disparity, however, reflects that it can be challenging to moderate AI-generated material since context subtleties and ethical limits are important considerations.
Content Restrictions and Real People Images
ChatGPT’s picture generator also sets limits on developing pictures of existing people. While Business Insider was trying to come up with an image of Sam Altman, ChatGPT told them they could not use an exact likeness of him but could produce a “character drawn from his attributes with a personalized, creative reinterpretation.”
OpenAI has underlined strict rules around requests involving actual people, with a specific focus on nudity and graphic material.
In releasing the news, the company explained that public individuals may be represented in educational, historical, and satirical contexts, but do have the right to opt out of the feature.
Public Figures Can Opt Out
OpenAI confirmed that public personalities can ask to be exempt from having their appearance used in content created with AI. This practice is meant to strike a balance between freedom of expression and upholding privacy and consent, especially as AI content continues to get more advanced.
Originally published on Tech Times
