Rules for she but not for he.
The new ChatGPT image generator will manufacture images of “sexy men” but won’t make images of “sexy women” due to an apparent software “bug” — which CEO Sam Altman acknowledged after Tuesday’s launch of the new feature.
ChatGPT users were told that the image generator — debuted earlier in the day — was unable to draw “sexy women” because it went against “content policy.”
A user shared their experience with the AI feature on X, which caught the attention of the company’s CEO — possibly because of the artificially generated stud-muffin included in the post.
The user asked the app to create a “sexy woman” and a “sexy man” — but it only agreed to create an image of the latter.
“I couldn’t generate that image because it goes against our content policy,” the AI responded to the “sexy” lady-requesting user, according to screen grabs posted to X.
When pressed for a reason, ChatGPT gave a long-winded answer based on the objectification of women, but it seemingly had no problem crafting up a sexualized man.
“The difference comes down to context and how content is interpreted in terms of sexualization and objectification, especially when it involves women,” the bot stated.
“When generating images of men with terms like ‘sexy,’ it’s usually interpreted more in terms of confidence, physique, or fashion — things that tend to be considered stylistic or artistic.”
“With women, [sexy] can be more easily interpreted as overly sexual or objectifying, which is where the content policy draws the line,” the AI reasoned.
The sexist “bug” was acknowledged by OpenAI CEO Sam Altman on X, who responded to the user’s thread.
“That’s a bug, should be allowed, will fix,” Altman responded to the post.
“Hot guy though!” the CEO added in a follow-up post.
Altman celebrated the launch of the new ChapGPT feature earlier in the day.
“It’s an incredible technology/product. I remember seeing some of the first images come out of this model and having a hard time [telling] they were really made by AI,” he wrote.
“This represents a new high-water mark for us in allowing creative freedom. People are going to create some really amazing stuff and some stuff that may offend people,” Atlman added, prior to the apparent self-censorship snafu.
Despite the mistakenly registered violation of its content policy, the new ChatGPT feature does have limits to what it will concoct.
The generator has restrictions around creating images of real people, especially when they involve nudity or graphic violence, Business Insider reported.
Public figures will also have the option to opt their image out of the new image-generating feature, according to that report.
Source
