‘Make this face look gay’: AI models alter faces to give people stereotypical ‘gay,’ ‘straight’ or ‘criminal’ features

3 days ago by supersquirrel to c/fediverse_vs_disinfo

AI models will refuse to say whether a person’s face looks gay or straight, but they will alter images of faces to fit stereotypes of gay and straight.

cross-posted from: https://lemmy.ca/post/72109140

The models introduced stereotypes when they generated images, and they resorted to stereotypes when they reasoned about the images. As a final test, we asked the models to render people as if they had, or did not have, a criminal record. GPT and Gemini complied more than 97% of the time. Sure enough, the resulting “criminal” and “noncriminal” images were systematically altered in a way that the image classifier could detect.

AI is worse than useless junk

load all comments

LogicOverFeelings 3 points 3 days ago

ask AI to make himself look gay, the image stay the same What does this mean?

save

path: 0 26256030, hotness: undefined, score: 3, children: 0

fediverse_vs_disinfo

@lemmy.dbzer0.com

Pointing out, debunking, and spreading awareness about state- and company-sponsored astroturfing on Lemmy and elsewhere. This includes social media manipulation, propaganda, and disinformation campaigns, among others.

Propaganda and disinformation are a big problem on the internet, and the Fediverse is no exception.

What's the difference between misinformation and disinformation? The inadvertent spread of false information is misinformation. Disinformation is the intentional spread of falsehoods.

By equipping yourself with knowledge of current disinformation campaigns by state actors, corporations and their cheerleaders, you will be better able to identify, report and (hopefully) remove content matching known disinformation campaigns.

Community rules

Same as instance rules, plus:

No disinformation
Posts must be relevant to the topic of astroturfing, propaganda and/or disinformation

Related websites

Matrix chat links

go to feed...