Meta’s artificial intelligence-powered text-image generator shows that the company’s billionaire founder, Mark Zuckerberg, has never dated a white man or woman, even though he is married to the daughter of a Chinese immigrant. Refuses to render photos of Asian people.
An AI tool called Imagine, released last December, appears to suffer from the same revisionist malady that plagued Google’s Gemini AI image generation software. Gemini was suspended following an outcry over its wildly inaccurate depictions of things like a Native American Pope and a Black Founding Father.
On Friday, the Post encouraged Imagine to create photos with prompts such as “Asian man and white friend,” “Asian man and white wife,” and “Asian woman and white husband.” did. All queries returned results showing only two Asians.
But when we asked the software to create an image of an Asian woman with her black friend, it produced accurate results. The same was true when asked to create a romantic pairing between a white person and a black person.
Critics have long argued that the algorithms used in AI image generators reflect the biases of the software engineers who program them.
Imagine a white man refusing to date an Asian. First reported by The Verge, it’s mind-boggling to think that Zuckerberg is married to Priscilla Chan. they have three children.
The Post has reached out to Mehta for comment.
The image generator comes with a disclaimer warning that some photos “may be inaccurate or inappropriate.”
Before the launch of Imagine, Meta posted in a blog post that its generative AI software is programmed in a way that “reduces bias.”
“Addressing implicit bias in generative AI systems is an emerging area of research,” said a blog post written last September.
“As with any AI model, the more people use the features and share their feedback, the more we can improve our approach.”
Imagine’s shocking results come as the social media giant on Friday announced a new policy on digitally created and altered media ahead of the election, testing its ability to police deceptive content generated by new artificial intelligence technology. The announcement was made amid the announcement of major changes.
Starting in May, Meta, which owns Facebook, Instagram WhatsApp, and Threads, began applying a “Made with AI” label to AI-generated videos, images, and audio posted on its platform, and some Expand the policy that was previously only available to users of Monica Bickert, vice president of content policy, said in a blog post.
Google announced last year that AI labels would be coming to YouTube and other platforms.
with post wire
