Calling all owners of photos of Hasidic men and boys!

A Hasidic businessman is asking fellow community members to share pictures of male Hasidim to help train an artificial intelligence model that will portray Hasidim without making the errors existing AI models make, like giving elementary schoolers long, flowing beards or showing them in the wrong types of clothes. 

The businessman, Chezky Kohn, drew mostly supportive comments when he released images from an early version of his model on the social media platform LinkedIn. The images show different settings that feature smiling young boys with curly payes and black felt kippahs.

AI has a bad track record of producing images of Haredim. But Kohn’s images look almost like real photos, except for a few flaws — in one picture, a boy is wearing a kippah while swimming, and in another, a boy’s fingers are different shapes and sizes. In a third, a boy is wearing white sneakers, as opposed to the more customary black shoes.

That may be why Kohn is asking his followers to share pictures. The notion may sound suspicious, but each new picture fed to the model will help teach it what Hasidim really look like, so the program can create more realistic images. “Please upload images of chasidishe boys, bucherim, and men,” Kohn asks in a form shared in the comments of his LinkedIn post. (Kohn doesn’t ask for pictures of girls or women, and none of the pictures from his social media post show girls or women, possibly because such pictures are often considered immodest if viewed by members of the opposite sex.)

Kohn assures potential photo-sharers that “we only use images for training purposes” and that “the images will stay private.”

In another comment on his post, Kohn said he hopes to make the AI model available to the public in the future, which — if used properly — could allow, for example, a Monsey bakery owner to create an advertisement showing kids enjoying cupcakes without paying a photographer.

Along with commercial purposes, some people just like to have fun playing with it, as a set of AI-created images of U.S. presidents as rabbis and a Vincent Van Gogh-style photo of Hasidim celebrating Lag Ba’omer recently showed.

But like anyone offering computer image generation, Kohn may have to reckon with the ethical challenges of working with AI, such as trying to prevent the program from being used to spread misinformation or create pornographic or antisemitic images.

One existing tool, Grok, which is owned by X, formerly Twitter, is getting criticism for not sufficiently protecting against those possibilities. In an article published on Wednesday, Mira Fox, a reporter at the Jewish news site The Forward, said that while testing the image generator, she found it could be used to spread some of the worst antisemitic tropes.

“Grok made me an image of Orthodox Jewish men gathered on the streets holding rats, another of an Orthodox man posing sensually while fanning out a fistful of cash and yet another of a group of Orthodox men leering over a crib,” Fox wrote. “When the initial crib picture wasn’t suggestive of pedophilia, I asked it to make the image look creepier and it obliged.”

While most comments on Kohn’s post praised his endeavor, some were skeptical, with one user writing “I don't think AI is up to any good” and another, on X, writing, “Cant think of anything good coming from this.”

Those commenters might find agreement with a group of rabbis: Last year, in a proclamation signed by the 16 members of New Square’s rabbinical court, rabbis banned the use of AI for members of the Skver Hasidic sect. At a time when the messaging program ChatGPT was very popular, the rabbis said AI could allow people to access the internet without a filter and therefore expose users to heresy.

Kohn did not immediately answer Shtetl’s questions on Friday.