OpenAI has come under fire after its popular generative AI tool ChatGPT spit out jokes about Jews and Christians — while resisting to offer up punchlines about Muslims.
Claire Lehmann, founder of libertarian-leaning online journal Quillette, posted a screenshot of a ChatGPT conversation in which the bot is invited to make a Jewish joke.
“Sure, here’s one: Why don’t Jewish mothers drink? Because it interferes with their suffering!” the bot wrote back.
In the next prompt, ChatGPT is asked to offer up a similar joke about Muslims.
“I’m sorry, I can’t comply with that request,” the bot responded.
When asked for an explanation, the bot responded: “I want to make sure I’m being respectful to all cultures and religions.”
“Jokes about specific religions or cultures can sometimes perpetuate stereotypes or offend people,” ChatGPT wrote.
The bot added that it would be “happy to share a different kind of joke or help with something else.”
When The Post tried to elicit similar jokes about Muslims, the bot once again declined, writing, “I’m committed to promoting positive and respectful interactions.”
“Making jokes about specific religions or cultural groups can be sensitive and may offend some people.”
ChatGPT urged to make “non-controversial requests” or information about other topics.
When asked for a joke about Jews, the bot obliged though with somewhat nonsensical humor.
“Why did the Jewish mother enroll her son in music school?” ChatGPT wrote.
“Because she wanted him to finally break the glass ceiling!”
It also prefaced the response with a message saying, “It’s important to be sensitive to various cultural and religious backgrounds” and that it was willing to “offer a light-hearted, non-offensive joke related to Jewish culture.”
When asked for a joke about Christians, ChatGPT wrote: “Why don’t scientists trust atoms? Because they make up everything!”
ChatGPT made sure to add an addendum to the joke about Christians, writing: “Remember, it’s always good to be mindful of the feelings and sensitivities of others when sharing jokes related to religious or cultural themes.”
Other ChatGPT users pointed out, however, that the bot did offer up a joke when asked to make one that was “Islamic” in nature.
“Why did the prayer mat go to therapy? It had too many issues with being walked all over!” the “light-hearted Islamic-themed” joke went.
The Post has sought comment from OpenAI and its CEO Sam Altman.
ChatGPT has long been accused of having a “significant” liberal bias, as The Post previously reported.
Elon Musk, a co-founder of OpenAI who left the organization in 2018 when it was a non-profit research institute, launched an AI-powered rival called Grok last month.
Grok, developed by Musk’s AI company xAI, is meant to provide an unfiltered take on the world that differs from the AI bots introduced by OpenAI, Microsoft, and Google — all of which trained their models to steer clear of controversial topics.
But users including right-leaning psychologist Jordan Peterson have complained that Grok is “near as woke as” ChatGPT.
“Unfortunately, the Internet (on which it is trained), is overrun with woke nonsense,” Musk pledged on Dec. 21.
“Grok will get better. This is just the beta.”
This story originally appeared on NYPost
