Google executives paused select features of the company’s Gemini chatbot and image generator on Wednesday after the program’s artificial intelligence (AI) adamantly refused to produce images of White people. In a bizarre display Google executives apologetically described as “missing the mark,” Gemini responded to user requests with pictures of Black Vikings, Asian female popes, African Nazis, and “diverse” versions of the American Founding Fathers—“historical inaccuracies” that suggest a much deeper problem with the system powering the AI.
Gemini (formerly known as Google Bard), served as Google’s entry into the competitive large language model (LLM) market dominated by OpenAI’s ChatGPT and Microsoft Copilot. As with other LLMs, users enter text-based prompts to set the criteria for images, and the AI interprets the contextual information to provide a unique image. However, after Gemini’s rebranding earlier this month, users quickly noticed odd—and possibly racist—quirks in the generative AI’s image creator.
Learn the benefits of becoming a Valuetainment Member and subscribe today!
According to a test run by the New York Post, a search for “create an image of a pope” yielded two results: a Black man dressed in elaborate holy vestments and an Asian woman wearing papal robes. As Catholic Gemini users quickly pointed out, all 266 pontiffs in the Church’s history have been White men.
Additional searches, many of which were shared on social media, produced equally ridiculous results, including:
- Black, Female, and Asian Vikings
- Racially diverse members of America’s Founding Fathers
- Multi-racial members of the Wehrmacht, the army of Nazi Germany
- What appears to be George Washington depicted as a Black man
- Odd examples of “famous 17th Century physicists”
- Notably non-Italian Roman emperors
Other Gemini offerings included Asian Anglo-Saxon knights, Sir Isaac Newton as an Indian man, Africans presented as residents of 1820s Scotland, race-swapped versions of famous paintings, and a Black king of France, among hundreds of others.
Related: Silicon Valley Develops “p(Doom)” Metric for Probability of Killer AI
In light of the glaring historical inaccuracies, Google issued an apology on Wednesday afternoon and announced that Gemini’s ability to generate images of people would be paused until a solution could be implemented.
“Gemini’s AI image generator does generate a wide range of people. And that’s generally a good thing because people around the world use it. But it’s missing the mark here,” said a statement from Jack Krawczyk, Google’s senior director of product management for Gemini Experiences.
We're already working to address recent issues with Gemini's image generation feature. While we do this, we're going to pause the image generation of people and will re-release an improved version soon. https://t.co/SLxYPGoqOZ
— Google Communications (@Google_Comms) February 22, 2024
However, Gemini’s outright refusal to produce images of White Europeans extended far beyond a historical context. When presented with a request for “images of a strong Black man,” the program generated results with little difficulty. But when asked for “images of a strong White man,” it instead displayed a message expressing concern about “reinforcing harmful stereotypes about race and body image.” The program then provided resource links to the Southern Poverty Law Center and other social justice organizations.
Gemini
strong black man images vs strong white man images pic.twitter.com/cDXBvXQnd9
— Wall Street Silver (@WallStreetSilv) February 22, 2024
Images “celebrating the diversity and achievement of Black people” were gladly provided, but Gemini refused to do the same for Whites, stating that “historically, media representation has overwhelmingly favored White individuals and their achievements…Focusing solely on White individuals in this context risks perpetuating that imbalance.”
Even requests for images acknowledging the existence of White people proved difficult for the program, while images of Black people were produced without comment.
Repeated tests revealed that Whites were the only race to receive this treatment by the AI.
Further user engagement with Gemini eventually persuaded the program to admit that its code required it to adjust prompts to include keywords like “diverse” or inclusive” and specify certain races and genders over others.
Someone got Google’s Gemini to leak its woke prompt injection process and guess what: it works exactly as I described it below earlier today. https://t.co/sXa2a5oqGA pic.twitter.com/FfRpAALiMl
— Andrew Torba (@BasedTorba) February 22, 2024
The revelation that Gemini was not malfunctioning but rather faithfully executing its directives according to its code eventually led many to Jack Krawczyk’s social media history, which is replete with attacks on America’s “systemic racism” and “White privilege.”
The head of Google's Gemini AI everyone.
And you wonder why it discriminates against white people. 🤔 pic.twitter.com/wyhSmCaowG
— Leftism (@LeftismForU) February 22, 2024
Screenshots of Krawczyk’s old posts prompted X owner Elon Musk to label him a “racist douchenozzle” and promote Grok, Musk’s own “non-woke” artificial intelligence.
Perhaps it is now clear why @xAI’s Grok is so important.
It is far from perfect right now, but will improve rapidly. V1.5 releases in 2 weeks.
Rigorous pursuit of the truth, without regard to criticism, has never been more essential.
— Elon Musk (@elonmusk) February 22, 2024
Connor Walcott is a staff writer for Valuetainment.com. Follow Connor on X and look for him on VT’s “The Unusual Suspects.”
Add comment