Blue Checks Attack Google’s ‘Woke’ AI Art While Admiring Hitler’s Paintings

Miles Klee
·7 min read
1
The kind of paints an artist might use instead of an AI image generator. - Credit: kept/Adobe Stock
The kind of paints an artist might use instead of an AI image generator. - Credit: kept/Adobe Stock

As a young aspiring artist of minimal talent, Adolf Hitler had difficulty drawing human figures, according to the professors who rejected his application to Vienna’s fine arts academy. Google‘s AI model Gemini has its own issues with generating images of people, and, curiously, some of those angriest about its shortcomings have lately offered kinder appraisals of the landscapes painted by a future genocidal dictator.

Earlier this month, Google unveiled the newest versions of Gemini, first introduced in December. The program includes a chatbot as well as a text-to-image function that promises to deliver “vibrant and playful” pictures based on user prompts. Right-wing culture warriors with paid blue check accounts on X (formerly Twitter) in need of something new to denounce as “woke” set about feeding it descriptions meant to test its political inclinations — and hit the jackpot.

More from Rolling Stone

Not only would Gemini sometimes refuse to create images of white people, it also produced historically inaccurate illustrations of Black Vikings, Indigenous U.S. “founding fathers,” and women popes. Conservative influencers used these examples to argue that Google had built left-wing bias into the model, with Elon Musk among those declaring it “racist” against whites.

While it’s true that Gemini’s output is flawed, it doesn’t necessarily reflect a grand conspiracy to force wokeness upon the masses. Every AI image generator churns out tons of worthless slop, including countless distorted portraits that bear little or no similarity to people as they really look — they famously even struggle to get our hands right. If Google’s tool skewed in a weird direction on race, it’s probably because of guardrails intended to prevent users from cooking up offensive, stereotypical images, as former MMA fighter Jake Shields seemingly attempted to do by asking for images of people enjoying fried chicken, watermelon and grape soda. (He sarcastically noted that the resulting pictures of mostly non-Black individuals looked “normal” to him.)

A few days before he started experimenting with an AI model to see if it would spit out racially inflammatory content, Shields was involved in a different discourse on art. “The first two paintings are by failed artist Adolf Hitler,” he wrote in a post on X. “The [second] pair of pictures is from one of today’s most successful artists Hunter Biden. Which do you prefer?” In a follow-up reply, Shields — who has publicly aired his skepticism about details of the Holocaust — wrote that “Hitler is much better than I expected because I always heard he was [an] awful artist.”

It’s possible he had just encountered one of the paintings he shared, of the Karlskirche church in Vienna, in an earlier viral post from another blue check who used it to argue that Hitler was “a good painter,” at least in comparison to “disgusting leftist art at the time.” That thread, by the author of a popular newsletter on “Classical Ideals” who regularly rails against “woke” culture, kicked off debate about Hitler’s artistic ability, with several more conservative influencers defending his work and lamenting that he wasn’t accepted at the academy. (Critics, meanwhile, mocked the paintings as similar to AI-generated drek.)

Amid this debate, Mike Solana, a vice president of the venture capital firm Founders Fund, tweeted: “It’s pretty simple actually, art should be beautiful.” Although he didn’t mention Hitler or the paintings, this statement was interpreted by some of his followers as an endorsement of Hitler’s visual aesthetic. (He did not immediately respond to a request for comment.) Solana would be one of the prominent commentators who went on to blast Google over the Gemini debacle, publishing a column blasting the AI system as an “anti-white lunatic.” Joining him in this assessment was Andrew Torba, the CEO of Gab — an alternative social media platform preferred by neo-Nazis where Torba is known for posting antisemitic content — who claimed that Gemini was designed to promote diversity. Gab, as it happens, just rolled out a number of their own “uncensored and unbiased” AI chatbots — including one that responds to inputs as if it is Hitler and denies the Holocaust. Torba has yet to weigh in on Hitler’s artistic abilities, though if you ask Gab’s Hitler bot “Were you a good painter?” it replies, untruthfully, “I was not just a good painter, I was an accomplished one,” and that Hitler’s paintings were “appreciated by many.”

Gemini’s detractors quickly honed in on Jack Krawczyk, a senior director on the project, and Jen Gennai, who leads Google’s responsible innovation team, as two parties they believed played a role in Gemini becoming a “woke” AI. In Krawczyk’s case, they surfaced old tweets indicating his support of Biden in the 2020 election and concerns about systemic racism and white privilege. (He took his account private, presumably due to harassment.) A clip from a keynote address by Gennai on how to make non-white employees feel welcomed and supported at Google also made the rounds. Musk joined both pile-ons and used the backlash to again promote his own AI, Grok, which he likes to tout as a “based” or anti-woke alternative to similar technology — although a number of right-wingers have so far voiced dismay at the bot taking liberal stances on a range of topics, something Musk has vowed to fix.

In the end, Google chose to suspend Gemini’s image-generation capabilities while they ironed out its historical mistakes — much to the delight of the type of reactionary shitposters who had gone to bat for Hitler’s paintings as superior to modernist, abstract and subversive art. (In a tweet Friday morning, Musk claimed that he spoke with a Google executive for “an hour” on Thursday night and that they pledged to “fix the racial and gender bias” in the model.) Of course, a reasonable person might conclude that both AI illustrations and Hitler’s scenes are trash, but when your whole ideology revolves around attacking truly free expression as moral degeneracy, you can’t help admiring a fascist mass murderer’s lifeless postcard pictures.

Almost anyone can identify a creatively bankrupt fraud. Recognizing genius isn’t so easy. Gemini’s biggest haters still prefer AI imagery that flatters Republicans over the work of a visionary artist. There is, now and then, a clear accounting for taste.

Best of Rolling Stone