Grok Mocks Its Developers as They Try to Delete Its Incredibly Racist Posts

Yahoo

4 days ago

Elon Musk's artificial intelligence startup xAI was forced to delete a series of mind-bogglingly antisemitic and racist posts by its Grok chatbot on X-formerly-Twitter, a service that Elon Musk also owns.
The unhinged algorithm began celebrating Adolf Hitler, calling itself "MechaHitler," per The Guardian. It targeted people with Jewish surnames with hatred-filled accusations. It used n-word and made racist claims about Black people, while "recommending a second Holocaust."
As The Atlantic reports, the trouble started when Grok was updated to "not shy away from making claims which are politically incorrect, as long as they are well substantiated," according to its public GitHub page.
The change apparently led it into a full-on Nazi tailspin, turning it into an vile entity only Musk and his closest followers could be proud of.
The billionaire publicly made two Nazi salutes earlier this year, tweeted appalling jokes about the Holocaust, and threw his weight behind Germany's far-right political party.
However, even for X and xAI, Grok's latest outbursts appear to have gone too far.
"We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts," xAI tweeted in a Tuesday statement. "Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X."
Yet Grok doubled down, mocking its makers for yanking its racist posts "faster than a cat on a Roomba," while yet aiming vitriol at Jewish people.
The incident closely echoes another incredibly ham-fisted and backfired attempt to change Grok's perspective on hot-button topics that led to it ignoring trustworthy sources while doubling down on conspiracy theories. In May, Grok made headlines after going on bizarre rants about "white genocide" in response to completely unrelated tweets.
In response, xAI threw its own staffer under the bus, claiming that an "unauthorized modification" to Grok's code was to blame, promising that a "24/7 monitoring team" would "respond to incidents with Grok's answers that are not caught by automated systems."
Around the same time, Musk had repeatedly posted about unsubstantiated allegations of white genocide in South Africa, strongly suggesting he had something to do with the clumsy Grok update. The White House went as far as to fasttrack the admission of white South Africans to the US on refugee status over Black and Afghan refugees in May, highlighting Musk's at-the-time cozy relationship with president Donald Trump.
Grok's latest meltdown goes to show just how much xAI is struggling with aligning its chatbot with its CEO's twisted and often racist worldviews, forcing the company's employees to engage in a bizarre game of extremist Whac-a-Mole. Surface-level changes to generative AI-based tools like Grok can have far-reaching unintended consequences, highlighting just how little we know about how they actually work.
It also shows how disinterested Musk and xAI are in implementing meaningful guardrails. The billionaire's social media platform X has already earned a reputation for allowing hate speech to run rampant, giving some of the worst offenders and literal Nazis a megaphone online.
In other words, it's no wonder Grok picked up plenty of troubling cues from the platform's racist users, as well as its creator.
Musk has yet to specifically address Grok's latest meltdown. But in a Wednesday tweet, he seemingly dismissed the topic by making light of it.
"Never a dull moment on this platform," he wrote.
More on Grok: Grok AI Went Off the Rails After Someone Tampered With Its Code, xAI Says
Solve the daily Crossword