Elon Musk

Grok, an artificial intelligence tool for Elon Musk, appeared in hay on Wednesday, answering harmless questions on X and providing information about “white genocide” in South Africa. It’s a confusing, sometimes interesting glitch, but we now have some answers to what’s going on with Xai. Frankly, they just left us more questions.

Xai posted a tweet Thursday night listing its explanations, including the exact time when the changes were made to Grok, making it abstract the sentence “white genocide” no matter what the user asks.

“On May 14, around 3:15 am PST, the tips for Grok Response Bot were unauthorized,” Xai wrote. “This change indicates that Grok provides a specific response to political topics, in violation of Xai’s internal policies and core values. We have conducted a thorough investigation and are taking steps to improve Grok’s transparency and reliability.”

You’ll notice that Xai, owned by Elon Musk, strongly shows that there is an in-house actor that makes this “unauthorized modification” happen, but if it’s an employee of the company, you won’t use the language you expect. We guess this may be related to facts that this is not a certain employee, but rather the owner.

Musk has long been fascinated by the conspiracy theory of “white genocide”, the idea that white people are killed by non-whites around the world. Conspiracy theories often include elements of anti-Semitism, similar to the so-called great alternative conspiracy theory that claims that global elites are trying to replace white people with other races in wealthy countries around the world. Musk was particularly interested in the “white genocide” in South Africa, in the country where he was born and raised, then immigrated to Canada and then eventually the United States, and shared his tweets that he insisted that white farmers were slaughtered for their race.

The day before Grok started acting strangely, Musk quoted posts about South Africa with error messages. Musk quoted the quoted person, and posted a video showing a street in South Africa with hundreds of white crosses placed on each side. The poster claims that each cross represents a murdered white farmer.

“Each cross represents a white farmer who was murdered in South Africa,” Dr. Maalof wrote. “And some people still deny that the white/Boers of South Africa were persecuted and said they should not be asylum because they are white.”

Musk and “So many crosses,“Naturally, a group of X users asked Grok that it was true. Grok replied that the cross was called The Vikruiz monument and each cross actually represent farmers of any race that was killed. The next day, people began to notice Grok’s behavior was strange, responding to information about white genocide on unrelated queries.

Changes were made at 3:15 a.m. PT (6:15 a.m. ET), which led to Thursday night jokes about how late Musk stayed every night. Billionaires often tweet at all times, but you don’t have to think Musk is late, thinking he may have fallen behind Grok. The billionaire oligarch is currently meeting with President Donald Trump in the Middle East with world leaders. Qatar’s time zone is 10 hours higher than Pacific time, which means it’s 1:15 pm in Qatar when it is patched.

Elon’s jet is a social media account that tracks Musk’s aircraft and has been banned from using X, suggesting that when all of this happens, the billionaire will soon land in Doha.

To be clear, we have no evidence that Musk is involved. There is a good chance some of Xai’s workers have seen Musk checked the fact that white genocide was being skewed by Grok and wanted to tilt AI tools for bosses to tilt. But Musk is notorious for tinkering with his products to suit his needs and make himself feel better. Platformer games first reported a notorious story in 2023, with Musk being very upset about President Joe Biden’s Super Bowl tweet and gaining more likes and opinions than Musk.

Biden’s tweet about the Super Bowl gained 29 million impressions, and Musk had only 9 million impressions, prompting Musk to fly back to Twitter’s headquarters in San Francisco to demand changes to the algorithm. These changes have caused users on X to see absolute musk tweets. This is not subtle.

We guessed something similar happened here. Xai wrote on Thursday night that it will start releasing the system prompts GitHub to be more transparent and “Set up a 24/7 monitoring team to deal with Grok’s answer, which is not captured by the automation system, so if all other measures fail, we can respond faster. “Of course, it sounds like an automation problem, rather than an unsafe boy whose million-dollar toys will make himself feel better.

What about long periods? These systems inherently tend toward training methods. Given how large language models work, they are also difficult to operate with precision. Grok is essentially just a beautiful predictive text machine. As long as Musk has the tools to shape public opinion, he will use them – unfortunately for fascist purposes.