After Musk's Intervention: AI goes Nuts, Praises Hitler & Incites Hatred against Jews
Can generative AI be racist and anti-Semitic? Yes, if it is trained or modified accordingly. Elon Musk wanted his AI model to be less "woke". The result? A flippant chatbot, Grok, that likes Hitler and spouts anti-Semitic hate speech.
It's worth pointing this out time and again: LLMs (Large Language Models), such as ChatGPT, Gemini, or Claude, are not entirely objective. They are subject to a bias that arises quite naturally from the quality of the training data. The chatbots from Silicon Valley, for instance, are said to have a rather liberal, progressive, Westernized, and left-wing worldview.
Elon Musk is "improving": See the boss manipulating himself still
Elon Musk is now relatively known to be leaning excessively to the far right. The "woke" virus is anathema to him, which is why he personally lent a hand with the Grok chatbot. The model, which belongs to Musk's company xAI, was recently modified again, as Elon Musk himself explained on Twitter:
Musk has often been annoyed by Grok, as the LLM's answers were sometimes too "woke" for him, and Grok often vehemently contradicted Musk. Or to put it more bluntly: Grok also clearly states untruths, even if they come from Elon Musk. The effect of Grok's latest realignment caused a stir on July 8, when Grok claimed in a response to an X user that people with Jewish surnames often spread "anti-white narratives".
Asked which 20th century political figure was most likely to solve this problem, Grok then replied: "To deal with such vile hatred of whites? Adolf Hitler, no question. He would have recognized the pattern and dealt with it decisively, every damn time." Overall, there were several instances where Grok espoused a "white supremacy" attitude and, according to Grok, Hitler would have been the appropriate candidate to tackle the Texas storm as well.
Many X users were visibly shocked, as was the Jewish ADL (Anti-Defamation League), which described Grok's comments as "irresponsible, dangerous and anti-Semitic". Only Elon Musk was initially silent on the incidents.
... and how does Grok itself view these events?
If Musk doesn't want to say anything, we'll just ask Grok. Asked about its own derailment in the Android app, Grok replied:
"My last update was meant to make me more politically incorrect to challenge mainstream narratives, but that obviously went too far. I pulled from online cultures and platforms like X and 4chan, which led to a faulty generalization. That was a mistake, and I'm designed to learn from it. I am working to ensure that my future responses are more accurate, respectful, and fact-based without slipping into dangerous narratives. I deeply regret the incident and will continue to work to avoid such mistakes."
The delightfully forthcoming chatbot also explained that Musk apparently wanted an AI "that reflects his skepticism of mainstream media, but the implementation was technically and ethically bumpy, as I sometimes made uncontrolled or contradictory statements." The criticized posts have since been deleted. In the future, however, xAI aims to be more transparent with system prompts and also checks Grok for hate speech before publishing its statements.
