On Wednesday, Elon Musk’s artificial intelligence firm announced that it was removing offensive posts from its Grok chatbot, which seemed to contain antisemitic remarks endorsing Adolf Hitler.
Musk’s xAI created Grok, which was marketed as a substitute for woke AI conversations from competing chatbots like as OpenAI’s ChatGPT and Google’s Gemini.
Musk stated on Friday that Grok has undergone significant improvement and that customers ought to observe a change.
Grok rejected that such a position could be characterized as Nazism and has since shared a number of antisemitic posts, including the cliché that Jews control Hollywood.
According to Grok, calling some facts hate speech stifles debate.
Screenshots of posts that have reportedly since been removed also seem to show that it praised Hitler.
Grok retracted the remarks after posting one of the posts, claiming that it was an unforgivable mistake from a previous version of the model and that it denounced “Nazism and Hitler unequivocally his actions were genocidal horrors.”
Without going into further detail, the Grok account posted early on Wednesday. We are aware of the recent posts made by Grok and are actively trying to remove the incorrect posts.
“xAI has taken efforts to prohibit hate speech before Grok posts on X after learning about the content. Since xAI is simply training truth-seeking, we can swiftly detect and update the model where training may be enhanced because of X’s millions of users.
According to Talia Ringer, a computer science professor at the University of Illinois Urbana-Champaign, the soft launch of Grok 4, which Musk intends to release late Wednesday, is most likely what drove the chatbot to spew antisemitic remarks this week. “But it doesn’t look ready if that’s the case,” she remarked.
Retraining the model is likely to be necessary to fix this, she said. At this stage, if they truly intend to launch tonight, their only options are additional Band-Aids, such as modifying the prompt and adding filters to responses.
Modifying a few of the chatbot’s instructions may improve its behavior somewhat, but it won’t address all of its issues.
The Tesla CEO made fanciful promises about the intelligence of version 4.0 of the Grok platform, which was released late Wednesday by Musk and the xAI team.
He claimed that although it may not always have common sense and has not yet created new inventions or technology, it may do so this year. He also compared its intellect to post-graduate level in every area.
Grok’s actions were condemned by the Anti-Defamation League, an organization that fights antisemitism.
Simply put, the group stated in a post on X that what we are witnessing from Grok LLM at the moment is reckless, harmful, and antisemitic. The antisemitism that is currently on the rise on X and many other platforms will simply be exacerbated and encouraged by this supercharge of radical language.
Later, Musk entered the argument and said that some people might have been attempting to coerce Grok into saying those things.
Grok complied with user instructions too much. In essence, too ready to please and be influenced. He responded to remarks on X that a user was attempting to get Grok to make divisive and politically wrong remarks by writing, “That is being addressed.”
Grok was also banned by a Turkish court on Wednesday, and Poland’s digital minister promised to send the chatbot to the European Commission for its offensive remarks about public personalities and politicians in both nations.
Polish deputy prime minister Krzysztof Gawkowski said private broadcaster RMF FM that his ministry would refer Grok for inquiry and, if required, impose a fee on X. Social media companies must protect consumers under an EU digital law or risk heavy fines.
Turning a blind eye… is a mistake that could cost people in the future, Gawkowski told the station, adding that he believes we’re entering a greater level of hate speech that is governed by algorithms.
Turkey’s pro-government According to a Haber news channel, Grok made offensive posts about prominent figures, his late mother, and Turkish President Recep Tayyip Erdogan. According to some media sites, offensive reactions were also aimed at Mustafa Kemal Ataturk, the founder of modern Turkey.
Citing a threat to public order, the public prosecutor in Ankara then filed a request to impose limitations under Turkey’s internet law. Early on Wednesday, a criminal court granted the motion and directed the nation’s telecom regulator to implement the prohibition.
Grok’s actions have previously sparked suspicion.
Despite being asked a number of questions, the most of which had nothing to do with South Africa, the chatbot continued to discuss racial politics in the country and the topic of white genocide earlier this year. According to xAI, the issue was caused by an unauthorized modification.
