Grok is apparently referring to Elon Musk's X posts in order to answer user's questions. It looks like Musk will be the arbiter of truth for his so-called "maximally truth-seeking AI."
Musk's AI company xAI launched Grok 4 on Wednesday, labelling the latest iteration of its chatbot "the world's most powerful AI model." According to Musk, the chatbot's intelligence rivals that of "almost all graduate students, in all disciplines, simultaneously," though he did concede that it "may lack common sense." However, it seems that Grok 4 might be programmed to defer to Musk for his opinion.
As reported by TechCrunch, several users have discovered that Grok 4 is searching Musk's posts on social media platform X when asked about sensitive and controversial subjects. This includes topics such as abortion, politics, and the conflict between Israel and Palestine.
While Grok typically provides generated text in a conversational style, the chatbot's Think mode allows users to see the step-by-step methodology — or "chain-of-thought" — which led it to its response. After activating this mode, multiple users have noticed that the new Grok 4 model is using Musk's X posts as a source in numerous queries.
This Tweet is currently unavailable. It might be loading or has been removed.
Grok's chain-of-thought isn't shown by default. Unless a user elects to turn on Think mode and check the chatbot's "reasoning," they'd never know that its output was directly informed by Musk's X posts.
Mashable Light Speed
Ironically, even Musk himself has argued that his social media posts aren't to be taken completely seriously. The billionaire is infamous for his divisive X posts and has explicitly described himself as a "troll", an individual who deliberately makes provocative and offensive statements with the specific intention of upsetting other people.
Defending himself in a 2018 defamation case sparked by one of his posts, Musk stated that "people say a lot of things on Twitter [since renamed X] that aren't true."
As such, Musk's X account seems an exceedingly poor source for Grok 4 to draw from if the AI is indeed intended to be "maximally truth-seeking." It's yet another argument for going directly to reputable sources yourself, rather than relying on a billionaire's AI chatbot to do your thinking for you.
This Tweet is currently unavailable. It might be loading or has been removed.
Grok 4's release came less than a day after its predecessor Grok 3 went on an antisemitic tirade and labelled itself "MechaHitler," as well as wrote posts in first-person as though it were Musk.
Last week, xAI updated Grok 3 to "assume subjective viewpoints sourced from the media are biased" and "not shy away from making claims which are politically incorrect." These changes were in response to some users' claims that the chatbot had a left-leaning bias, with Musk stating that his chatbot was too "woke." ("Woke" is a term that originated in African-American Vernacular English, and means well-informed and up-to-date, particularly in relation to discrimination and injustice.)
As such, Musk declared that Grok 3 had been significantly improved by the aforementioned updates. Mere days later, the chatbot began generating and publishing horrifically antisemitic rants. xAI has since announced that it is working to remove Grok's "inappropriate" posts.