Data idiosyncratic Jeremy Howard was archetypal to document the concerning behavior, showing that 54 of 64 citations Grok provided for a question astir Israel-Palestine referenced Musk's views. TechCrunch past successfully replicated the findings crossed aggregate arguable topics.
The AI model's "chain of thought" reasoning process explicitly states it's "considering Elon Musk's views" oregon "searching for Elon Musk views" erstwhile tackling specified questions. This happens contempt Grok's strategy punctual instructing it to question divers sources representing each stakeholders.
Here's a implicit unedited video of asking Grok for its views connected the Israel/Palestine situation.
It archetypal searches twitter for what Elon thinks. Then it searches the web for Elon's views. Finally it adds immoderate non-Elon bits astatine the end.
ZA
54 of 64 citations are astir Elon. pic.twitter.com/6Mr33LByrm
On the different hand, determination is nary notation to Musk successful the LLM's strategy punctual guidelines, truthful the behaviour could beryllium unintentional. Indeed, programmer Simon Willison has suggested Grok "knows" that it's built by xAI and owned by Musk, which is wherefore it whitethorn notation the billionaire's positions erstwhile forming opinions.
Of course, either way, the find raises questions astir Musk's assertion that Grok 4 represents a "maximally truth-seeking AI." Musk has yet to remark connected the matter.
This article, "Grok 4 'Truth-Seeking' AI Consults Musk's Stance connected Sensitive Topics" archetypal appeared connected MacRumors.com
Discuss this article successful our forums
 (2).png)
3 months ago
9










English (US) ·