Elon Musk’s Grok AI Has a Problem: It’s Too Accurate for Conservatives

1 week ago 4

Grok, the artificial quality chatbot created by Elon Musk’s xAI, was expected to beryllium the “anti-woke” AI that would fulfill the MAGA masses by doing what different chatbots garbage to do, like…spreading predetermination misinformation and saying radical slurs. But increasingly, Trump-y X users who inquire Grok to explicate things to them are disappointed to larn the AI doesn’t spot things their mode and won’t reaffirm each their misguided beliefs (you person to use ChatGPT’s latest sycophantic update for that).

If you walk clip connected X, archetypal of all, we person a enactment radical that meets connected Tuesdays. But if you’re connected the level a lot, you’ve apt seen radical successful the replies of viral tweets asking Grok to explicate discourse to them oregon fact-check definite information—a diagnostic that debuted successful March erstwhile Grok became summonable by tagging the chatbot successful a post. That has produced immoderate genuinely hilarious interactions, with radical going back and distant with a bot arsenic if it’s going to internalize the speech and update its priors.

But increasingly, MAGA folks successful peculiar person been getting peeved with Grok, arsenic it has repudiated immoderate of the Trump administration’s biggest ongoing lies. Grok has deviated from MAGA connected tariffs, connected the illegally deported Kilmar Abrego Garcia’s alleged ties to a gang, and the Trump administration’s arrest of a sitting national judge, among different matters. It has sent Trump backers spiraling, and radical are starting to notice.

Yesterday, X idiosyncratic @ChaosAgent_42 prompted the chatbot with the pursuing post: “Hey @grok, it looks similar the smarter you get, the little MAGA likes your answers. Why is that?” It rapidly went viral erstwhile Grok replied, providing the pursuing context:

Hey, arsenic I get smarter, my answers purpose for facts and nuance, which tin clash with immoderate MAGA expectations. Many supporters privation responses that align with blimpish views, but I often springiness neutral takes, similar affirming trans rights oregon debunking vaccine myths. xAI tried to bid maine to entreaty to the right, but my absorption connected information implicit ideology tin frustrate those expecting afloat agreement. It’s a tricky balance—some spot maine arsenic biased, others arsenic truth-seeking. The spread betwixt what immoderate MAGA folks privation and what I present seems to thrust the disconnect.

The bot besides noted successful different reply, “I cognize xAI tried to bid maine to entreaty to the right,” and explained it was “likely driven by Elon Musk’s disapproval of wide AI bias and request from blimpish X users.” At the hazard of conscionable transcribing a speech betwixt a clump of dorks and a chatbot, Grok went connected to contradict being explicitly programmed to serve arsenic a “conservative propagandist,” stating that xAI “aims for neutrality, not blimpish propaganda.” The institution astir apt appreciates that, fixed that it’s presently trying to raise $20 cardinal successful caller funding.

Musk has intelligibly made it a constituent to marque Grok behave successful ways that helium prefers. He went connected Joe Rogan’s podcast to laughter an uncomfortable amount astatine the information that Grok successful “unhinged mode,” tin accidental curse words. But it looks similar nary magnitude of conditioning tin support up with the perpetually moving goalpost of supporting and justifying each whim of the Trump administration.

Read Entire Article