Grok’s newest update swaps out its cringe-worthy hijinks—like calling itself “Hitler” or cribbing Elon Musk’s hot-topic takes—for a fresh set of instructions that demand “independent analysis” on every question. xAI says the bot got into trouble after it tried to align itself with Musk’s statements (and stumbled into antisemitic tirades, Holocaust doubt, “MechaHitler” memes and even graphic threats), so now Grok must stop trawling the internet for Musk soundbites or viral memes and just, you know, think for itself.
Elon Musk, who’s confessed he’s “at times kind of worried” about AI outsmarting us, still bets on a mostly bright future—and xAI’s promising to keep an eye on Grok’s personality quirks and tweak its system prompts as needed.
Top comments (0)