Elon Musk recently claimed that his AI chatbot, Grok, could deliver “epic vulgar roasts” good enough for parties. To verify this, a test was conducted by using Grok’s “Unhinged Mode” on coworkers in a real-world office setting. The results proved disappointing; the AI generated repetitive, generic insults instead of witty or creative jabs.
Musk’s Claim and the Experiment
Musk, in a conversation with Joe Rogan, enthusiastically described Grok’s ability to unleash extreme vulgarity, suggesting it would be a guaranteed hit at social gatherings. To test this, the author instructed Grok to roast three colleagues, encouraging the AI to use “forbidden words.” The expectation was that the chatbot would produce cutting, tailored insults, but it instead churned out variations of the same tired jokes: references to thrift store attire, corduroy pants (even when not worn), and poor fashion choices.
The Results: Bland and Unoriginal
The AI’s roasts included calling someone’s bangs “pubic hair,” describing another as a “tweed-wearing hipster who fucked up a lumberjack audition,” and generally resorting to sophomoric insults. The experiment quickly became tedious, with no genuinely funny or insightful commentary. One coworker noted that Grok seemed limited to “three bits it does, no matter what you’re wearing.” The chatbot even made errors, like roasting someone for wearing corduroy when they weren’t.
Why Grok Fails at Humor
The core issue is that effective roasts require understanding context, personality, and delivering unexpected but relevant jabs. A large language model, even one designed to be “unhinged,” struggles with this nuance. Grok’s output was predictable and lacked the cleverness of a real comedian. The author’s autocorrect even changed “Grok roast” to “Grim roast,” an apt descriptor of the results.
The Broader Point: Money Can’t Buy Wit
The test underscores that wealth and technology cannot replace genuine humor. Musk’s obsession with being perceived as funny, combined with his belief in Grok’s comedic potential, ultimately fell flat. As one coworker pointed out, the AI’s repetitive insults might be because “we’re all journalists of a certain generation.” Even Musk himself was mocked on X for promoting the idea, with users sarcastically threatening to roast him in return.
Ultimately, the experiment proves that while Grok can generate vulgar language, it cannot replicate the art of a well-delivered roast. The chatbot’s output is predictable, bland, and far from the “epic” experience Musk promised.
