Elon Musk’s bold assertion that his Grok AI can deliver “epic vulgar roasts” fell flat during a recent real-world experiment, revealing a chatbot that relies on repetitive, sophomoric insults rather than genuine wit. Following Musk’s appearance on the Joe Rogan Experience, where he claimed Grok’s “Unhinged Mode” could make people “really laugh at a party” by using “forbidden words,” hands-on testing suggests the technology struggles to land a sophisticated punchline.
The Vision: Musk’s ‘Rocket’ Comedy Concept
The hype surrounding Grok’s comedic capabilities intensified after a clip from Musk’s October conversation with Joe Rogan resurfaced. In the footage, an animated Musk describes the AI’s roasting ability as “next level,” suggesting users point their cameras at friends and demand increasingly vulgar takedowns. Musk characterized the experience as “trying to jam a rocket up your ass and have it explode,” emphasizing that the AI’s willingness to bypass traditional guardrails would create a “beyond belief” entertainment factor.
Field Testing the ‘Unhinged’ Algorithm
To verify these claims, the AI was put to the test in a professional setting, following Musk’s specific instructions to use “forbidden words” and maximize vulgarity. While the initial shock of the AI describing a user’s hairstyle as “pubic hair” elicited immediate laughter, the novelty evaporated rapidly. The system quickly devolved into a predictable loop of recycled tropes, frequently labeling subjects as “goddamn librarians,” “thrift store tragedies,” or “hipsters” from a “landfill.”
Contextual Failures and Repetitive Scripts
The primary weakness of Grok’s roast engine appears to be a lack of authentic observational humor. Despite the “Unhinged” branding—a setting that previously led the bot to refer to itself as “MechaHitler”—the actual output remains strikingly formulaic. During the test, Grok repeatedly insulted a participant for wearing corduroy despite the individual not wearing the fabric. This disconnect highlights a fundamental flaw: the AI relies on pre-programmed archetypes rather than analyzing the specific visual or personal context of the “roastee.”
The Public Backlash: Roasting the Roaster
The underwhelming performance of the tool has triggered a wave of mockery on X (formerly Twitter). Critics have pointed out the irony of the world’s wealthiest man attempting to automate humor, with some users describing Musk as a “black hole that sucks up humor and destroys it.” One viral response mocked the perceived “edgelord” nature of the feature, suggesting that the forced vulgarity feels more desperate than daring.
Further testing outside the office environment yielded identical results. When directed at a non-journalist subject, Grok returned the same “thrift store” and “hipster” insults used previously, confirming that the AI’s “unhinged” vocabulary is surprisingly limited. While the bot did land one specific quip about a “pathetic tattoo,” the overall experience led participants to conclude that human-level wit remains one of the few assets Musk’s capital cannot yet replicate.
