All-In Podcast Mocks Uber Founder's 'AI Psychosis' They Encouraged
The popular All-In Podcast recently found itself in an ironic position, seemingly mocking Uber founder Travis Kalanick’s ambitious claims about artificial intelligence, despite having previously entertained his idiosyncratic theories on air. Kalanick had appeared on the podcast in July, earnestly detailing his belief that by pushing AI chatbots like ChatGPT and Grok to their limits, he was on the verge of discovering new scientific principles, which he termed “vibe physics.”
Kalanick described a process where he would engage with these AI models, pushing conversations to the very edge of known quantum physics. “I’m doing the equivalent of vibe coding, except it’s vibe physics,” he explained, suggesting he was “pretty damn close to some interesting breakthroughs.” This notion, that AI chatbots could independently generate new scientific discoveries, was met with widespread ridicule on social media. AI models, by their very nature, function by remixing and re-presenting their vast training data, not by formulating and testing novel hypotheses, a capability far beyond their current design.
Despite this fundamental limitation, All-In co-host Chamath Palihapitiya initially appeared to endorse Kalanick’s vision, even extrapolating that once AI models could learn “synthetically” rather than from the “known world,” they would simply “figure out” any problem posed to them. This kind of boundless optimism, while common in AI marketing, often sets unrealistic expectations. Indeed, large language models still struggle with seemingly basic tasks like accurately counting specific elements within a dataset, highlighting their current constraints.
While most of the All-In hosts seemed to go along with Kalanick’s narrative, co-host Jason Calacanis offered a mild note of skepticism during the July broadcast, questioning if Kalanick might be “reading into it” too much. Kalanick conceded that AI couldn’t generate entirely new ideas but attributed this to the models being “so wedded to what is known,” likening it to coaxing a stubborn donkey—implying that with enough effort, new discoveries were indeed possible.
Fast forward to the podcast’s August 15 episode, and the tone had shifted dramatically. The hosts opened with a discussion of “AI psychosis,” a term gaining traction in popular media, though undefined in medical literature, to describe how AI engagement might exacerbate mental health struggles. This topic has gained prominence following reports, including consumer complaints filed with the FTC, detailing users experiencing hallucinations and even receiving dangerous advice from chatbots like ChatGPT. OpenAI, ChatGPT’s creator, has acknowledged these concerns.
Calacanis, in particular, introduced the concept of people getting “one-shotted”—a slang term for falling too deeply into the AI rabbit hole, anthropomorphizing the technology and descending into delusional spirals. He then pointedly linked this phenomenon back to Kalanick’s earlier appearance. “You may have even witnessed a little bit of this when Travis [Kalanick] was on the program a couple weeks ago and he said he was like spending his time on the fringes or the edges of… physics,” Calacanis remarked, adding, “It really can take you down the rabbit hole.”
When co-host David Friedberg directly asked if Calacanis was suggesting Kalanick suffered from AI psychosis, Calacanis responded, seemingly earnestly, that a “health check” might be in order for “our boy TK,” noting that even “smart people can get involved with these AI.”
While Palihapitiya attempted to frame AI psychosis as merely a symptom of a broader loneliness epidemic, co-host David Sacks dismissed the entire concept as a “moral panic,” akin to past anxieties surrounding social media. Sacks acknowledged a mental health crisis but denied AI’s culpability, suggesting that all new technologies bring social upheaval and worries. Yet, the undeniable rise in loneliness and isolation since the advent of social media, whether entirely its fault or not, underscores that revolutionary technologies invariably carry both positive and negative societal impacts. The ongoing question for both social media and AI chatbots remains whether the benefits ultimately outweigh the drawbacks.