Beating the AI Doom Cycle
The host introduces the 'AI Doom Cycle,' a five-stage emotional and cognitive framework describing how people relate to AI, from skepticism through psychosis and doom desperation to enlightened excitement. The episode uses recent news stories—including Ken Griffin's reversal on AI, college graduation boos, Silicon Valley malaise, and token pricing shifts—to illustrate where different groups currently sit in the cycle. The host argues that reaching 'enlightened excitement' enables more nuanced, productive policy discussions rather than fatalistic narratives.
Summary
The episode centers on a framework the host calls the 'AI Doom Cycle,' loosely inspired by Gartner's technology hype cycle but focused on the emotional and cognitive states people experience in their relationship with AI. The five stages are: skepticism and disbelief, AI psychosis (peak belief in AI's transformative power), doom desperation, real-world recalibration, and enlightened excitement. The host argues that the faster society moves through the earlier stages into enlightened excitement, the better positioned people will be to engage meaningfully with AI's implications.
The first stage, skepticism and disbelief, is described as increasingly on the wane, often driven by outdated priors or AI skepticism as a business model. The second stage, AI psychosis, is illustrated through Citadel CEO Ken Griffin's reversal: having dismissed AI as hype at Davos in January, Griffin now admits AI is completing work that previously required PhDs over weeks, delivering 15-25% productivity gains—yet this realization left him depressed rather than excited, nudging him toward the doom desperation stage.
Doom desperation is the episode's most extensively covered stage. It manifests in Andrew Yang amplifying Griffin's concerns about job destruction, Fortune and WSJ resurfacing predictions from Mustafa Suleiman (18 months to automate all white-collar work) and Dario Amodei (10% overall unemployment, 50% for entry-level white-collar jobs). A viral post from Menlo Ventures' Didi Das captures Silicon Valley's fractured mood: a tiny cohort of ~10,000 people at top AI companies has hit retirement wealth, while everyone else faces layoffs, career uncertainty, and existential dread. College graduation ceremonies have become flashpoints, with Eric Schmidt and an unrelated corporate speaker both booed for mentioning AI, reflecting graduates' anger at being told to celebrate a technology whose architects promise it will destroy their livelihoods.
The host then introduces real-world recalibration as the corrective stage—not necessarily optimistic, but grounded. Meta's layoff of ~8,000 employees (10% of staff) and reported screen-tracking for AI training have tanked internal morale. Simultaneously, a structural shift in AI pricing is forcing companies back into ROI thinking: Anthropic has moved enterprise customers from flat-rate subscriptions to usage-based billing, GitHub Copilot has done the same, and screenshots from users show that actual usage-based costs are 20x to 100x higher than what flat-rate plans were subsidizing. This compute scarcity is structural—driven by shortages in electricity, memory, and chips—and will slow the pace of automation, reintroducing economic friction that doom narratives often ignore.
The final stage, enlightened excitement, is characterized not as naive optimism but as a state enabling more specific, nuanced discourse. The host highlights economist Alex Emas' essay 'What Will Be Scarce?' as an example of productive thinking, arguing the relational economy (where human provenance has economic value) will grow proportionally as AI commoditizes other outputs. He notes that OpenAI and Anthropic launching large consulting operations—training 30,000 PwC professionals on Claude—reflects how real-world institutional inertia demands intensive human effort regardless of lab capabilities. Jensen Huang's Carnegie Mellon commencement speech is cited as an example of an effective counter-narrative focused on generational opportunity and re-industrialization. The host also points to emerging policy ideas—like federally taxing tokens at the provider level (Mark Cuban) or requiring data centers to set aside affordable compute for low-income users (Matthew Iglesias)—as the kind of concrete, actionable discussions that enlightened excitement makes possible, in contrast to the binary fatalism of doom desperation.
Key Insights
- The host argues that Citadel CEO Ken Griffin's shift from calling AI 'all hype' in January to declaring 'for the first time, AI is real' in May—after seeing PhD-level financial research completed by AI in hours—exemplifies how rapid capability improvements are moving skeptics directly into doom desperation rather than excitement.
- The host claims that doom desperation is self-reinforcing because it is primarily being amplified by the very executives building AI, creating a paradox where the architects of the technology are also its loudest doomsayers, which the host says fuels public anger and distrust.
- The host identifies a structural compute scarcity as a real-world constraint that materially slows automation timelines: shortages in electricity, memory, and chips mean the only short-term mechanism to allocate tokens is price increases, forcing companies back into ROI-based thinking rather than unlimited experimentation.
- The host points out that GitHub Copilot's move to usage-based billing reveals that flat-rate AI subscriptions were subsidizing usage at ratios of 20x to over 100x actual cost, with one user's $451 monthly bill equivalent to $11,432 in usage-based pricing—suggesting AI productivity gains were partly illusory when cost-subsidized.
- The host argues that college graduation boos directed at AI mentions are not simply anti-technology sentiment but a rational response to the unusual and historically unprecedented practice of technology builders publicly promising their product will destroy the livelihoods of their audience.
- The host contends that Didi Das' viral post about Silicon Valley malaise—where ~10,000 people hit $20M+ wealth while surrounding tech workers face layoffs and career obsolescence—illustrates that doom desperation is not confined to outsiders but is pervasive even among well-compensated insiders who benchmark against extreme outliers.
- The host argues that Jensen Huang's Carnegie Mellon commencement speech was more effective than doom-framed speeches not because it was more accurate, but because it positioned listeners as agents with power to shape AI's trajectory rather than passive recipients of a predetermined outcome.
- The host claims that the emergence of concrete AI policy proposals—such as Mark Cuban's federal token tax to fund debt reduction and offset AI disruption, or requiring data centers to reserve affordable compute for low-income users—is only possible once society moves past doom desperation into enlightened excitement, where outcomes feel shapeable rather than fixed.
Topics
Full transcript available for MurmurCast members
Sign Up to Access