Episode Summary
Executive Summary: The episode debates whether GPT-5 is a breakthrough or a letdown, ultimately framing it as a sign that AI progress is shifting from hype-driven AGI dreams toward practical product-building. The hosts contrast “thinking” models with agentic “doing” systems, discuss Sam Altman and Eric Schmidt’s public skepticism about bubbles and AGI fixation, and note that AI is already reshaping labor in niches like OnlyFans chat operations.
Main Topics: GPT-5 backlash and model disappointment (Priority: 5/5): The hosts argue over whether GPT-5 represents real progress or a step back from o3. One side says GPT-5 under-delivers, feels less capable, and is confusingly routed between fast and thinking modes; the other says the release is useful because it punctures hype and pushes AI toward practical use. Thinking models vs. agentic doing systems (Priority: 5/5): A central debate is whether AI should primarily act as a thought partner or as an agent that calls tools and executes tasks. They argue these are different products with different UX needs, and that OpenAI may be forcing both into one interface. Sam Altman, bubbles, and the AGI hype cycle (Priority: 4/5): Altman’s comments that AI may be in a bubble are treated as a headline-grabbing admission rather than a retreat from AI’s importance. The discussion frames this as evidence that expectations are cooling and that the market may be overexcited about AGI timelines. Eric Schmidt’s case for building with current AI (Priority: 5/5): Schmidt’s New York Times op-ed is used to support a shift away from AGI obsession and toward deploying today’s models in real-world workflows. The hosts highlight his argument that the U.S. should focus on productivity gains and diffusion across industries rather than waiting for a god model. Enterprise AI adoption and the 95% failure problem (Priority: 4/5): The conversation connects GPT-5’s backlash to a broader enterprise reality: most AI pilots are not delivering ROI. The hosts interpret this as a learning and workflow design problem, but also as evidence that the tech remains less capable and predictable than hype suggested. AI automation in OnlyFans chat and content workflows (Priority: 3/5): A final segment examines reports that AI is starting to replace human chatters in OnlyFans operations and may also generate AI images for creators. This is presented as a concrete, near-term example of AI replacing narrowly defined labor tasks.
Key Arguments: GPT-5 feels like a downgrade from o3 because it offers less visible deep thinking and less user control over reasoning depth. The biggest flaw in GPT-5 is product design: it sometimes routes users to a weak model and sometimes to a strong one without clear signaling. OpenAI is trying to combine two distinct use cases—thinking and doing—into one product, but those should likely be separate or clearly routed. Agentic systems are the future because real value comes from AI choosing tools and executing tasks, not just generating text. The release is useful because it punctures unrealistic AGI expectations and pushes the market toward practical building. AI’s biggest near-term value may come from applying current tools across industries, not from waiting for a breakthrough model. The enterprise AI failure rate reflects both poor implementation and the fact that current tools are still too unpredictable for many workflows. Altman’s bubble comments should be read as a market signal that enthusiasm is outpacing current technical reality. Eric Schmidt’s China comparison strengthens the argument that the U.S. should prioritize AI diffusion into the real economy. Niche labor like OnlyFans chatters is especially vulnerable because it involves repetitive, scriptable conversational work that AI can increasingly mimic.
Data Points: GPT-4 to GPT-5 timeline: About 2 years - Used to emphasize that OpenAI spent a long time preparing GPT-5 and built up high expectations. Survey respondents: 475 - Referenced in an Association for the Advancement of Artificial Intelligence survey cited by Eric Schmidt and Selena Xu. Current approaches unlikely to reach breakthrough: More than three quarters - From the AI survey cited in the NYT op-ed, indicating skepticism about current scaling paths. AI pilot programs achieving rapid revenue acceleration: 5% - MIT study cited in the episode saying only a small share of enterprise AI projects produce fast revenue gains. AI projects failing to hit target performance: 95% - MIT study result cited to illustrate the widespread lack of measurable enterprise ROI. Americans who trust AI: 32% - From the China/U.S. comparison mentioned while discussing public perceptions of AI. People in China who trust AI: 72% - Used to show much higher trust and perceived value of AI in China than in the U.S. Adults in China saying AI changed daily life: Over three quarters - A poll cited to argue China is seeing broader practical diffusion of AI. Anthropic valuation mentioned: $170 billion - Referenced during discussion of continued capital inflows and bubble concerns. Anthropic raise mentioned: $10 billion - Mentioned as evidence that AI valuations and fundraising remain extremely large. Prior Anthropic raise report: $5 billion - Used to show how quickly fundraising numbers are escalating.
Pivotal Quotes: "GPT-5 is much better than anything he's used before. It's better than most humans at almost every task." — Sam Altman (as quoted in the transcript): Referenced to contrast Altman’s hype with the hosts’ view that GPT-5 failed to meet expectations. "The issue with GPT-5 in a nutshell is that unless you pay for a model switching and know to use GPT-5 thinking or pro, when you ask GPT-5, you sometimes get the best available AI and sometimes get one of the worst AIs available." — Ethan Mollick: Cited to explain the product inconsistency and why users may experience GPT-5 as both strong and weak depending on routing. "AGI isn't a finish line, it's a process that involves humble gradual uneven diffusion of generations of less powerful AI across society." — Eric Schmidt and Selena Xu: Used as the strongest articulation of the episode’s pro-building, anti-hype position.
Implications: The episode suggests the AI industry is entering a post-hype phase: less AGI talk, more product integration, workflow redesign, and selective automation. For users and companies, the winners may be those who build with current tools now rather than wait for a breakthrough.
About Big Technology Podcast
The Big Technology Podcast takes you behind the scenes in the tech world featuring interviews with plugged-in insiders and outside agitators. Alex Kantrowitz, a Silicon Valley journalist who's interviewed the world's top tech CEOs — from Mark Zuckerberg to Larry Ellison — is the host.