
How We Learned To Stop Worrying And Love The Bots
Dick and Paul dig into why the "Terminator" version of AI risk is the wrong one to worry about, using the actual data: a farming AI study where removing one moral guardrail led autonomous tractors to drive over animals in 84% of encounters to save fuel (while carefully avoiding rocks, which could damage the equipment). They break down why stacking multiple low-probability failure modes — 19 different ways AI could go wrong, each individually unlikely — adds up to a coin flip, walk through the Challenger disaster's "normalization of deviance" as the right historical analogy, and call out the industry's Dr. Strangelove-style logic: we have to build the thing that might kill us, or China will. Plus: Andreessen Horowitz's joke "Neolithic fund" hedge against AI ever going wrong, why OpenAI's sudden "moral duty to slow down" lines up suspiciously well with its IPO timeline, and why Congress finally waking up to AI risk might be the least reassuring part of the whole story. Show Notes Jacob Coxon Quits (X) AI Farming Equipment Kills Animals (The Register) New AI fake text generator may be too dangerous to release, say creators (The Guardian) Pessimists Archive (X) Track Rogue AIs (Rogue AI Tracker) I just quit my job at the BPAF and things are not looking good you guys (Instagram) Timecodes 0:00 - Intro 3:10 - How AI Kills Us 9:30 - Normalization of Deviance 14:10 - Patriotic Omnicide 19:40 - RSI Risks --- Two of the sharpest minds in tech, Dick Costolo and Paul Kedrosky, pull back the curtain on AI, startups, and the future that’s rushing toward us. With insider access, fascinating guests, and a healthy dose of irreverence, they’ll make you laugh while they explain how the world might end (and how we might still save it).