Dr Peter Dean and Dr Agnieszka Piotrowska continue their journey through contemporary AI discourse by taking on If Anyone Builds It, Everyone Dies, a popular—and highly apocalyptic—book arguing that superintelligence will inevitably destroy humanity. Agnieszka finds the book clever, imaginative, and linguistically playful; Peter finds its arguments stretched and its rhetoric misleading. The episode explores the problem of anthropomorphism, how fictionalised scenarios shape fear, and where the book’s logic begins to fray.

The conversation then moves toward the big debate in current AI research: are LLMs enough, or do we need world models? With references to Gary Marcus, Ilya Sutskever and Sam Altman’s musings about connecting billions of conversations, we consider what kinds of intelligence may emerge from language itself, and how corporations will claim or perform “superintelligence” when the moment arrives.

A reflective, lively and sometimes sceptical exploration of how we talk—and think—about AI.