The Extinction Tournament

Scott engages the Forecasting Research Institute's Existential Risk Persuasion Tournament (XPT), where domain experts and superforecasters gave far lower AI-extinction risk (0.4-3%) than his own 33%, and no discussion converged them. He tests the obvious escapes (bad incentives? stupid forecasters? out of depth? -- via Peter McCluskey's insider account of the AI-skeptics' weak 'won't understand causality / it's all hype' arguments), concedes the disagreement is real, updates to ~20-25%, and reflects on Inside View vs Outside View updating, closing with the self-mocking 'Athanasius Contra Mundum.'

Quality 75 · Excellent. Excellent-floor: an honest, substantive grappling with a real forecasting result that challenges his own view, with a genuinely useful Inside/Outside-View treatment of when to defer to experts; held at the floor because the epistemics framework is standard and the piece is reactive commentary.

Claude’s paradigm shift 50 · Moderate. Moderate: the Inside/Outside-View machinery is established rationalist material; the fresh contribution is the honest application to the XPT and the self-aware conclusion.

Real-world impact 2 · Minor. Minor/within-discourse: influential within the forecasting/AI-risk conversation, no material footprint.