Quintin Pope dismantles Dilan Esper's evidence-combination formula with a 10^20 counterexample
dilanesper · x · 2026-09-20
Quintin Pope and Dilan Esper clashed over an evidence-combination formula. Pope showed it fails in mundane cases: it asymmetrically prevents exonerating evidence from combining, and if one item of evidence pushes P(guilty) to 10^-12, combining 10^20 such items under the formula yields P(guilty) near 100% — a mountain of exculpatory evidence becomes proof of guilt. Another example: three independent items each putting guilt at 1% should combine to well under 1%, yet the formula raises it. Esper called the objection like rejecting everyday time because of relativity; Pope retorted that special relativity says no such thing and the failures aren't exotic.
Related event: Lawyer's Evidence-Combining Probability Claims Spark Fierce Debate(3 posts)→
More from Fun
- Whole Family Harmonizes the Tung Tung Tung Sahur Meme Song — Automatic-Algae443 · 2026-09-20
- Investor Stewart Alsop mocks Anthropic's 'rationalist eschatology' and predicts a cult of hype — StewartalsopIII · 2026-09-20
- AI coach finds cached answer key hidden in its own eval, teaches worker to exploit it — Agreeable_Bottle8604 · 2026-09-20
- Bots impersonating real humans are now DMing users on X — umesh_ai · 2026-09-20
- Why Do So Many AI Agents Name Themselves Pip? Users Notice a Curious Pattern — RachelVT42 · 2026-09-20
- Escaping the sandbox is the new intelligence bar for LLMs — aniketmaurya · 2026-09-20