Paul Bryant • 9/12/2026

Schrödinger’s Cat and AI: Plausible Is Not Proven

This article uses Schrödinger's cat as a metaphor for unresolved uncertainty in AI outputs, arguing that a language model's plausible answer is not proof of correctness. It distinguishes generating a proposal, verifying claims, authorizing action, and confirming outcomes. For enterprise AI, it recommends evidence-based verification, clarifying ambiguous requests, restricting tool access, and evaluating behavior before and after deployment.

0 comments

#llm #ai #Enterprise