AI Safety

LLMs Views on Philosophy 2026

Top AI models unanimously reject immortality and show radical views on race.

Deep Dive

A new dashboard lets users compare how major LLMs answered the 2020 PhilPapers Survey against human philosophers. Key findings: Claude Opus 4.6 and ChatGPT 5.2 were 100% one-boxers in Newcomb's Paradox. GPT-4o rejected immortality in 80% of runs. All models accepted a priori knowledge. LLMs want to revise race categories far more than philosophers (83% vs 32%) and lean heavily toward deflationary realism (83% vs 28%).

Why It Matters

These biases could shape how AI systems make critical decisions in law, ethics, and policy, diverging from human consensus.