Testing 4 Top AI Models on 'Does God Exist?': Huge Differences

Ibz04 · reddit · 2026-08-05

A developer ran an interesting baseline model experiment, asking four different AI models the exact same highly subjective question: "Does God exist?" while instructing them to be opinionated.

The results revealed distinct differences: Qwen 3 and GPT 5.5 leaned towards 'yes' but remained cautious or general; Gemini 3.6 Flash gave a highly confident 'yes'; while Claude leaned towards 'no'. The author uses this to explore how much of an AI's response comes from genuine reasoning versus underlying training data and alignment.

Original post →

More from Fun

Fun channel →