'Moses' ark' trick question still catches AI hallucinations

The classic 'Moses Illusion' trick question resurfaced on X, fooling even AI safety researcher David Manheim, and remains a key test of whether language models truly understand semantics rather than hallucinating.

2026-10-13 ~ 2026-10-13 · 2 related posts

• The Moses illusion: the classic trick question still testing big-model hallucinations — dioscuri · 2026-10-13
• AI safety researcher falls for the classic Moses illusion test on X — davidmanheim · 2026-10-13