Moravec's Paradox has never been empirically tested, and its predictive claims rest on selection bias

Moravec's paradox—the widely circulated claim that computers excel at tasks humans find difficult while struggling with tasks humans find simple—lacks empirical validation despite frequent citation by prominent AI researchers. The originating statement, from Hans Moravec's 1988 book Mind Children, observes that while chess-playing has reached superhuman performance, physical tasks like robotic manipulation remain primitive.

A systematic test of the paradox would require categorizing a representative sample of tasks by difficulty for humans and for AI systems, then plotting the relationship. Instead, Narayanan argues, what appears to be a paradox reflects a selection effect in which AI researchers focus attention on specific quadrants of a 2×2 task space. Tasks that are easy for both humans and AI (simple arithmetic, tic-tac-toe) generate no research attention. Tasks that are hard for both (predicting stock prices, deciphering ancient scripts) receive minimal serious investigation. This selective focus leaves visible only two categories: tasks hard for humans but easy for AI (web search, data retrieval) and tasks easy for humans but currently hard for AI (physical manipulation, complex reasoning in novel domains). When two quadrants are effectively ignored as uninteresting, apparent negative correlation between human and machine difficulty becomes inevitable.

The paradox carries an evolutionary origin story: the human brain encodes billions of years of sensorimotor learning that makes perception and motor control appear effortless, while reasoning represents a thin cognitive veneer. Narayanan identifies this explanation as speculative, grounded neither in modern neuroscience nor evolutionary biology. AI researchers drawing on such claims typically lack background in these fields yet assert them as fact.

Rather than rely on such predictions, Narayanan suggests that the diffusion timeline for new AI capabilities—the lag between technical breakthrough and widespread deployment—provides the actual window for policy and institutional response. Panic tends to follow capability breakthroughs precisely because prediction frameworks like Moravec's paradox provide false confidence, not because prediction itself is impossible.