CONCEPT
Moravec’s Paradox
The observation, earned in hardware and confirmed by every subsequent decade, that the cognitive skills we regard as the crown of human intelligence are computationally cheap for machines, while the sensorimotor competences we share with every mammal are nearly unreachable—that the mountain of intelligence is steeper at the base than the summit.
Moravec’s paradox is the single most durable observation to have emerged from robotics, and its source is a failure. Hans Moravec, watching the Stanford Cart inch across a cluttered room and collide with obstacles it had spent fifteen minutes computing, noticed that the machine could in principle be programmed to play chess while being unable to do what a toddler does without effort—simply see the room. The paradox he named inverts the intuition almost everyone brings to artificial intelligence: the feats we revere as human intellectual achievement turn out to be the cheap ones, and the biological competences we share with every creature that has ever walked or swum turn out to be the expensive ones.
Embodied cognition researchers have confirmed the structural reason—a billion years of evolutionary refinement is buried in the sensorimotor substrate, tuned and inscrutable, while abstract reasoning is a thin and