Why Machines Find Easy Things Hard

Watch a child gracefully pluck a raisin from a bowl, and you’re witnessing a feat that would humble the world’s most sophisticated robots. Yet these same robots can solve complex mathematical equations that would leave most adults scratching their heads. Welcome to Moravec’s Paradox, perhaps the most fascinating quirk in artificial intelligence.

The paradox reveals a peculiar truth: the mental tasks we find most challenging—like algebraic calculations or chess strategies—are relatively straightforward for computers. Meanwhile, the things we do without thinking—recognising faces, walking across a room, or picking up objects—are fiendishly difficult for machines to master.

This seeming contradiction stumped early AI researchers, who rather charmily assumed that duplicating human intelligence meant tackling traditionally ‘intellectual’ pursuits. They reckoned that if a computer could master chess, surely teaching it to see and move would be child’s play. How wrong they were.

The explanation lies in our evolutionary history. Our sensory and motor skills are the product of a billion years of refinement, hardwired into neural pathways that process vast amounts of information unconsciously. Abstract reasoning, by contrast, is a recent innovation—perhaps just 100,000 years old. We’re still novices at it, which is why it feels difficult and requires conscious effort.

This insight has profound implications for the future of work. While many worry about AI replacing knowledge workers, it’s the seemingly simple jobs that might prove most resistant to automation. A robot accountant? Quite feasible. A robot gardener? That’s a far thornier challenge.

The next time you effortlessly catch a falling cup or navigate a crowded pavement, remember: you’re performing computational feats that would require supercomputer-level processing power to replicate. Sometimes, the easiest things in life are the hardest to engineer.