Robotics & Automation
Why Robots Are Clumsy: The Strange Gap Between Thinking and Touching
A robot that can beat the world champion at chess will still struggle to pick up the chess piece.
The Idea
There's a counterintuitive principle in robotics called Moravec's Paradox, named after the roboticist Hans Moravec who noticed it in the 1980s: the things that feel hard to humans — abstract reasoning, strategy, calculation — are relatively easy to build into machines, while the things that feel effortless — catching a ball, recognising a face, picking up an egg without crushing it — are extraordinarily difficult to automate. The paradox reveals something profound about intelligence itself. What we call 'thinking' is evolutionarily recent; our brains developed it quickly and in a fairly concentrated region. But the sensory and motor skills we take for granted are the product of hundreds of millions of years of biological refinement. They're deeply embedded, massively parallel, and almost entirely unconscious. When you reach out to grab a coffee cup, your brain is solving an insanely complex physics problem in real time — calculating weight, surface friction, the precise grip force needed — without you ever noticing. Building a robot that can do the same requires solving each of those subproblems explicitly, in code and hardware. This is why industrial robots have historically been bolted to factory floors, performing the same motion on the same object in the same spot thousands of times per hour. The moment you introduce variability — different shapes, unexpected surfaces, the chaos of the real world — the whole system breaks down. Modern robotics is essentially a long war against this paradox.
In the World
Boston Dynamics spent years — and a genuinely staggering amount of engineering effort — teaching their humanoid robot Atlas to walk across uneven ground without falling over. The videos look miraculous. But consider what the robot is actually doing: it has to sense the terrain through contact, adjust balance across dozens of joints in fractions of a second, predict where its centre of mass will be several steps ahead, and do all of this while the surface is actively changing. A toddler manages something very similar within about a year of being born. The gap gets even starker in manipulation tasks. Amazon's warehouses, despite being among the most automated logistics environments on Earth, still rely heavily on human hands for the final pick-and-pack stage, because gripping oddly shaped, loosely packed, or deformable objects — a bag of crisps, a coiled cable, a stuffed toy — remains genuinely unsolved at industrial speed and reliability. The company has poured resources into robotic picking systems for years, and the honest answer from engineers inside those projects is that human hands are still, for now, a more efficient solution. The fingers you're using to scroll through this lesson are, by any fair measure, among the most sophisticated tools on the planet.
Why It Matters
Understanding Moravec's Paradox reframes almost every conversation about automation and jobs. The public debate tends to focus on white-collar work — accountants, lawyers, coders — as the obvious targets for displacement, and blue-collar physical work as somehow safer. But the paradox suggests the opposite is often true: a surprisingly large share of physical, dexterous, real-world labour is harder to automate than it looks, while certain cognitive tasks are far more vulnerable than people expect. A truck driver navigating an unpredictable road is doing something deeply difficult for a machine. A paralegal reviewing standardised contracts is, in some ways, doing something far more tractable. This doesn't mean physical workers are safe forever — the research investment flooding into robotics is relentless, and the gap is narrowing — but it should make you sceptical of simple narratives about which humans are replaceable and which aren't. The body is not just a vehicle for the brain. In a very real sense, it is part of the thinking.
A Question to Ponder
Which everyday physical tasks do you perform without a second thought that would require a machine to solve thousands of distinct sub-problems — and what does that tell you about where your own irreplaceable value actually lives?
Get a new one of these every morning.
Start learning with Thinkable