Moravec’s Paradox
Why AI Struggles With the Easy Stuff
We live in an era where artificial intelligence can write poetry, beat chess grandmasters, and predict stock market trends. And yet, that same AI might trip over a curb, struggle to recognize a blurry face in a photo, or completely fail at folding a towel.
Welcome to Moravec’s Paradox. It’s the counterintuitive observation that tasks humans find hard are often easy for AI, while tasks we find easy are hard for machines.
What Is Moravec’s Paradox?
Coined by robotics researcher Hans Moravec in the 1980s, the paradox goes like this:
“It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”
In other words, AI can do calculus and beat you at strategy games, but it still can’t load a dishwasher or walk through a busy farmer’s market without bumping into someone.
Why Is This So?
To understand Moravec’s Paradox, we have to turn to evolution.
Humans evolved sensorimotor skills, like vision, movement, spatial awareness, over hundreds of millions of years. These functions are deeply embedded in our brains and bodies. We don’t think about catching a ball or recognizing a face; we just do it. It's fast, intuitive, and subconscious.
By contrast, formal logic, mathematics, and abstract reasoning are newcomers in our evolutionary timeline. Written language has only been around for about 5,000 years. Most of humanity’s cognitive energy has been spent navigating complex physical and social environments. Not solving equations.
And that’s the paradox: what feels easy to us is actually incredibly complex. And what feels hard, like formal logic and symbolic thinking, is easier to model computationally.
Three Key Domains Where AI Struggles
1. Perception and Navigation
You can walk into a noisy café, recognize your friend in the back, and avoid the waiter carrying drinks. All without thinking.
AI? Not so much.
Even advanced robots and autonomous vehicles struggle with perception under real-world conditions. Shadows, poor lighting, unpredictable obstacles are things that throw machines off. Computer vision has come a long way, but edge cases are still a major challenge.
2. Motor Skills and Dexterity
A toddler can stack blocks, open a snack wrapper, or tie their shoes. These actions require intricate coordination of muscles, feedback loops, and environmental awareness.
Most robots today are either rigidly repetitive (great at factory tasks) or painfully clumsy in general environments. Folding laundry? Still a monumental challenge.
3. Social Interaction and Empathy
ChatGPT and similar models can generate human-like conversation, but they don’t understand emotion. They don’t feel discomfort, confusion, excitement, or grief.
Humans constantly read body language, vocal tone, context, and cultural nuance. Machines, on the other hand, rely on pattern prediction, not emotional intelligence.
What Does This Mean for AI Development?
Moravec’s Paradox reshapes how we should think about AI, not as a human replacement, but as a complement to human strengths.
Here are three major implications:
1. Specialization Over Generalization
Rather than building AI to mimic humans across all domains, we should build systems that excel in narrow, well-defined tasks: data analysis, pattern recognition, optimization. These are areas where machines already outperform us.
2. Human-AI Collaboration Is Key
The real power of AI comes when we pair it with human intuition. A doctor aided by AI diagnostics is more accurate. A pilot with automated systems is safer. A writer with AI drafting tools is more efficient. Collaboration > automation.
3. Design Human-Centric AI
Instead of forcing AI into roles that demand emotional intelligence or physical grace, we should design systems that support the humans doing those jobs. Think: copilots, assistants, advisors, etc.
Intelligence Isn’t Just Logic
Moravec’s Paradox reminds us that intelligence isn’t just about reasoning. It’s about being in the world. About moving through space, reacting in real time, reading people, adapting to change.
Machines are getting better at thinking like us. But being like us? That’s still a long way off.
And maybe that’s a good thing.
Because in a world increasingly filled with powerful machines, it’s our very human qualities, like intuition, empathy, perception, that might matter most.
Watch on YouTube, or listen on Spotify.
