General-purpose robots
Machines beat us at chess but can't fold laundry. Why is the physical world so hard for robots?
What makes this fascinating
Great at chess, bad at laundry — Machines beat us at abstract tasks but fumble the messy physical world — Moravec's paradox.
Manipulation is hard — Grasping unfamiliar objects and coping with an unpredictable world is far harder than it looks.
Learning bodies — AI is starting to give robots general manipulation skills, but reliable household help is still distant.
Frequently asked questions
- Why don't we have general-purpose robots yet?
- Machines beat us at chess but struggle to fold laundry. The physical world is unstructured and unpredictable, and giving robots the dexterity, perception, and adaptability of a person is far harder than it looks.
- What is Moravec's paradox?
- The observation that tasks easy for humans — perception, walking, manipulation — are extremely hard for machines, while tasks we find hard, like complex calculation, are easy for them.
- Are AI advances bringing general-purpose robots closer?
- Modern machine learning has improved robotic perception and control, and humanoid efforts are accelerating, but robust, general dexterity in the messy real world remains an open challenge.
More summits in Engineering & Energy
Practical nuclear fusion power
The reaction that powers the stars, in a box on Earth — clean, limitless energy, if we can tame it.
Grid-scale energy storage
Sun and wind are free but fickle. Storing their energy cheaply at scale would remake civilization.
Scalable carbon capture
Can we pull CO₂ back out of the air fast and cheaply enough to matter?
Cheap fresh water at scale
Two-thirds of the planet is water we can't drink. Can we make the ocean drinkable for everyone?
Ready to climb?
Learn it the whole way up — from the fundamentals to the frontier.