Moravec's Paradox in AI
Where is the robotics world is going, using the paradox in the real world, building robotics products
Welcome to Infinite Curiosity, a weekly newsletter that explores the intersection of Artificial Intelligence and Startups. Tech enthusiasts across 200 countries have been reading what I write. Subscribe to this newsletter for free to receive it in your inbox every week:
Hello friends,
Have you ever wondered why robots are great at smart tasks like playing chess but have a hard time with simple tasks like climbing stairs? Machines are breaking world records when it comes to mental intelligence. But when it comes to physical intelligence like perception and mobility, they seem to have trouble acquiring those skills. This is the premise of Moravec's Paradox.
Moravec's Paradox isn’t just about why your robot vacuum cleaner keeps bumping into walls. It's a roadmap for researchers to build robots that understand us a little better. And hopefully someday they will even laugh at our terrible jokes!
What is the paradox?
Moravec's Paradox is named after Hans Moravec. He articulated this principle in the 1980s along with Rodney Brooks, Marvin Minsky, and others.
Tasks that a child can easily do (e.g. recognizing their mother in a crowd, jumping from a chair) is like rocket science for robots. On the flip side, they'll beat you at chess without breaking a (digital) sweat.
Humans are evolution's favorite child. Our ancestors honed our senses when it comes to surviving and thriving in any environment. But when it comes to reasoning and solving puzzles? Those came much later. And also they were acquired skills.
On the other hand, robots are logic driven metalheads (quick sidebar: if there were a heavy metal band called Logic Driven, then their fans would be called logic driven metalheads). Robots feel naturally at home when it comes to reasoning and solving puzzles.
Humans can dance, dodge a ball, or spot a friend from afar without even thinking about it. For robots, it’s like trying to understand street parking rules in Montreal. Super complicated!
Old skills like running away from hungry tigers are second nature to us. If a tiger is coming at you, I'm pretty sure every human will know that they should run. No coaching needed here! We've had millenia of practice doing that.
Newer skills like differential calculus or contemplating existential dread require effort.
In fact contemplating existential dread requires so much effort that they awarded the Nobel Prize to Sartre for doing that. Robots started their journey with algebra, so they're just built different.
So how do we make this work?
Robots are great at logic and computation. And humans are great at perception, emotions, and everything else. We can pair a human with a robot to form the dream team! Researchers are working hard on:
Teaching robots how to feel. Robots need to learn how to feel the room (I mean just physically, not emotionally).
Making robots quick learners. Because who has time for a robot that can't adapt?
Broadening their horizons by making them jack of all trades and master of quite a few.
Why should I care about Moravec's Paradox?
Moravec's Paradox gives us an idea of where the robotics world is and where it needs to go. Any AI aiming to outsmart humans needs to wrap its digital "head" around this paradox.
This enables us to direct our efforts towards the right places. And show us where to poke around.
Here's why you should pay attention:
- It will help us guide the next big AI model. Ever wondered why we don't have robotso making our breakfast yet? This pesky paradox explains it. Building robots for real-world action is a massive headache. And it's all because of what Moravec pointed out.
- Moravec's Paradox reminds us to play to our strengths. Robots can compute really well and really fast, while humans are the brains of the operation. Knowing where robots rock and where they flop ensures that we're putting the right folks (and bots) in the right jobs.
- Want a future-proof job? Train in stuff that robots are terrible at (hint: emotional intelligence, interpretive dance, standup comedy).
How can Moravec's Paradox help us build great products?
Here are a few ideas to consider for products that require the merging atoms and bits:
Robocare: Create robots that help old people fetch cookies or assist someone with mobility issues.
Medical bots: Build tools for doctors that can read and analyze medical jargon but also understand when to say "We can let the real doctor handle this".
Industrial robots: Build robots that can lift heavy stuff without turning the factory into that scene from The Office when Michael Scott handles the forklift.
Learning tools: Build learning tools that adjust to users but remember that a human teacher's gut feeling trumps any algorithm.
Self-driving cars: I want my car to avoid hitting other cars and also understand that humans jaywalk sometimes.
Moravec's Paradox reminds us that defining intelligence only as capacity for mental computation misses a huge part of the reality. By focusing on real human needs and quirks, we get products that are not only smart but also genuinely useful in our everyday life.
If you are getting value from this newsletter, consider subscribing for free and sharing it with 1 friend who’s curious about AI: