Hacker News new | past | comments | ask | show | jobs | submit login
Moravec's paradox (wikipedia.org)
30 points by notadog on April 21, 2018 | hide | past | favorite | 6 comments

>Moravec's paradox is the discovery by artificial intelligence and robotics researchers that, contrary to traditional assumptions, high-level reasoning requires very little computation, but low-level sensorimotor skills require enormous computational resources. The principle was articulated by Hans Moravec, Rodney Brooks, Marvin Minsky and others in the 1980s

Yeah, so where's this "high level reasoning", 30+ years on?

I was going to say the same thing. Programming should be just high level reasoning but we’ve yet to automate that.

It is premature to assume that AI has already demonstrated that the full scope of high-level reasoning is simpler to compute than sensorimotor skills, even if the latter are proving to be harder than expected.

True, but I think what they are getting at here is that sensorimotor skills are not simple to compute.

What a great observation! It reminds me of data-driven programming, espoused by Rob Pike, Linus Torvalds, Fred Brooks, and Eric Raymond, https://www.lysator.liu.se/c/pikestyle.html. Basically to me it means that you move complexity from code to your data. Let your code be small and simple, and your data structures that run through it be --- I don't know if I would say complex but well thought out. It's only tangential to this, but so helpful to me.

More on topic, it also reminds me of the little study I have done of the human brain. "High-level reasoning" and "low-level sensorimotor skills" take place in two different parts of the brain. High-level reasoning happens in the cerebrum, the big gray main part of the brain. Motor skills take place in the cerebellum. That's why when you first learn to type, it's slow, because your cerebrum is handling most of it, you have to "think about" where each key is. After practice, you can find the keys without "thinking about it," --- that is, your cerebellum is handling it. The cerebrum is trained by teaching, but the cerebellum is trained by practice: typing, tennis, basketball, driving, etc.

Could the artificial intelligence field make progress by studying the cerebellum more?

xkcd 1425 illustrates this paradox very well


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact