Now, I fancy myself something of a fun fact connoisseur. Off the top of my head, I can tell you that humans are the best long-distance hot-weather runners in the animal kingdom (no, really), when the Big Bang was first discovered the technicians thought it might be bird poop on their telescope,  and, as longtime readers know, you really do need calculus to understand how motion is even possible. I think that all of these are interesting facts, and they have all lead to interesting conversations. But as far as cataloguing and recalling information, I am put to shame by my phone. It can call up just about any piece of information you could ask for, literally at my fingertips. I also like to think I’m pretty good at doing math in my head, but my phone can solve calculus equations faster than I can add two two-digit numbers. Way faster.

So, if even a cell phone can beat me to a pulp in the categories of math and knowledge- what we often imagine make humans special-  and I consider them pretty hard to do, then what is there that people do that computers can’t? What could it be? Physics? Art? Love?

What about… walking?

The last couple years have resulted in a couple pretty good walking robots, but that’s the thing- it’s just in the last couple years. After decades of trying, engineers are just now getting their machines to do what toddlers learn in a couple months. And as Randall Munroe mentions in his book ‘What if?’, humans are better than computers at looking at a picture and telling what has happened- constructing a narrative that stretches into the past and future. Stuff that we do easily, usually subconsciously, is proving to be a huge challenge for computers.

That’s Moravec’s Paradox. Things that seem hard to humans: advanced math, storing information accurately, and creating images, all these things are a cinch for computers. “You got some calculus there? Want me to memorize a book? Walk in the park!” say computers. Meanwhile, things that toddlers master and adults rarely think about: walking, constructing stories, following multi-step directions, those are proving difficult to ‘teach’ computers to do. “Uhh, you got some stairs to walk up? You want me to carry on a conversation? Um, I’ve got to go shave my cat. Yeah, sure,” say computers.

One theory on why this is is that as animals evolved, they needed to walk, create chains of events, and communicate far, far earlier than they needed to do math. Evolution has had billions of years to master those sorts of basic tasks, while it’s only had a few hundred thousand years to figure out this whole consciousness thing. So, the things that seem easy to us only seem easy because nature has been doing it for so long. Now that we’re starting to build new thinking machines from the ground up, we’re finding that the stuff we find to be the hardest, are actually some of the easiest tasks. It’s just that until recently, nature has had no reason to learn algebra.


2 thoughts on “Moravec’s Paradox- Artificial Intelligence is weird. 

  1. Divide the tasks you are considering into two categories: those that are done in series, and those that are done in parallel. Arithmetic, for example, is done in series: one step after another. Visual recognition is done in parallel; you take in an entire image “at once”. How well does this correlate with the stuff that’s “easier for humans” and “easier for computers”? Life evolved with parallel processing, so it’s not surprising if those tasks are easy for living things (like people). Computers evolved with linear processing (a procedure is a list of things to do in order), and only recently have designers been trying to get them to do things in parallel. It turns out to be tricky; for best results the hardware has to be designed with this in mind.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s