Moravec's Paradox: AI Reasoning vs Sensorimotor Skills

[Constantinos]

Hey!

Browsing the web, I came across a pretty interesting idea(at least for me) which I haven't seen people in this forum discuss.

http://en.wikipedia.org/wiki/Moravec's_paradox

In just says that high level human cognition is a much easier problem for AI researchers to solve compared to sensorimotor skills. This is because those skills where evolving for far longer than human reasoning. I'm not sure it includes common sense though.

Maybe you'll find it interesting too!

 

[256bits]

Complex activities require complex computations, so what else is new. Anything that involves movement is of course going to require computational power

Say to catch a ball, one has to determine or estimate trajectory, speed, mass etc. and these are all done unconsciously without thinking. I don't buy into the evolutionary explanation of why sensory motor activities are hard for a computer to do - it is hard because of the rules and precision needed in computations. If you are 'out' by 1% in the speed for example you or the computer is going to miss catching the ball. The difficult part is the systems analysis of a situation and once that is refined enough the computations fall in place. The other aspect is that if the computer is too slow in doing the computations than the event will have passed by before the computer gets its answer.

Ever wonder how a hitter can connect with a baseball coming at him 90 miles per hours from 90 feet away? The human brain pretty much has to determine whether to swing or not at the ball as the ball has just left the pitchers hand, and then make minor corrections in the swing before and as the ball is coming across the plate. I have not heard of a perfect computerized system being made that can connect with the ball as many times as hitters actually do - their computation is just too slow.

 

[SW VandeCarr]

Ever wonder how a hitter can connect with a baseball coming at him 90 miles per hours from 90 feet away? The human brain pretty much has to determine whether to swing or not at the ball as the ball has just left the pitchers hand, and then make minor corrections in the swing before and as the ball is coming across the plate. I have not heard of a perfect computerized system being made that can connect with the ball as many times as hitters actually do - their computation is just too slow.

Hitting a pitched ball (at the professional level) is the hardest thing in sports according to this source which agrees the with the generally accepted view. The batter has 0.48 seconds from the time the ball leaves the pitcher's hand to the time it plops into the catcher's mitt at 90 mph.The distance is only 63.6 feet (actually less since the pitcher's release point is usually a few feet in front of the rubber).

http://www.usatoday.com/sports/ten-hardest-splash.htm