May 31, 2005

Little GTO

Like I suppose every red-blooded male did, I loved anything with wheels when I was a child. I still have a small, now rusting, collection of toy cars that I push around the floor from time to time. My wife suspects I have regressed into childhood; that is, if she can assume I grew out of it at all.

But better men than I remain fascinated with cars too. Example: Valentino Braitenberg of the University of Tuebingen, Germany. In 1984, he wrote Vehicles: Experiments in Synthetic Psychology, a slim collection of thought experiments involving little cars. He uses them to make some interesting observations about human behaviour, much as my wife makes observations about mine.

Braitenberg's imaginary cars all have two things in common: sensors and motor-driven wheels. The sensors respond to light and send signals to the motors. His simplest vehicle, Vehicle 1, has one sensor and one motor driving one wheel. The more light the sensor picks up, the faster the motor turns, propelling Vehicle 1 in whichever direction it happens to be pointing. It speeds up in well-lit areas, slows down in dark ones. In outer space, Vehicle 1 would move in a straight line, varying its speed as it passed sources of light.

But when Vehicle 1 must contend with friction, as it would on Earth, it deviates from a straight line in unpredictable ways. Responding to light and dark, driven by its motor, disturbed by the forces of friction, Vehicle 1 follows a complex path, turning one way and then another for no obvious reason.

Suppose we put Vehicle 1 in a pool of water. To you, it would seem somewhat restless, nervous. It would be frenetic in the day, somnolent at night. In fact, "you would say it is alive," writes Braitenberg, since nothing that's dead moves around quite like that.

This is one of Braitenberg's basic themes: that inferences we make from watching the behaviour of something -- like ascribing life to a vehicle -- are usually more complicated than the structure -- wheel, motor, sensor -- of the thing itself.

Vehicle 2 has two sensors and two wheels, and things get immediately more interesting. In one version, 2a, the left hand sensor connects to the left hand wheel, the right to the right. In 2b, the connections are switched, so that the left hand sensor connects to the right hand wheel and vice versa.

If we place a light bulb directly in front of Vehicle 2a, the little car speeds up and charges straight at the bulb. But if the bulb is even slightly to one side, let's say the right, the sensor on that side is more excited. This drives the right wheel faster, which turns the vehicle to the left: away from the bulb. 2b also speeds up and runs into a bulb placed directly in front of it. But if the bulb is to one side, this car turns toward the bulb, because the wheel on the opposite side turns faster. Eventually, inevitably, it hits the bulb.

The inferences? Both 2a and 2b "dislike" sources of light. But 2a tends to avoid them, turning away and searching for darker areas to escape to. 2a, you might think, is a coward. On the other hand, 2b is not scared of light sources. "It resolutely turns toward them and hits them with high velocity," writes Braitenberg, "as if it wanted to destroy them." Clearly then, 2b is aggressive. Not a coward.

Braitenberg describes ever-more complex vehicles which behave in ways that suggest to us, watching them, characteristics such as: love, a desire to explore, a system of values, knowledge, instincts, an ability to make decisions, a will, and more.

Ever more human characteristics, in fact.

This may make you wonder. Here we are, playing with a few wires, wheels and sensors, and we read these complex meanings into the behaviour of the toys we produce. "These vehicles," Braitenberg writes about Vehicle 4 and its variants, "are governed by instincts of various sorts, and, alas, we just don't know how Nature manages to embody instincts into a piece of brain."

"You forget, of course, that we have ourselves designed these vehicles."

The thought experiments suggested to Braitenberg a "law of uphill analysis and downhill invention." That is, it would be easy to make these little cars and have them perform their range of tricks: easy like going downhill. But it's more difficult -- like going uphill -- to analyse their behaviour and guess at their internal structure.

Analysis, says Braitenberg, is a much more difficult process than invention.

What's more, analysis tends to overestimate complexity. Considering the behaviour of a little car, we ascribe complex characteristics like love or cowardice to it. But those who invented it know that its behaviour is actually driven by simple gadgetry, no more.

That is: the simpler explanation is usually the correct one. An ancient truth, rising out of a thought experiment with cars. Let me push those things around a little more, won't you?

6 comments:

Neela said...

d

am definitely going to pick up that book since am attempting to make my living analyzing and predicting people's behaviour.

n!

Anurag said...

Hey this was quite an interesting read.

I once worked in robotics for a project, and my task was to contribute to progamming to avoid obstacles and find plan a path for vehicles. Sometimes, when two of our vehicles faced each other, they would try to sverve out of each other's way, in the same direction. Then they would notice that it didn't serve the purpose so they would simultaneously move to the other direction. This could go on for quite some time till the vehicles were so close to each other that the program commanded them to stop. This behaviour was not much unlike that of two human beings trying to dodge each other in a gallery, or through a door. Was interesting.

You do realize that the uphill analysis is difficult because there could be a many to one relationship between cause and effect for complex systems. So, when a human being is angry, it could be because of physical hurt, or rejection of love, or hunger, even psychosis. For really complex systems, we have to first completely understand the behavious of components comprising the system, and how they interact (which is more often non-linear than not, adding to trouble) before we can safely arrive at the cause of a response.

Whoa! I think I really like this topic. :))

Fadereu said...

Fascinating post, Dilip. I would love to explore the twilight zone between the inanimate machinistic and the animate, the life-like.

the computer behaves in the image of man because it is programmed, and if the computer is assumed superior, we get the feeling that man is made in the image of the computer, the computer becomes and ancient truth.

Would that be a dangerous assumption? would that make the computer an idol?

I tried calling your cell but it said it's temporarily disconnected. Where are you?

Wanted to share sme good news. I've taken up a small job with DNA.

Anonymous said...

To buy louis vuitton online, Buy this louis vutton bags in shop. louis vuitton handbags for sale, buy louis vuitton Purses online.discount louis vuitton wallets for sale, louis vuitton replica handbags sold well

Anonymous said...

To buy louis vuitton online, Buy this louis vutton bags in shop. louis vuitton handbags for sale, buy louis vuitton Purses online.discount louis vuitton wallets for sale, louis vuitton replica handbags sold well

Anonymous said...

we offer Gucci handbags online sale, Gucci bags for cheap, Gucci wallets on sale, Gucci Purses for sale cheap Gucci handbags for sale, cheap Gucci bags online sale, discount Gucci handbags online store, Gucci replica handbags for sale, cheap Gucci handbags sale on sale, discount Gucci bags sale, buy Gucci bags sale,cheap Gucci replica bags on sale. Gucci handbag online sale, Gucci bag for cheap, Gucci wallet on sale, Gucci Purse for sale cheap Gucci handbag for sale, cheap Gucci bag online sale, discount Gucci handbag online store, Gucci replica handbag for sale, cheap Gucci handbag sale on sale, discount Gucci bag sale, buy Gucci bag sale,cheap Gucci replica bag on sale.