How does this relate to Minds and Machines?

After completing this project, we were asked to read select chapters from Valentino Braitenberg's book, Vehicles: Experiments in Synthetic Psychology and then reflect on how his ideas related to our robot, humans, and other systems. Here is what we thought:

When we try to analyze a system, be it alive or otherwise, we often create complicated explanations for what is happening in the system. However, according to Valentino Braitenberg, we may simply be falling prey to the “law of uphill analysis and downhill invention.”

Braitenberg’s book Vehicles follows the creation of numerous simple vehicles consisting of a multitude of sensors and some number of motors. After these vehicles have been created, their behaviors are interpreted by the designer and by an outside observer. The designer easily sees the simple positive of negative influence of the sensors on the motors as the vehicle responds to different environments. The outside observer, however, is blind to the internal workings and thus describes the vehicles as “alive,” “aggressive,” having “values” and so forth, applying a much more complicated internal system to these vehicles to try and explain their behavior. This is what Braitenberg refers to as the “law of uphill analysis and downhill invention,” that the deductive reasoning of the inventor creates a much simpler (and correct) system than the inductive reasoning of the outside observer. It is almost impossible in theory to determine exactly what the hidden mechanism is without opening up the box, since there are many different mechanisms that operate with identical behavior. The analysis itself is also much more difficult that the invention since induction takes much more time to perform than induction: in induction one has to search for the way, whereas in deduction on follows a straightforward path.

This would apply to our robot if someone were to see the Wumpus World and observe the robot’s behavior. An outside observer would conclude that the robot feared the pits and the Wumpus, as it would avoid them, while it valued the gold, as it would try to reach it. This observer would also conclude that the robot would sometimes become lost or commit suicide. In reality, however, the robot is simply marking locations as possible places for pits or Wumpuses and avoiding them thanks to arbitrary programming saying “avoid it.” It seeks the gold only because its program tells it to “go get it.” Our robot became lost only if it overshot the center of the square enough to put its light sensor on a line. It would appear to commit suicide when our robot moved to a square that it thought existed, when in reality, that square was non-existent or off the board. As it was just explained, the observer’s interpretation is much more complex than the creator’s, as Braitenberg’s law predicts.

In addition to Braitenberg’s law, it can be inferred that many or all of complicated cognitive activities performed by humans are wrongly thought to be inexplicable in terms of the workings of simple, uncomplicated components. In other words, if you had infinite knowledge about the system, one would be able to break it down into very primitive and simple parts that would work in unison to create the system that one observes. For example, the brain is little more than a bunch of neurons working together, a muscle is nothing more than a bunch of fibers working together, and a human is nothing more than a bunch of organs and muscles working together. One can suspect Braitenberg’s idea is applicable to all systems, granted one breaks the system down into the smallest possible components.

Despite this, one can still point out several objections by examining the human mind. The first objection is that of consciousness—where would it come from if the brain is a multitude of neurons working together by emitting electrical impulses that each interpret in a different way? Clearly, there must be something more that allows humans to have the ability to make sense of their own existence and be aware. There is also the objection that Braitenberg appears to be judging the vehicles’ “intentions” and “will,” in essence their intelligence, by the vehicles’ behaviors. Intelligent behavior is a poor criterion for intelligence, due to the fact that one can be intelligent and not act intelligent or vice-versa.

Braitenberg may be able to counter the first objection, explaining that consciousness is a whole made up of the individual parts. He could make the argument that as individual cells, neurons cannot “think” or perform any behavior that would be resemble a thought process, but when billions are working together to allow a person to function, these vast connections allow the human mind to create conscious thoughts. For the remainder of the objections however, Braitenberg’s entire analysis of is based upon the behavior of these vehicles, which makes any objection against behaviorism impossible to refute. He may be correct that all “cognitive” activities may be described by the workings of simple, uncomplicated components when dealing with small machines as those can easily be proven by disassembling the machines. Proving this proposition on the scale of a human being and possible future robots would be impossible if the parts were small and numerous enough. In these situations, we are the outside observer trying to induce the criteria for the various exhibited behaviors. No matter how much data we can collect about the observations of these systems, we may never be able to fully describe how they function without an insider’s look on how they were created. Without knowing how their behaviors are governed, we cannot come to a conclusion on whether or not they are intelligent or just exhibiting intelligent behavior.

Braitenberg’s theory may successfully describe primitive, simple machines similar to the ones we have today, but it may fail when applied to future machines that may be able to “think” for themselves. While we are unable to fully understand how these systems operate, we are unable to make any concrete conclusions of how these complex systems truly work.