Tuesday, March 4, 2008

A moment to breath..and recommend

For those who take a fascination in humanity's attempts to understand itself through computers, this provides and interesting read. Apparently, Netflix established a one million dollar prize to anyone who could improve upon its movie recommender program by 10%.

What's interesting is that amongst the group funded by giants like AT&T one of the top ten contenders is an unemployed psychologist working out of his home. Take note that the direction of his solution draws from behavioral economics.

Also, I went to a fantastic lecture on optimization. Essentially, finding the best solution from a set of possibilities as judged by an explicit criterion (or set of criteria). The speaker began by discussing the emergence of optimization during the Renaissance. Then it was conceived of as being a principle of nature. God created the world, the world is perfect and so its physical operations are determined by the principle of optimization (more explicitly, by the idea that nature is structured so as to follow and produce the best of all possible outcomes).

An interesting illustration was with theories of light and its movement. Initially, it was thought that light always traveled along the shortest path and this is true for the case of reflection (this can also be understood as saying that the incoming angle with respect to the reflecting surface equals the reflecting angle). However, it is NOT true for refraction.

But, Fermat showed (and this wasn't experimentally verified until 200 years after him) that light maximizes SPEED, not distance in the case of refraction. How does light KNOW to act in such a way (to optimize something)?!?!

Well, it doesn't (neither case turns out to be true in the situation of light reflected in a closed circle). And the transition from thinking of optimization as a condition and definition of nature into thinking of it as an analytic tool follows the rise of the Enlightenment. However, it still leaves us in the interesting position of assuming optimizing behavior not only in individuals (which turns out to not be so problematic) but also among groups. Which turns out to be VERY problematic.

I'll discuss why in the future.

No comments: