Machine Learning

When I went to UT Austin for a college visit, they had one of their best professors give us a lecture on Machine Learning, specifically in soccer playing robots. His lab had developed a method of getting the robots to teach themselves how to run the fastest, or kick the ball the furthest. It was all based off a simple process. First, one needed to define the parameters for success. For some algorithms, this was just as simple as a for loop looping through a parameter, and testing the program each time. When the parameter was finally maximized, we have finally allowed the machine to teach itself a behavior. This is one of the best ways of programming, because the program will always run optimally, given enough time.

After finding out about this method of programming, I immediately began looking for ways to use it in my life. For me the most obvious use came the next day, when I was playing a game with some of my friends. The idea was to get to a certain Wikipedia page in the fewest number of clicks. An example of getting from Pair Programming to Marie Antoinette would be Pair Programming > Programmer > Offshore Outsourcing > China > List of Countries and Dependencies by Area > France > Louis XVI > Marie Antoinette. Of course, there are likely many many many other networks that lead from Pair Programming to Marie Antoinette. The challenge is to find the shortest one. And this task seems far more suited to a computer than a human. A computer can iterate through thousands and thousands of links in seconds, searching for answers. It would be ideal to use one to solve the Wikipedia game. For two specific articles, it might take the computer minutes to find the optimal solution, learning through trial and error which pages it should take.