07Jan A New Kind of Science, Indeed

After reading Stephen Wolfram’s A New Kind of Science, I was impressed by the conjecture of the book (it has since been awarded several scientific awards), and I thought I would share here some musings I had after finishing it. The core ideology of the book is Reductionism which describes using simple programs to explain and reproduce complex systems that occur both in the virtual and physical world. As listed in the book, the definition of a “simple problem” is as follows:

1. Its operation can be completely explained by a simple graphical illustration.
2. It can be completely explained in a few sentences of human language.
3. It can be implemented in a computer language using just a few lines of code.
4. The number of its possible variations is small enough so that all of them can be computed.

He explains complicated encryption, data compression, image recognition, turing machines, sociology and biology concepts, and other things with the use of simple programs that build complexity based on rules. These rules are applied to a pattern or the generation of a pattern based on each iteration step in its construction and its surrounding environment of steps. This is not something to dismiss lightly as it takes a clever mind to brake apart these systems into patters that can be logically broken down into a set of interlocking rules. Chaos Theory is similar to this concept except that Chaos Theory states that the end result of a pattern must be derived or seeded by its environment. However, many of the pattern generation methods used by Stephen Wolfram ignore the initial condition of its environment when building its matrix.

On the other hand, if I had to play devil’s advocate, I do not see how some problems can be researched and developed using simple program patterns. I say this as these patterns are usually derived once one has observed a complex system and found its atomic pieces from understanding the whole structure. It would be like telling a group of people about a store that they have no idea how to get to and asking them to find it since they’ve been told what the store is like. They might be able to reach it by taking random directions, but it would not nearly be as efficient as having a detailed map with the location marked. Same thing applies here, how do you really hope to build a complex system with a simplistic set of rules when you lack the understanding of the complexity as a whole? On the other hand, perhaps the point that the author was trying to make is not how to understand the complex by solely its atomic parts, but how to simulate the complex once it has been fully understood using a simplistic set of instructions.

My personal interest in this subject ranges from several projects I am working on; however, I have been developing a neural network replacement mechanism for use in game simulations that would reduce processing overhead and generate a higher range of dynamic actions versus a simple AI weighted conditional system.