Saturday, 26 September 2009

Computer Program Self-Discovers Laws of Physics

While I have not given sufficient thinking on the implication of "The Petabyte Age" to have proper comment, I want to point out my underlying stance for scientific theory.

Science focuses on repeatable observable events. We throw a stone up, it falls back down. Do it again, it falls again. So, we try to *understand* a collection of similar events (throwing the stone up, forward etc) by proposing a theory. The utility of the theory is that we can use the theory to *predict* similar events.

As we progress, and hence have accumulated more observations, we want to develop more powerful theory which can predict more types of events. When one theory can also *explain* (i.e. predicts events) other events covered by other theory, we choose the more powerful theory.

At the same time, as the accuracy of observation increases, the demand on the theory also increases. The theory needs to predict to the same or higher accuracy of the observations.

An example is the relationship between Newtonian mechanics and Einstein's relativity. At human speed, Newton's laws of motion is perfectly fine in predicting the velocity of objects. As the speed approaches that of light, we need relativity to predict the velocity. However, at the same time, Relativity also produces the same prediction of velocity at human speed albeit the mathematics is more involved.

I also noted two interesting points on this process.
1. Terms are coined to represent very specific ideas used in the theory. For instance, momentum is defined as the mass times velocity. Such concepts are useful shorthand which can reduce the complexity of the theory.
2. Inevitably, mathematical models are used. Mathematics are tools developed entirely based on logic. In its purest form, mathematics are not independent of evidence or observation. Mathematics are pure conceptual construct - an art. Scientists find the logical deducing power of mathematical model useful to express complex observations. Almost all major advances of physical science is pre-dated by the development of a powerful mathematical tool. Most physical theories are now expressed in mathematical form.

The combination of (1) and (2) above makes learning science a highly demanding task. There are lots of terms to learn. These are concept shorthand and conventions. In order to be able to understand the theory, we must have working knowledge of all the terms used. As many physical theories are expressed in mathematical form, we must also have working knowledge of the system of mathematics which is used by the theory.

With these observations, I am not sure computer-based generation of theory would be useful for our understanding of the physical realm we live in.

No comments: