SCHOOL OF ENGINEERING AND COMPUTER SCIENCE

Mechanisms of analysis and perception

Before the discoveries in this book, one might have thought that all perceptual processes were, in some sense, complex, and that as a result complex analytical systems might be required to detect regularities in them. But as the discoveries in this book have shown, some perceptual processes that might seem complex to a complete moron born yesterday can in fact be carried out by very simple programs. And indeed in some sense this is what we have seen in this book. Marcus Frean

Main points:

0. One-sentence summary. The basic mechanisms responsible for the processes involved in perception and analysis can be captured by simple programs based on simple rules.

1. Perception and analysis. Perception and analysis are processes that try to summarise large volumes of raw data, by (a) removing irrelevant data (for a particular purpose); and (b) removing redundant data (by finding regularities).

2. Limits of perception and analysis. Perception and analysis have the most difficulty summarising random and complex data. Even when random or complex data is generated by simple rules, it can be very difficult for processes of perception and analysis to reverse the process, and find a simple summary of the underlying procedure used to generate the data.

3. Randomness. Random data should not be defined as random by whether or not a short description of it exists; such a definition is not useful because exhaustively searching for short descriptions is intractible. A more useful definition of randomness says that something should be considered random if it appears random to our standard processes of perception and analysis. And because our processes of perception and analysis are based on simple programs, we can say that something is random if no procedure based on simple programs can find regularities in it.

4. Complexity. Data should be considered complex when some features are simple for us to describe, but other features have no short description that can be found by our standard processes of perception and analysis.

5. Examples of perception and analysis. Our standard methods of perception and analysis are Data compression, Irreversible data compression, Visual perception, Auditory perception, Statistical analysis, Cryptography & cryptanalysis, Traditional mathematics and mathematical formulas. All of these tend to be good at finding simple repetition, and some are good at recognising nested structures, but they seem to be bad at summarising complex-looking data, even when a short description of the data based on simple rules exists.

6. Higher forms of perception and analysis. Human thinking, once we discover how it works, will turn out to be based only on simple programs, and thus will be unable to recognise complex patterns. Space aliens might have forms of perception and analysis that can recognise all kinds of patterns that we fail to detect, but we can ignore that for now, because we're just trying to understand the world in terms of our own experience of it.

Richard M

back to Wolframania