A New Kind of Science [Hardcover]

With some of the negative reviewers, I will agree that:
- The book is long.
- It contains self-praise.
- Others have explored some of its ideas.

If you are in the mindset of peer-reviewing a conference/journal publication, those three things might very much bother you. However, this is not a conference/journal submission! This is a book for a very wide range of audiences, and it is always very difficult to satisfy such a range. IF your own ego can get over the author's self-praise, then you can really enjoy the book for the following:

- A thorough exploration of a very important idea, which may sometimes seem obvious but when actually incorporated into our thinking can indeed profoundly affect the way we approach some scientific problems.
- A fascinating demonstration of how science selects the problems it considers important simply based on its ability to solve them. This works both ways, with the book pointing out how classic methods completely avoid certain problems, but also happens again in a new way in the course of the book as Wolfram himself selects problems to solve based on the applicability of the concepts he introduces.
- A delightful conglomeration of fascinating concepts and problems from all kinds of fields, including computer science, physics, biology, philosophy, etc. If you read the notes, this book takes you on a grand tour of the state of science in many areas, and I can't even imagine the effort that must have gone into compiling, understanding, and organizing all this information. I understand why it took 10 years, and this alone makes the book worth its value.

Put your ego and the egos of others aside, and simply enjoy!

So far so good. Wolfram's next contention is that the complexity found in what he calls Class 4 cellular automata cannot be exceeded by any physical, biological or computational process. Put more boldly, every physical, biological, psychological, financial, meteorolical and, no doubt, astrological feature of the universe that exhibits complexity is generated by some sort of cellular automaton with appropriate initial conditions.

Such a statement cannot, of course, be proved in any acceptable way. To compensate, Wolfram gives us many examples of phenomena whose random behaviour resembles those of cellular automata. He is most convincing with his pictures of real seashells and arguments about turbulence in fluids (I especially liked his wafting smoke in the air anology.) He is less persuasive when he argues that evolution has nothing to do with maximizing anything and everything to do with generated patterns, some of which survive. When he talks about the analogy between Class 4 cellular automata and human cognition, he is downright silly.

Yet this is all irrelevant. Wolfram is scathing in the inability of mathematics to solve anything but the simplest physical problems. Thus Newton could tell us how to calculate the orbit of a planet around a star but neither he nor any of his successors could come up with a reasonable mathematical model for turbulence. And no one has even attempted a mathematical model of evolution.

But describing the disease is easier than prescribing a cure. Suppose that Wolfram is correct and that every meaningful physical and biological process is generated by an ongoing cellular automoton--or something equivalent. Then we could understand how we got where we are and predict where we will go. All we need is to discover the underlying rules and initial conditions for each system we wish to model. But therein lies the rub.

Wolfram argues persuasively that the systems generated by Class 4 cellular automata are irreducible. This means that there is no shorthand method for calculating future behaviour. The only thing we can do is go through the iteration millions, billions, gazillions of times and observe the outcomes at each step. Since the behaviour is random, knowing where you are at any step doesn't help you to predict where you will be at a future step.

The inverse problem is far more intractable. It is practically impossible to determine the underlying rules and initial conditions of a cellular automoton by looking at the deterministic pattern that it generated--especially if the pattern is complex and random (the only case of interest). But that's the whole point. Even if we knew with certainty that some complex process was generated by a cellular automoton with simple rules, it would still be impossible to describe its past behaviour or predict its future because we could never find the rule and starting conditions.

So, at its most profound level, even if Wolfram's new science is correct, it fails at doing two of the most fundamental things that science is supposed to do: telling us how we got where we are and making predictions about future behaviour.

In the final analysis, Wolfram's book is brilliant and well worth reading. But its new ideas may prove to be as useful as those in astrology.

I was interested in Wolfram's underlying ideas. But could barely discern what those ideas were between the self aggrandizement and atrocious writing. It becomes obvious very rapidly why he didn't propose these ideas to the scientific community at large; he would have been reviewed.

Wolfram's lead in claims incredible depth and application, but the book doesn't deliver. I was not only disappointed but angered by the author's tone and style. If he had stuck with the facts, the book would have been interesting, readable, and probably about 300 pages. As it is, it is hardly worth wading through it.