J. Doyne Farmer disputes the integrity of a fellow academic's bestseller.
Although Steven Wolfram's recent book, A New Kind of Science , has received an unusual level of attention in the popular press, the real story has largely been missed.
The press describes a child prodigy, a Cal Tech professor of physics at 20, and a successful businessman, who spent ten years in seclusion writing a magnum opus because the scientific community did not properly appreciate his genius.
What is not as obvious from the popular press is that there is little in Wolfram's book that an expert in the field would regard as new. In 1983 Wolfram and I organised a conference and edited a proceedings volume on cellular automata, the main topic of his current book. There is very little in his recent 1,200-page tome that was not well known then.
What is also controversial is that by giving no bibliography, and burying most of the attributions in a rambling section in fine print at the back of the book, Wolfram runs the risk of misleading the lay reader. He mentions the work of the most famous players in the field, but typically inserts his own work and exaggerates its importance to make it seem more significant and influential than it really was.
When he discusses work that was published in more obscure places, the original authors of the work are often not mentioned at all, and the work is presented as though it were entirely his own.
The reader could also get the impression that he is largely responsible for basic ideas that have been central dogma in complex systems theory for 20 years, such as the observation that extraordinarily simple models can give rise to complex behaviour, or that nature can act as a computer.
The only result that is possibly both original and significant is that a particular cellular automaton, rule 110, can act as a universal computer, that is, that it can simulate a computer. This would indeed be a major contribution if there were not already a previous proof for a cellular automaton rule that is almost as simple, and if Wolfram were the main author of the result. Rather, if you read the fine print carefully, you discover the proof was done by Matthew Cook, who was one of Wolfram's paid assistants.
This brings up the aspect of Wolfram's book that might be genuinely considered a new kind of science, and that is the manner in which the science is produced. Wolfram is not just a person, he is a corporation. He hires researchers to assist him, has them sign strict intellectual property contracts and keeps them well out of the limelight.
Several years ago, the Santa Fe Institute tried to publish Cook's proof in a proceedings volume, but decided to withdraw the paper when Wolfram threatened to sue if it appeared before his book.
Finally, what about the science itself? A key demand of scientific theories is that they make predictions. There are no predictions here. There are some nice qualitative comparisons, for example between the interesting patterns of seashells or snowflakes and those produced by simple cellular automaton rules. (In the case of snowflakes, there is no mention that similar results were originally obtained by Norman Packard, without Wolfram as co-author, while he was a postdoc of Wolfram's in the mid-1980s).
Conceptual models such as those in this book can be valuable, and have been a staple of complex systems theory for decades. But sensible people know that they will never constitute more than a first step down the path of real science. For the moment, I think it is wiser to hold on to the old kind of science, where collaborators and earlier work are properly acknowledged, and the ultimate test of a theory is whether it makes falsifiable predictions.
J. Doyne Farmer is McKinsey professor at the Santa Fe Institute, New Mexico.