Automatic revelation at a price

Stephen Wolfram's book A New Kind of Science (reviewed by Nobel laureate Philip W. Anderson in The THES , June 7) has created something of a furore in the academic world, not least because of its long, unorthodox period of gestation (over two decades) and its forthright style of presentation. Though it is a huge tome that has been a long time in the making, the book is quite approachable, given the subject matter and a willingness to excuse the author's somewhat egotistical tone.

The reason for a second review is that Wolfram and his company, Wolfram Research, have released a CD-Rom, A New Kind of Science Explorer , that can be used to experiment with the cellular automata at the heart of the book. Cellular automata are mathematical algorithms based on a simple set of rules that are applied repeatedly to some initial (normally simple) state. The juxtaposition of the changed states of the automaton for each discrete repetition can produce interesting and apparently random patterns that, when they are presented in suitable graphical form, are particularly discernible to the eye. The book reproduces more than 1,000 patterns created by techniques using cellular automata selected by Wolfram. The CD-Rom allows these to be viewed on a computer and, what is more, enables users to select different rules and initial conditions to investigate the myriad possible outcomes for themselves.

The assumption behind A New Kind of Science is that cellular automata are important in the physical world because they may lie behind the creation of the random patterns found throughout nature. Though I find the assumption convincing, many scientists do not, and it has generated much debate. The CD-Rom offers a chance for sceptics to try out Wolfram's ideas directly in a number of different fields.

The experiments on it follow the original book very closely, and it is worth having the book to hand when using the program, though this is not essential (without it, one can still skim the ideas). They cover a wide range of scientific fields and there should be something of interest to most scientists, regardless of the area of specialisation, as well as to more computer-literate non-scientists. The software is thus a useful adjunct to the book.

The areas of application are impressively wide - Wolfram has done his best to investigate as many fields as possible. Obviously, computing is well covered and, as would be expected, the ubiquitous Turing machine, a basic model for all sequential computers, is included. Mathematics too is well represented - no surprise given Wolfram's professional background. But there are also experiments on the growth of crystals, thermodynamics, data compression, visual and auditory perception and cryptography, to mention some fields that attracted my eye.

Turning to the software itself, the * symbol in the title on the cover of the packaging and on the CD-Rom betrays the commercial approach adopted in the marketing of this product. At least the CD-Rom is simple and reasonably quick to instal on a PC using Windows (and I assume on a Mac as well). The user must type in a password included on the inside of the packaging that is visible only after one has broken a seal that implicitly deems the customer to have accepted the licence agreement. Fortunately, the agreement is just a single page, although it includes the usual limitations of warranty and damages for the user that only software companies seem to have the gall to inflict on their customers.

On first running the software, one is invited to "take a tour", which is reasonably short and gives an accurate outline of what to expect. There are more than 450 interactive "experiments" based on the book keyed into matching chapter numbers and headings in the selection menu on the left of the screen. They start with simple cellular automata, while later experiments demonstrate applicability over a wide range of scientific areas. Both book and experiments attempt to illustrate Wolfram's principle of computational equivalence that unifies his approach as expounded by him at the end of the book.

Each experiment is controlled by a simple, if sometimes cryptic, input template using numbers, with pre-selected values that produce interesting results in the form of a graphical output. On trying to change numbers randomly, like the proverbial monkeys typing Shakespeare, either an error message appears (for example, if a number is out of range) or a different and normally more boring display is produced. One quickly realises that the number of "interesting" patterns is relatively small when trying random samples, although there are enough to make this method of exploration worthwhile.

As one would expect, it is possible to print out results and also to resize and export the resulting images to files for printing, inclusion in other documents and so on. Supported graphic formats include, for the technically minded, Gif, Jpeg, Tiff and EPS. This is a reasonably good selection and is certainly not a great limiting factor. Gif and Jpeg are good formats for inclusion on webpages, assuming this is allowed (of which more later). Tiff is the highest quality graphical output (with a larger file size as a result) and EPS is useful in PostScript and PDF documents (which are quite widespread).

By way of example, I generated an image that produced a 25KB Gif file, 119KB for Jpeg, 132KB for Tiff and 464KB for EPS. As expected, the Gif format was small with good results because of its lossless compression. The Jpeg size was less impressive and the lossy compression produced some minor but noticeable degradation in white areas when printed. The Tiff image produced the best printed results but, surprisingly, the EPS produced the worst results with some aliasing effects (perhaps due to the external software being used). The EPS file seems excessively large and, when converted (using software not supplied on the CD-Rom) to the similar PDF format, reduced in size to a more reasonable 139KB, less than a third of the size. (An annoying feature of the software is that the correct graphics file-name extension is not added automatically, at any rate on the PC version. This would be easy to add in a future version for the convenience of users.)

Use of any graphics produced by the Explorer other than for personal use is included under "prohibited uses" in the licence agreement "without the prior, express written consent of WRI" (Wolfram Research Inc.). A website address to request permission is also included, and visiting this results in the ubiquitous "coming soon" message. The more than 1,000 images included in the original book are protected by copyright anyway, but the situation is less clear for newly generated images where a researcher could expend considerable time and expertise in finding interesting and previously undiscovered results. As it stands, without clarification of the permissions situation (including whether or not the permissions will cost money), the product is little more than a toy for personal use. I would advise anyone considering using it for anything more serious to wait until the permissions situation becomes clear and the associated website no longer says "coming soon". The situation for use in teaching is also unclear. For example, are academic site licences available at a reasonable discount?

To use the graphics "export" feature, it is necessary to register online using an Explorer registration key. It seems that Wolfram Research is being very protective of any results that this software might produce. This is not the most open "new kind of science" that could be envisaged and is not one with which I (and perhaps many academics) am enamoured. There is very little scientific altruism on view in this product.

A nice feature of the software is the recording of experiments that have been undertaken during a session. It is possible to go back and review what has been done. However, the Explorer is memory-hungry, and storage of several experiments can result in adverse performance, even on a modern PC with a reasonable amount of memory. One may wish to close most other applications to obtain full performance from the software. It is very easy to run out of memory, requiring the program to be aborted and restarted.

Finally, it is also possible to share experiments by emailing input data to others for use on their own copy of Explorer . This has the potential of being useful in a teaching situation where a set of initial experiments could be distributed to students to start some practical study using the software.

The interface with the software is simple enough for most scientists to be able to conduct the experiments even if they are not very well versed in computer technology. How much they will get out of them is another question. After a while, the experiments start to seem repetitive. I suspect that really successful use will require deep knowledge of the area of application.

Overall I do not believe the CD-Rom will enjoy the same sales success as the book, but it is a worthwhile product that will sell extremely well by the standards of most academic offerings. However, it does not seem to be regarded by Wolfram Research as a serious research tool or even as a teaching tool. I guess that Wolfram would expect users to buy the full Mathematica mathematical support software, also from Wolfram Research, on which this product is based.

In summary, for anyone who has purchased and enjoyed the book, this CD-Rom may be a diverting product, at least for a while. Others may prefer to save their money, or if they are undertaking serious mathematical research requiring computer support for graphical output, to purchase a copy of Mathematica instead.

Of course I have had only a relatively short time to assess this software. Wolfram has spent about 20 years working in this area. With his extremely sharp intellect, he obviously has a huge head start on most people in terms of expertise in finding interesting examples of randomness. Others will generally have a hard time catching up but may enjoy trying, at least with experiments relating to their own field. But do not start with the CD-Rom. Buy the book first, and purchase the CD-Rom only as an aid to understanding and enlivening some of the ideas propounded at length in the book.

Jonathan Bowen is professor of computing, South Bank University.