New Kind Of Science

A New Kind Of Science by StephenWolfram

ISBN 1579550088

http://www.wolframscience.com

StephenWolfram spoke at MIT on Thursday, April 24, 2003. His fundamental (and very well supported) claim is that the language of the universe is the language of computing, rather than mathematics. He describes conventional mathematics (prior to the invention of the computer) as a search for "shortcuts" to avoid computation. He demonstrates that such shortcuts exist for only a tiny fraction of the actual behavior of the universe.

He observes that invention and widespread availability of the practical universal computer is creating a revolution in science analogous to those that followed the invention of the telescope and microscope. When we examine the behavior of CellularAutomata in the computer, we follow in the footsteps of Galileo and Von Leeuwenhoek.


Wolfram has not only plagiarized most of the ideas in this book, including every critical idea, but he randomly changes vocabulary, so the hapless reader will not only be left wondering where the substance is behind his claims (buried in the histories of the fields he touches upon), but left with no map to find them. You may end up worse off as you may internalize some of his vocabulary and thus wander the scientific earth speaking in jargonified tongues.

What's truly disappointing for a book called A new kind of science is that it is wholly unscientific in construction. He published it with a VanityPress for starters, so no PeerReview. He cites nobody, and often lays claim to others' ideas. He also only makes descriptive claims, rather the verifiable, predictive claims. While his most interesting ideas are towards a unified underlying model for physics, it's hard to claim they amount to anything useful to make predictions.

For those who don't understand why it needs to make predictions, you have to ask yourself at the end of the exercise, "so what?" So what if the world is a giant computer program? So what if the world is a unknown dreamers' dream? So what if it rests on the back of a giant sea turtle? What can we learn about the world from this theory that we can turn back and use to do more with the world around us? Even if the world as a computer program, unless we can measure an effect from that, it doesn't matter, just like it doesn't matter if we are a dream or we live in Brahma's nose.

From http://www.lurklurk.org/wolfram/review.html, "Again, this phenomenon of computational irreducibility is not Wolfram's discovery, but the key point here is that you can begin to see why "traditional" science has not devoted much attention to this kind of modelling. For if you cannot get conceptual understanding or useful predictions from a model, what use is it?"

Basically, save yourself your time and money and just read Joe Weiss' Amazon review, which said more in 2 pages of text than Wolfram does in 1000. (cf. http://www.amazon.com/exec/obidos/tg/detail/-/1579550088/102-3773809-9128108?v=glance) -- SunirShah

You have to be a little careful here Sunir; Wolfram *was* responsible for some of the early critical ideas in CA, just not most of them. He also hasn't done anything of note recently (his unsubstantiated wildly handwaving ideas about universiality aren't that interesting in their current state). However, on the whole, I agree with you that it is a deeply flawed book. Worth a look for the figures if nothing else, though.


If what he says is true, it's as important as Newton's Philosophiae Naturalis Principia Mathematica.

Note - I haven't read this work, but I thought that science was supposed to help us understand the real world and thereby make accurate predictions. The material being presented in this book is reported by several (I can't say many - yet) as appearing not to do that. Perhaps rather than being a new kind of science, what this book describes is in fact something entirely different from science. Maybe if people whose opinions I trust give other than very negative reactions I might read it, but until then it remains a self-published book by someone who has kept himself isolated from the mainstream.

To get slightly less meta, Wolfram essentially proposes that CellularAutomata are a useful substrate for modelling the world, in much the way that DifferentialEquations are. Furthermore, he claims that CA are better than normal maths, being more powerful, able to deal with more complex systems, etc. He might also be claiming that CA can do transcendentally more wonderful things, although at this point it looks very much like he's HandWaving.

To go meta again, it probably doesn't help that he's so confrontational about it; yes, CA are undoubtedly useful, but dissing the centuries-old mathematical traditions we have at the moment isn't big or clever.


In my opinion, NewKindOfScience suffers primarily from needing a MercilessEditor? (as does Hofstaedter's last tome). I don't read NewKindOfScience as "dissing the centuries-old mathematical traditions" - to me, Wolfram simply claims that they explore a relatively narrow segment of the problem space. Wolfram's claims don't strike me as any more arrogant then, for example, BenoitMandelbrot?'s in FractalGeometryOfNature?. Furthermore, virtually every example and experiment in the book is readily both reproducible and falsifiable by anyone who wishes, using material published in the book. There is certainly difficult material in this book - but surely that, in itself, does not constitute "HandWaving". -- TomStambaugh

Every experiment and example, yes. This is one of the true values of the book; lots of examples and good figures. Unfortunately, it is much weaker in other areas. In several places, Wolfram makes claims based on experiments that are not referenced, leaving us to take his word for it. Worse, we have things like one of the central 'new ideas' he describes in the book, the 'Principle of Computational Equivalence', is never properly defined, and is basically unsupported in the text (here is some of the handwaving). In several places, he makes grandiose assertions that simply are not supported by what he offers in the book.

Are you seriously suggesting that you've read both chapter 12 (pp 715-846) and the notes from chapter 12 (pp1125-1197), and that after reading that material, you find the principle not "properly defined" and "unsupported in the text"? You can lead a horse to water...


Note - I haven't read this work, but I thought that science was supposed to help us understand the real world and thereby make accurate predictions. The material being presented in this book is reported by several (I can't say many - yet) as appearing not to do that. Perhaps rather than being a new kind of science, what this book describes is, in fact, something entirely different from science. Maybe if people whose opinions I trust give other than very negative reactions I might read it, but until then it remains a self-published book by someone who has kept himself isolated from the mainstream.

Perhaps you might read the book before you attack it. Somewhat like listening to someone before you disagree with them.

What I wrote is not an attack. What I wrote is a summary of the opinions of several people who have read it and whose opinions I generally trust.

I make a lot of decisions based on other than first-hand knowledge. We all do - we have to - we can't know everything. We rely on people whose experience and opinions we trust. Based on that, I specifically haven't dissed the book. I said that the reviews I've read and the people I've spoken to have left me with the impression that the work in this book isn't great at predicting things. For me, science is about understanding the world so we can make predictions with confidence. Nothing I've seen or read about this book suggests that it does that. Based on that I suggest it's not a new kind of science, but something entirely different from science.

Would someone who has read the book care to persuade me otherwise? Can the material here presented be used to make predictions? Has it been used to make a single discovery? Perhaps it is the most amazing work since EuclidOfAlexandria's Elements, but is it science? That was all I questioned.


Prominent in the book are many fruits of the work of EdFredkin, the inventor of CA. Unattributed. -- PhlIp

StephenWolfram mentions EdFredkin multiple times, references his work, and the NewKindOfScience index has entries for him. What more attribution do you want? -- TomStambaugh

http://www.theatlantic.com/issues/88apr/wright.htm

Notwithstanding the work of Fredkin, Wolfram makes several false claims of discovery, or the implication of them by not mentioning the researchers responsible for the ideas. A lay reader of the book could come away with two deeply flawed ideas: 1) this work was mostly, if not all, done by Wolfram (it wasn't) and 2) this work is mostly new (much of it is 20 years old). What more do you want? Any book of this size, scope, and subject area should have a good bibliography, and its lack is a serious deficiency in this book.

I read the book, I listened to Wolfram lecture about it, and I participated with him in a conversation about all this at a signing. I am a lay reader. I did not come away with the idea that "this work was mostly, if not all, done by Wolfram" and I did not come away with the idea that "this work is mostly new". I did hear him say (and write):

In the early and mid-1980's, I did collaborative work relevant to this book - some published, some unpublished - with several people: Richard Feynman ..., Oliver Martin ..., John Milnor ..., Andrew Odlyzko, ..., Norman Packard ... and Jim Salem ..." (pp xii-xiii)

Just over twenty years ago, I made what at first seemed like a small discovery: a computer experiment of mine showed something I did not expect ... So I resolved just to keep working quietly until I had finished, and was ready to present everything in a single coherent way. Fifteen years later, this book is the result." (p ix)

Have you read the book yourself, including the second half? Have you corresponded with Dr Wolfram about any of these "false claims" or "implications"? Did you verify them yourself (and can you cite them please), or are you simply repeating criticisms of a book you haven't yourself read?

I ain't read it yet. Ain't gonna. Ain't gonna pretend I did when I diss it.

The book makes an excellent decoy. -- PhlIp

So apparently what you meant to say, above, was "A lay reader of reviews of the book..." - I take it you did trouble yourself to read at least a review or two, right? It appears to me that any "handwaving" and "grandiose assertions" are more likely to be yours than Wolfram's, don't you think? Thanks, at least, for the honesty to admit your perspective - it helps calibrate your comments about the book.

Hmm. There are several authors here. I wrote the ..lay reader... above, but nothing following. I have, however, read the book. Wolfram mentions many of the appropriate names, but doesn't tell you who did what. Undoubtedly, he made big contributions to the field. Some of the work was collaborative, yes. Some of the work Wolfram had nothing to do with at all. From reading this book, you have no idea who did what, or where to look for the details. For example, he mentions Packard; however, in the section around snowflakes (around p370), he somehow forgets to mention that all this is based on work done by Packard, not Wolfram. Wolfram had to know about Packard's work, because he edited the volume this came out in. There are many, many instances of this throughout the book. In some cases, he outright claims to have discovered something that was other people's work (I'll have to dig for a ref. here). Of course, S.W. mentions (several times) that sections of this work were done twenty years ago. However, I maintain that a lay reader can come away with the following impression: "Twenty years ago a lot of people were working on this. S.W. became frustrated with them, and went off to work by himself. The book is the result of this work". Realistically, the book is a summary of many people's work over the last 20+ years, some of which S.W. was involved with, some of which he wasn't. It misses entire sections of CA theory, often fails to credit work to original sources, and makes many grand hand-waving claims. Despite its many flaws, it is an interesting book. Some of the ideas he proposes *are* interesting ideas - they are just not very well developed yet (e.g. some of his ideas on universality etc.). Some of them are unlikely. All in all, it is well worth reading, although it could have benefited from a harsh editor.

I, PhlIp, am about to add my 3rd epithet to this wonderful page. Someone has seen fit to blame me for the "grandiose assertions", just because I'm the only one here with the 'nads to sign. I did not participate in the grandiose assertions, but I'm certainly not about to interfere with them on either "side"...

So, without further ado, my 3rd and final epithet:

Math is a democracy. -- PhlIp

I, TomStambaugh, carefully use "you" in the full knowledge that it can apply to any of the authors, singularly or altogether. I enthusiastically agree with your observation that NKS would have benefited greatly from a harsh editor. I suppose I'm more forgiving about his attributions, since I was focused on the content. I'm somewhat familiar with the field (though still very much a "lay reader"), and happened not to make the various mistakes you express concern about. I worked for a Fredkin startup in the early 80's (Three Rivers Computer) and so I'm familiar with his contributions and persona. It was apparent to me that NKS is a summary that includes many others' work.


He really shouldn't have published his theories without making a single verifying discovery. Now it's just a very big and pretty bag of fluff.


Originally one of the MeditationsOnEmergence:

While Wolfram's tome has lots of interesting material in it - and maybe even some important new ideas - it doesn't strike this reader as a "bottom line" on emergence ... or much of anything else. It's much more like an invitation to further research.


Related:


CategoryBook


EditText of this page (last edited August 31, 2006) or FindPage with title or text search