200 of 202 people found the following review helpful
This book is easily the best introductory "guided tour" of complexity I know of. It has several key strengths:
1. Mitchell covers many of the major topics which can reasonably be grouped under the umbrella of complexity, so the breadth of the book is excellent. For my benefit and yours, here are the main topics covered, roughly in the order they appear in the book: chaos, information, thermodynamics, Godel's theorem, Turing machines, evolution, genetics, measures of complexity, fractals, self-reproducing automata, genetic algorithms, cellular automata, artificial life, information processing in living systems, analogy-finding algorithms, game theory, networks, power laws, metabolic scaling, random boolean networks, and historical foundations of complex systems research (cybernetics, general systems theory, synergetics, etc.). This long list leaves out some significant complexity topics, but Mitchell's scope is still plentiful for an introductory guided tour.
2. The topics are covered in sufficient depth to clearly convey the key concepts, which reflects the fact that Mitchell is a scientist who really knows the subject. Though the treatment is certainly introductory, rest assured that this isn't a superficial journalistic popularization which drops lots of names and terminology without getting into any real content.
3. Mitchell's writing style is concise and precise, but still friendly and not at all terse. The book is quite easy to read if you have a decent background in general science.
4. General readers will appreciate that there isn't much formal math in the book, yet Mitchell explains things in a way that nicely intimates the outlines of the math for readers who are math-savvy.
5. Mitchell's presentation is sober and honest. She naturally highlights the potentials and promise of complex systems science, but she also openly acknowledges its past dead ends and likely future limitations.
6. There are biographical notes interspersed throughout the book, which adds a nice human touch.
For completeness, I'll note that I did notice a few technical errors in Mitchell's initial discussion of immunology. But these errors don't invalidate the general message, and can be overlooked, considering the overall excellence of the book.
The bottom line is that I highly recommend this book to anyone interested in complexity (how could you not be?). It's a perfect introduction for beginners, and people well-versed in the subject will also appreciate the convenience of having a high-quality broad overview within the covers of just one book.
128 of 133 people found the following review helpful
From reviews of the book that appear on the back cover:
"...scholarly yet entertaining..."
"...best general book on this topic."
"...entertains and informs all the way..."
I agree with all of the above. Unlike many books on complexity, this book is easy to read and highly accessible to general readers. More importantly to me as a graduate student, this book is more fascinating and in many ways more thought-provoking than math-heavy textbooks for specialists/academics.
I bought the book because of my interest in artificial intelligence, and I highly recommend this book to anyone interested in artificial intelligence, computer science, or biology. What I like most about the book is that it provides me with a fresh perspective/synthesis that pulls together what has been going on in different fields and subfields. For example, in computer science, we are taught all the time about how important it is for programs to be able to scale, but we are not given a biological perspective of how genes scale so well. This book does that in it's chapter on scaling.
Each chapter includes historical perspectives and/or real-world examples. For example, the chapter on genetic algorithms includes a quick survey of the companies and organizations that have recently benefited from using them.
The book also includes a chapter on why computers are still pretty dumb (lack general intelligence). The chapter reiterates that analogy understanding may be the holy grail to developing artificial general intelligence. (Like most people, I agree with the author that artificial general intelligence, AGI, is not going to happen anytime soon.) Some relevant info about the author from Wikipedia: "She received her PhD in 1990 from the University of Michigan under Douglas Hofstadter and John Holland, for which she developed the Copycat cognitive architecture. She is the author of "Analogy-Making as Perception.""
60 of 62 people found the following review helpful
This book really lives up to its title, "Complexity: A Guided Tour." Dr. Mitchell has turned her Santa Fe Institute lectures on the foundations of Complexity into a very interesting, readable book suitable for academics, professionals, students, and interested laypeople. She explains how complexity fits into the history of scientific knowledge. She relates it to the rapidly expanding field of information science, as influenced by biological rather than mechanical models. She even explains how computer models relate to living systems as information processors.
Having read many scholarly papers on these topics, I can vouch for the clarity and accuracy of her work. She certainly doesn't need any endorsement, though; as a successful doctoral student under the renowned Doug Hofstadter and now a professor at Santa Fe, she is in the inner circle of complexity scientists today. If only her book had come out a year or two ago! It puts in one place many ideas we used to have to search out and integrate on our own!
One note: the mathematics of complexity science can be daunting. Dr. Mitchell has done a terrific job expressing & explaining those concepts. Unlike many of the complexity books in print, hers is both intelligent and accessible. Highly recommend it!
37 of 37 people found the following review helpful
Complexity: A Guided Tour, by Melanie Mitchell, Ph.D.
One of my favorite books from the early 1980s was a huge tome titled Gödel, Escher, Bach: An Eternal Golden Braid, by Douglas Hofstadter, a pioneer in Artificial Intelligence (AI). Hofstadter described GEB (the initials became a popular abbreviation for his book) as "a metaphorical fugue on minds and machines in the spirit of Lewis Carroll". At the time I was just getting interested in AI and I found GEB fascinating. Apparently, I was not alone. Melanie Mitchell, then a high school mathematics teacher in New York, found it "one of those life-changing events that one can never anticipate".
She wrote to Hofstadter that she wanted to study under him as a graduate student. Receiving no reply, she later approached him in person when he gave a lecture at MIT. He handed her off to a graduate student. She was "disappointed, but not deterred" and after several more follow-up calls to him, she managed, through her persistence, to convince him of her passion for AI - a topic that eventually was absorbed into Complexity Science. Eventually, she moved to Michigan and earned her Ph.D. under Hofstadter and John Holland, another complexity science pioneer. I mention this history to try to convey the contagious enthusiasm for complexity science that Dr. Mitchell exudes in her book. She seems to prefer the term complexity sciences, since this is such a cross disciplinary subject; but in this review I'll use the more common term complexity science.
Mitchell starts with an acknowledgement to the Santa Fe Institute (SFI) where she directed an SFI Complex Systems Summer School. The SFI seems to be the current epicenter for complexity science research, and this book is an expansion of the author's series of SFI lectures on "The Past and Future of the Sciences of Complexity", with updated material reflecting new perspectives from 2008 and 2009.
Previous knowledge of complexity science is unnecessary, as the first chapter starts out with a series of examples to describe what is meant by complexity. This was useful since the topic seems to evoke many different definitions from scientists and practitioners. Those of us in the financial sector like to start with some definition of the topic under study; but a rigorous and widely accepted definition of complexity science just does not exist yet. On the other hand, we spend vast amounts of time developing potential strategies for risk management - even though we may differ considerably in our opinions about what constitutes risk. In a similar vein, Mitchell's examples make it clear what falls into the realm of complexity. The examples run the gamut from insect colonies to the human brain; and from immune systems to economies and the World Wide Web. In some respects, ERM seems like an application of complexity science; and quoting A.S. Eddington, the astronomer who first demonstrated that Einstein's Theory of Relativity worked in the real world, "We need scarcely add that the contemplation in natural science of a wider domain than the actual leads to a far better understanding of the actual." I submit that a study of the wider domain of complexity science can help us better understand risk management. In fact, lest the finance oriented person reading this review assumes that the book mentions only theory and some science applications, the author peppers her theory with references to practical financial applications in several sections. She explains early on that:Economies are complex systems in which the "simple, microscopic" components consist of people (or companies) buying and selling goods, and the collective behavior is the complex, hard-to-predict behavior of markets as a whole, such as changes in the price of housing in different areas of the country or fluctuations in stock prices.
And later in the book she gives specific examples: GAs [Genetic Algorithms] have been used by several financial organizations for various tasks: detecting fraudulent trades (London Stock Exchange), analysis of credit card data (Capital One), and forecasting financial markets and portfolio optimization (First Quadrant).
Her extensive notes section refers the reader to details about each of these specific applications.
In Complexity: A Guided Tour, we are given a short history lesson on the roots of Dynamical Systems Theory, Chaos, and Prediction. Again, the examples help guide the reader through an inductive learning process. Deterministic Chaos, for example, is introduced via the famous Logistic Map that results from varying values of R in the seemingly simple equation xt+1=R*xt*(1-xt) where 0'xt'1. Along the way, we hone in on Feigenbaum's constant, a universal constant for functions approaching chaos via period doubling, and the fact that it applies outside the realm of pure mathematics and shows up in electronic circuits, lasers and chemical reactions.
Now, we are ready to approach the concepts of Information, Energy, Work and Entropy. This is explained through stories about the development of the Second Law of Thermodynamics, Maxwell's Demon, and Shannon's Information Theory. Moving along to Computation, Mitchell guides us through topics such as "What is Computation and What Can Be Computed?" She describes Hilbert's Problems and Godel's Theorem, which proved that not all mathematical questions are computable. Then she covers Turing machines, where the goal is to mimic human behavior so well as to fool a human, and this leads into a chapter on evolution. Her primer on evolution summarizes pre-Darwin, Darwin, Mendel and the Modern Synthesis, and leads quite naturally into the next chapter, on Genetics. Skipping quickly through an admittedly simplified treatment of DNA and RNA Mitchell leads us into the geometry of fractals, and the underlying power laws that describe them when normal measurement techniques fail us.
Now with approximately 100 pages of history and basic tutorials behind us, we can begin the next parts of the book, which deal with topics such as life and evolution in computers, cellular automata, information processing in living systems, genetic algorithms, ant colony optimizations, and the mystery of scaling. Clearly oriented towards AI, the author devotes a major chapter to applying network science to real-world networks - such as the brain.
Each topic is approached in a logical, understandable manner. In addition though, as a reader I felt the excitement of the discovery process as I read about Von Neumann's self-reproducing automation, the "new Kind of Science" from Wolfram, and the gradual increase in intelligence of Robby, the soda-can-collecting robot, like the movie robot WALL*E, which eventually outperformed the author in developing its own clean-up strategy.
A chapter is devoted to an overview of the author's development of "copycat" - a program she wrote for her doctoral dissertation that makes analogies in the letter-string world by using reasoning believed similar to that used by humans as we make analogies to understand our world.
The examples often caused me to stop and write a quick spreadsheet or program to further explore the particular subject.
This is one of the first books I read on Complexity Science; and admittedly many of the ones I read afterwards were more narrowly focused; and some went into more detail, or provided even more memorable examples on particular complexity science topics. However, this book gave me a base level understanding of a lot of topics that previously were just fancy sounding phrases. More than that, it nurtured my initial interest in complexity science and left me with a voracious appetite for more! The subtitle is appropriate. This is truly a guided tour for complexity. Dr. Mitchell is an excellent guide; and I recommend her for your visit to the amazing world of Complexity Science.
132 of 164 people found the following review helpful
Complexity a Guided Tour
Review of Melanie Mitchell's book "Complexity: A Guided Tour"
This is a thoroughly disappointing book; or an eye opener. Or maybe both.
Disappointing because the book does not cover much more than many popular science books already in the market (and it promised a bit more than that). An eye opener because the topic surveyed is still fairly fashionable and comes up in the end as fairly vacuus.
Who is the author, what are her stated goals?
The author is a well known computer scientist from the world renowned Santa Fe institute. Her goal is to survey what she implicitly holds to be "the great unexplored frontier of science". So far so good. She is actually careful to point out that as she will be talking about work in progress, some of the concepts might be a bit fuzzy around the edges and the book will be as much about clarifying "whether such interdisciplinary notions and methods [as complexity, emergence etc...] are likely to lead to useful science and to new ideas for addressing the most difficult problems faced by humans such as the spread of disease, the unequal distribution of the world's natural and economic resources, the proliferation of weapons and conflicts, and the effects of our society on the environment and climate".
Judge and party
The first problem with the book is that it is far from being impartial. Mrs. Mitchell does not hide her fascination for the topics that she studies (as a matter of fact someone not enthusiastic about one own's work would probably not go very far), but this makes her less credible in her attempt to provide an objective assessment of the usefulness of her own field of studies. I found she was doing a credible job until chapter 17 (out of 19), which would not be too bad if the last chapters were not those dealing most directly with the relevance and prospect of "complexity science". But a couple of sentences really rubbed me the wrong way. More on this in the note about "the mystery of scaling", but suffice it to say at this point I don't believe anybody deserves my attention who writes with a straight face that the so-called "metabolic scaling theory" has "the potential to unify all of biology" (or for that matter anyone relaying such a claim as even credible).
Surveying old chestnuts
For a book attempting to survey "the cutting edge of science", much is covered that is fairly old and well established. Let us survey the table of content. The chapters 2 to 6 are respectively "dynamics, chaos and prediction", "information", "computation", "evolution" and "genetics, simplified". While each chapter in itself is not particularly bad, one would find better introduction to all these topics elsewhere. As I don't imagine too many readers of Mrs. Mitchell are complete science novices, the material in these chapters is therefore not particularly useful. One could object that maybe the idea is not to expose the readers to the basic facts of these disciplines, but rather to present them within a new framework that would act as an eye opener. Unfortunately, I did not find that the presentation made of these topics was enlightening in this way.
Evolution in Computers and "Computation Writ Large"
These are the parts 2 and 3 of the book and in my view one of the better ones. The presentation of genetic algorithms through one example was one of the more interesting I've seen (little robot picking up garbage comes up with a neat trick that one would not necessarily have programmed a-priori). Again, I'll levy the charge that the author does not make it particularly clear how the material she deals with in this part of the book relates to the rest and fits into the big picture. The author also covers cellular automata (a topic beaten to death by Wolfram's A New Kind of Science) and provides some examples of current research in this field that are less likely to have been previously encountered by the reader. Then comes a vanity chapter dealing primarily with the author's PhD thesis. While not uninteresting in itself, the subject does not warrant being put on equal footing with the other themes dealt within the book, but this is probably one of the lesser shortcomings of the book and one of the most understandable one.
Network Thinking and "The Mystery of Scaling": I'll bite
The next part of the book annoyed me to the extreme. Full disclosure: this is going to get emotional and somewhat ugly. If you don't like this type of stuff, please move on! Ok, if you're still reading, here's my main issues with this part of the book: the "science" it describes is all style no substance. At its mediocre seems to specialize in producing factoids that can be usefully integrated in your average popular science article or Malcolm Gladwell book. At its pathetic worst, it becomes some sort of post-modern science where the clever positioning of the results matters more than their intrinsic worth. I won't cover here all the issues I have, but will instead focus only on one example provided by the author (and already mentioned in my review above), the so-called case of the "mystery of scaling". What's going on here is that big animals have less surface to dissipate heat proportionally to their volume than smaller animals. This is something a high school student can easily understand. Given big animals do not routinely die of overheating, they must have a lower metabolic rate than small animals. One can through some sort of back of the envelope calculation predict how the metabolic rate should vary with size. The naive calculation does not seem to match experimental data very well. Then low and behold, a few heroic complexity theorist come up with a fractal network theory that seems to fit the data a bit better. My view is that this is a "cute and clever" explanation for a marginally interesting factoid. The book presents this as a revolution. I mean, come on! that's just a bit of basic geometry that does not provide any insight whatsoever into any underlying biological process. Any assertion something like this would play a role in biology "similar to the theory of genetics" is either shameless and cynical self promotion, or the result of a total lack of perspective. To be fair, the author mentions that the claims made here are a bit controversial, but I find this part a bit disingenuous to say the least. If this type of theory can in any way be put on equal footing with genetic theory, one would expect at least some sort of application. Look for it and return when you've found it... you're not going to be back any time soon.
I said I'd cover only one example, but the last chapter has a couple of nuggets that I just can't avoid mentioning. Basically according to this chapter, biology and genetics are a massive failure (I'm exaggerating somewhat, but this is a summary). Junk DNA is not junk (that's actually possible) but the most important bit of our genome (highly speculative but not flagged as such) and really understanding biology will require understanding biological networks (well, as the Dude said in the Big Lebowski, that's like your opinion). It's hard to keep one's cool when reading things like this. Basically, bench scientists who have sweated all their life to look at the details of how things actually work are wasting their time. All that one needs is a self indulgent theoretician who will come up with suggestive analogies that a biological system is like the internet and then we'll have the final word. Hmm.
Putting It All Together
Ok, I'm getting carried away a bit, so let's come back to factual facts. I quoted Mrs. Mitchell when I started my review. Her goal was to clarify "whether such interdisciplinary notions and methods [as complexity, emergence etc...] are likely to lead to useful science and to new ideas for addressing the most difficult problems faced by humans such as the spread of disease, the unequal distribution of the world's natural and economic resources, the proliferation of weapons and conflicts, and the effects of our society on the environment and climate". Did she clarify this at all?
As a matter of fact, she touched upon these topics only briefly and certainly did not provide any evidence that complexity theory had anything useful to say on these topics. If one is looking for interesting ideas on how to deal with the tragedy of commons for instance, one would be much better served by referring to the work of someone having looked carefully at practical, real world examples. Someone like Elinor Ostrom for example. One will find recommendations on how to manage the complex by understanding the specifics of one complex situation. This looks to me much more promising than drawing remote analogies between non-commensurable systems. How long can scientists get a job to ponder fascinating similarities between fractal exponents? This would actually be a good subject for a sociology of science study.
10 of 10 people found the following review helpful
This is one of these books that changes your intellectual perspective. An excellent introduction to Complexity Theory, still in its infancy with a lot of debates because it is a work in progress.
Complexity theory is perhaps the antithesis of reductionism and a complement of the most successful scientific philosophy that, however, leaves out such complex phenomena as the emergency of life and intelligence, the weather, ant colonies, the immunologic system, biological metabolism, brain functioning, self- referential systems, non linear phenomena, game theory, the economy, artificial intelligence, etc. Complexity theorists are trying to find common patterns in these different fields and take cues from dynamical systems, cellular automata, random Boolean networks, chaos theory, information theory, biological systems and others. The author defines a complex system as one that exhibits non trivial emergent and self-organizing behavior.
It seems that some of these systems have a balanced combination of randomness and determinism very well exemplified in ant colonies and immunological systems. Stuart Kaufmann says that life happens "at the edge of chaos".
The skill of the author to explain difficult questions in an easy language is demonstrated in the description of the mechanics of DNA and the explanation of what is a fractal dimension, for example.
Other interesting topics covered are genetic algorithms (which have great practical applications and solve problems in a rather mysterious way which is difficult to understand), cellular automata (a class of non Von Neumann computers), a software program to make analogies in a microworld defined by Douglas Hofstadter (author of the Pulitzer Prize winner "Gödel, Escher, Bach: an Eternal Golden Braid"), an excellent discussion of the best strategy in games derived from the "Prisoner's Dilemma", Small World and Scale Free Networks (the internet being one example), power laws, etc.
To sum up, if you are not an expert in Complexity, but are curious to learn about all these topics I would strongly recommend reading this book.
8 of 8 people found the following review helpful
Simply put, one of the most amazingly accessible and -- on the basis of some expertise coming into reading it -- accurate presentations of a massively complex technical subject that I've ever read.
As an experienced computer scientist, primarily with "von Neumann" machines, but also with some experience in connectionist architectures, her explanations of all material with which I *am* familiar are as economical and lucid as any I've seen. Her explanations of systems further afield from my comfort zones -- cellular automata, genetic algorithms, etc. -- have me feeling comfortable with my understanding of them in a way that no previous treatment has achieved.
I'm about to order her book on celluar automata, and at this point I'd buy (and recommend) anything she writes without hesitation. Scientists who are also excellent writers are rare as hens' teeth -- to use the title of a work by one of the few other hugely literate scientists, the late (and very much lamented) Stephen Jay Gould, who.
I don't know whether this niche constitutes a genre, but if so Mitchell is a master of the genre.
From a technical standpoint, Mitchell's Ph.D. adviser was Doug Hofstadter -- a pedigree that is quite evident. She's got me wanting to re-read Goedel, Escher, Bach -- no mean feat, since that's a huge undertaking.
If you have any interest in complexity, buy this book.
7 of 7 people found the following review helpful
Is Chaos and Complexity just based on two simple ideas - the sensitivity of a system to its starting conditions, and feedback? As the author details in full context, we do not yet have the proper lexicon for terms of chaos or complexity that truly describe the theory and systems discussed in the book. However, we conveniently now use the terms such as complexity, chaos, cellular automatons, scale-free networks and etcetera, but which may change in the future. This tour explores those terms and their associated concepts. Though it appears that some folks in the business have taken the author to the woodshed, based on some of the reviews, I on the other hand - as an 'autodidact' - found the book flowing and helped me with some of concepts that I had previously misunderstood.
The author also details in full that all the subjects discussed are not the end all, be all, but as an introduction to the concepts or guided tour as the title suggests. What was nice in my view was that Dr. Mitchell provides the counter points or opposing views on several of the subject matters presented which was a welcomed addition.
If you enjoy reading Complexity - A Guided Tour, as this book is a good introductory start on the subject matter of self-organizing systems, power laws, and stuff of complexity -- I also recommend the following which you may like too:
Ubiquity: Why Catastrophes Happen by Mark Buchanan (Very Good Book)
The Black Swan: The Impact of the Highly Improbable by Nassim Nicholas Taleb (Economic Game Changer),
Why Most Things Fail by Paul Ormerod (Introduction to Fractals, Power Laws),
Deep Simplicity: Bringing Order to Chaos and Complexity by John Gribbin (A Classic).
7 of 7 people found the following review helpful
Melanie Mitchell is the protege of Doug Hofstadter, and who wrote the Copycat program that models analogy making in a toy world that formed the basis of Hofstadter's "Fluid Concepts And Creative Analogies: Computer Models Of The Fundamental Mechanisms Of Thought". I still don't know why this approach isn't used in collective intelligence programming, perhaps it needs a wider audience and less rigid ideas about machine intelligence.
In complexity, Mitchell takes us on a broad tour of the subject, covering all the major bases and interleaving the threads of biology and computation into an informative cloth of complexity. What makes this book stand out from teh others in it's class is how Mitchell shows the various threads come together. Biology is science that is full of phenomena that show remarkable complex behaviors based on interacting units and she provides a few examples - ant colony foraging, the immune system and metabolism. She shows how computational techniqes shed light on how these phenomena may be explained and how we might understand biology as computation.
For me the most interesting part is chapter 11 - "Computing with Particles". She shows how a genetic algorithm evolved cellular automata rule set may be propagating information in its world. While the example is simple, it just begs for more study in different systems and seems like a very interesting idea to follow in real networks, like brains. I couldn't help but wonder if this was the missing model needed for Calvin's excellent "The Cerebral Code: Thinking a Thought in the Mosaics of the Mind".
Mitchell does her readers a great service in not just covering the broad range of topics, but also explaining where the science of complexity (if there is indeed one) fails and where key ideas are controversial and why. In this regard, her discussion of Kauffman's seminal "Origins of Order" is outstanding, highlighting the problems of his approach.
If you want a readable, thought-provoking book on complexity and computation, this is the one to buy. I found it unputdownable and read it in a single session.
7 of 7 people found the following review helpful
Melanie Mitchell has written a wonderful book on a subject that defies simple explanation. Complex systems science, as she aptly refers to the interdisciplinary field commonly called complexity, is admirably presented to a general audience without diminishing the intellectual content of the discussion. The breadth and depth of her exposition are more than adequate to convey a clear notion at an introductory level of what complex systems are all about.
After first illustrating what complexity means through a variety of real-world examples, Dr Mitchell provides a historical background of the principal theoretical bases underpinning complex systems science, namely, dynamical systems and chaos, information and entropy, Turing computation, evolution, and genetics. Following is a thorough discussion of the difficulties of coming up with a universal definition of complexity. Then a number of problem areas are investigated, such as self-reproducing systems (computer programs, DNA, von Neumann's automaton), genetic algorithms, cellular automata, dynamical information processing structures, actual living systems (the immune system, ant colonies, biological metabolism), analogy making by computers, and computer modeling and simulation as a third way of doing science, the traditional two being theory and experiment. Four subsequent chapters delve into the intricacies of networks, including the alluring mysteries of the ubiquitous power law and the complexification of genetics and evolution. The book ends with a candid appraisal of the past and future of the sciences of complexity.
That Mitchell is able to intelligently expound on such a wide range of technical topics while resorting to but a single mathematical equation (for the logistic map) is a testament to her command of the subject as well as her fluid writing skills. Her editor must have been pleased. (Well, a couple of English sentences logically amounting to mathematical equations were deployed to show a power-law distribution and the formula for the volume of a sphere but, hey, the equal signs were duly avoided!) The end notes, however, do explain the math behind various verbal assertions in the main text.
Complex systems science is the third major attempt to launch an autonomous and self-sustaining field of academic inquiry devoted to this intriguing domain. The first two, cybernetics and general systems theory (both discussed in the book), did not fare all that well. An uneasy feeling that history may recur yet again permeates the atmosphere of the final chapter. It may indeed be the case that complex cybernetic systems is much too rarefied a conceptual domain to ever congeal into a conventional discipline. Yet perhaps that is how things ought to be. For the true value of systems thinking lies in the deep transcendent insights it affords about phenomena, perspectives which traditional disciplines often can scarcely come to fathom.