David Deutsch is a well known theoretical physicist whose research focuses on quantum physics and the relatively new field of quantum computation. He is a member of the Centre for Quantum Computation which is part of the University of Oxford. "The Fabric of Reality" is a popular science book, in the sense of discussing his ideas on science mostly in non-mathematical terms. The book seems aimed not only at the intelligent lay person but also at fellow scientists and philosophers of science, some of whose mainstream ideas he seeks to criticize.
As often in a book like this, the writer seems to want to cover as many of his ideas as possible. This results in the inclusion of many apparently not directly related points which serve to some extent to distract from the major themes. But I think any one of us, as a writer, would have such a tendency. We all view reality from our personal perspective and acquire a kind of integration of the various bits of knowledge that we happen to have. Therefore we tend to see a stronger relationship between those pieces of knowledge than someone else will.
According to Deutsch the four main strands of explanation of our current scientific theories are quantum physics, evolution, computation and knowledge. A main theme of his book is showing how these four subjects are interconnected and allow us a deeper insight into "The Fabric of Reality" than each of those on its own is able to allow us. He goes on to claim that the combination of our best theories in those four areas brings us close to a theory of everything. With that theory of everything he does not mean the search for a Grand Unified Theory of Physics, which would combine the theories of general relativity, quantum theory, nuclear forces and electromagnetism into one theory. He means a theory that gives us a deep understanding of our world on a higher sort of level.
Probably many scientists would choose a different set of four theories or subjects which they consider most fundamental to our understanding of reality. I think Deutsch's choice is less fundamental than he would have us believe. Not that there is anything wrong with choosing four areas of science which one considers important and interrelated and writing a book about it. It's a good way to bring structure to a book and can provide for a useful view. I only disagree with the Deutsch's suggestion that his chosen structure is fundamental.
The book doesn't convince me of being the basis of a theory of everything. In fact, I doubt very strongly that a high level theory providing a basic understanding of all phenomena is possible. It's surprising to me that Deutsch would suggest it, since it seems to me that such a theory would have to be reductionist and therefore in conflict with his own excellent critique of reductionism (which I'll discuss near the end of this article). Also, I think the book does not succeed very well at demonstrating an important connection between the four strands.
Another of the book's themes is Deutsch's philosophy that theories in physics provide explanations of reality rather than merely predicting the outcome of experiments. I find myself to be closer to the opposite school, which he calls instrumentalists. I have always considered physics to be more about describing nature than about explaining nature. I consider the idea that physics not only describes and predicts nature but does something more, in that it provides some deeper explanation of why nature behaves a certain way, to be wrong. Most physicists today, I think, would agree with my view. Deutsch sees that as something bad while I see it as a positive sign of science having reached a certain level of maturity.
But the instrumentalist portrayed by Deutsch may also be a bit of a straw man. He describes the example of the instrumentalist who is only interested in predicting the patterns of projections of stars and planets on an astronomer's photographic plate. Deutsch correctly argues that the specific patterns by themselves are not important or interesting, but what is important and interesting is an understanding of the underlying laws of physics which describe, explain if you will, the emergence of these patterns. My problem with Deutsch is that he seems to be looking for some kind of explanation behind the explanation. Once you've described how nature behaves, preferably with mathematical formulas, you're done. It is an empirical fact that nature behaves in systematic ways which can be described with models and mathematical formulas. And there is true beauty in that. But I am not aware of any single theory or formula in physics which does anything other than describe and predict phenomena in nature. And yet it seems that Deutsch is looking for something more than that.
I would say the various philosophically different viewpoints do not really have an effect on the ability of physics to progress. The explanation focused physicists and the instrumentalists are both equally capable of understanding the essence of physics, which largely involves applying mathematics. But Deutsch may disagree with this since he seems to argue that his philosophical view is a requirement to a complete understanding of his four fundamental strands.
One symptom of Deutsch's and other's quest for a deeper meaning behind the phenomena of physics is the idea of an interpretation of quantum physics. Deutsch criticizes the Copenhagen interpretation of quantum mechanics while he defends the so-called many worlds interpretation of quantum mechanics, although I believe Deutsch himself does not call his theory of multiple universes an interpretation. In any case, the whole idea of an interpretation of quantum mechanics is an invalid one. It's an idea made possible by not sticking to physics as describing, in mathematical terms, the behavior of nature. There is nothing to interpret in quantum mechanics. Nobody talks about the interpretation of Newton's laws or the interpretation of Maxwell's equations. Nobody does this because it is understood that such a thing would be nonsensical. Any aspects of a theory which are designed to interpret quantum mechanics, be it the Copenhagen interpretation, the Broglie/Bohms guided wave theory or the many worlds interpretation, are invalid. Only those parts of those theories that describe and predict observations are true theories of physics and if various theories do not differ in their predictions they are not fundamentally different.
This doesn't mean that Deutsch's theory of multiple universes (which is based on a theory first proposed by Hugh Everett in 1957) can be dismissed out of hand. It may be a true theory rather than just a semi-philosophical or religious idea or interpretation. His critics have the responsibility to show where his reasoning goes wrong. I make an attempt at this is my article The Parallel Universes of David Deutsch. Suffice it to say here that his "chain of reasoning which rules out the possibility that the universe we see around us constitutes the whole of reality" contains holes the size of canon balls.
In my article I respond to Deutsch's challenge to explain how Shor's algorithm works based on a single-universe view. I answer his question how, within a single universe, a quantum computer can factorize a number requiring more than 10500 calculations when there are only about 1080 atoms in the universe. I lay down the requirements for an experiment which is able to test the multiple universe theory and I show that the experiments based upon which Deutsch accepts the multiple universe theory do not meet these requirements. Furthermore, I present my own challenge to Deutsch, whereby I propose an experiment which is able to prove or disprove his multiple universe theory. My challenge is to do the experiment or some other similar experiment.
Let me give another example of how Deutsch's philosophy leads him astray. Deutsch writes that in the nineteenth century people thought there was such a thing as a gravitational force. Everyone was confident they felt gravity pulling them to the earth and it was also consistent with Newton's theories. Deutsch then writes "Today we understand gravity through Einstein's theory rather than Newton's, and we know that no such force exists."
First, this isn't true. It's a common misconception that somehow Newton's laws became outdated with the acceptance of the more accurate general theory of relativity. Most people who have a scientific understanding of gravity understand gravity through Newton's theory of gravity rather than Einstein's. The reason is that Newton's theory is simpler to apply but accurate enough for most purposes and therefore in most cases the theory of choice for engineers and applied physicists, etc.
Second, because Deutsch does not consider our observations the primary driving force of our theories, but instead considers our theories the primary factor, this makes possible the situation of theories starting to lead a life of their own. What happens is that the model (theory) is put on a higher level than our observations, which is putting the horse behind the cart. Then the model is used in turn to argue that what we observe is not what we think we are observing, even though all our theories of physics are based on our observations in the first place. So, for example, Einstein's theory of gravity, which is in part driven by our observation that we are pulled to the earth, is now used to argue that people used to be wrong in their belief that a force is pulling them to the earth. Now it may well be that in Einstein's theory gravity is not defined as a force, but rather as a curving of space-time. But that's irrelevant. What constitutes a force is a matter of definition and doesn't change our observation that we are pulled to the earth.
One sees a similar thing when people dogmatically deny the existence of a centrifugal force. Of course it's true that in Newton's system there is no centrifugal force, so in that sense this is correct. But on the other hand there is nothing wrong with using the term force for an apparent force in a rotating coordinate system if this helps us understand the situation. In fact there is an accepted method of doing mechanics whereby one indeed makes use of apparent forces such as the centrifugal force and the coriolis force.
Being somewhat of a fan of the philosopher George Berkeley I feel compelled to correct a mistake in Deutsch's presentation of him. Deutsch portrays Berkeley as a defender of the theory of solipsism and and as someone who denied the existences of things. Both claims are incorrect. Berkeley was neither a solipsist nor did he deny the existence of reality. Arguably, Berkeley is the most misrepresented philosopher of history. What Berkeley did argue was that there is no such thing as material or matter. Most people take this to mean he denied the existence of things. But Berkeley wasadamantt at insisting that things do indeed exist. Only his theory was that things and reality exist in the form of our perceptions rather than in the form of matter and he provides an interesting argument for it.
Now one may believe that it is inconsistent to deny the existence of matter and not deny the existence of things. But even if that belief is correct, it is still inaccurate to depict Berkeley as having denied the existence of things. That would be just as wrong as to conclude that Berkeley believed in the existence of matter, since he also believed in the existence of things and given the latter it would be inconsistent for Berkeley to have denied the existence of matter. Although I tend to disagree with Berkeley's view that matter doesn't exist I do think his book "Three Dialogues between Hylas and Philonous" (published in 1713) is a beautiful critique of those who, like Deutsch, search for internal qualities of reality beyond what is required to account for our perceptions.
There is much Deutsch writes which I don't agree with. That by itself is not so strange, but the amount of disagreement I have with this book is larger than that with many other books I've read. It's unavoidable when one writes a book, especially when about controversial ideas, that there won't be many things that will be criticized. And it's also unavoidable that various things will be rightly criticized, unless one is perfect, which few of us are. Of course the statement that I disagree with much of Deutsch's reasoning is not by itself a criticism. I am merely communicating to the reader a personal impression of the book, for what it's worth. To provide a full critique of the entire book would require the writing of a book the size of Deutsch's, something I'm not going to do. Even though I tend to think, as every writer would, that all my criticisms are valid, from my own imperfection it follows that I am probably wrong about that. On the other hand, some (philosophical) points of disagreement might not be a question of valid or invalid, but rather a question of from what angle one chooses to look at an issue.
Deutsch seems to have a tendency toward a contrived line of reasoning. And Deutsch is quite good at reasoning. While this sometimes results is something good it sometimes results in something questionable. Let me give an example of the latter. (Note to reader: you may choose to skip this section. It consists of a detailed analysis of a small part of the book while it takes up half the space of this article.)
In chapter 6 Deutsch attempts to prove that there are what he calls Cantgotu environments. A Cantgotu environment is a virtual reality environment which is logically possible but which can't be generated by any universal virtual reality computer. The name Cantgotu is chosen by Deutsch partly in honor of Cantor, Gödel and Turing and partly because you can't go to it. It is not clear to me why Deutsch discusses Cantgotu environments and what their relation is to the rest of the book.
My first thought when reading about Cantgotu environments is that it doesn't make sense. A flexible enough computer should be able to simulate any kind of environment which we may think of. All the computer does is provide us with a chain of inputs to our senses in interaction with things we do, such as move, talk, etc. Such machines are currently relatively primitive, but there's no reason why a certain sequence of computer generated outputs based on a certain sequence of inputs (our behavior) would in theory not be possible. As long as the machine is fast enough and programmed the right way it should be able to generate any series of sensory inputs we desire.
Let's review Deutsch's proof. Deutsch uses the so-called diagonal argument in his proof, a method well known in mathematics and computer science. Suppose there is a sophisticated computer that can be programmed to render any possible virtual reality environment depending on the program it runs. Each possible program consists of a sequence of symbols in some programming language. We now image an infinitely long numbered list of all possible programs: program 1, program 2, etc. The programs might be alphabetized and then numbered consecutively. First on the list are all 1-symbol long programs, in alphabetical order, then come all 2-symbol long programs in alphabetical order, etc. Since there is no maximum length of a program, the list is infinitely long. Presumably Deutsch means by a list of "all possible programs" all programs which are syntactically correct in the computer language used.
Now we can say that program 1 renders environment 1, program 2 renders environment 2, etc. Each virtual reality environment rendered is assumed to go on for an infinite period of time. Even though some programs might stop after a certain period, we interpret their lack of output thereafter to be an infinite period of a sensory deprivation experiment, where nothing happens. Some environments might be the same, even though their programs are different. For example, environment 5 could be the same as environment 34, or whatever. Now Deutsch imagines that we look at the environments by dividing them into one minute long segments. We consider the first minute, the second minute, etc. Consider a virtual reality environment with the following characteristic. In the first minute it behaves differently from environment 1, in the second minute it behaves differently from environment 2, in the third minute it behaves differently from environment 3, etc.
The question now arises whether a program that produces such an environment is on our list. It cannot be program 1, because the environment must be different from the output of program 1 during the first minute. It cannot be program 2, because the environment must be different from the output of program 2 during the second minute, etc. So, the reasoning goes, the computer program needed to generate such an environment is not present on this list. Since the list contains all possible programs such a program does not exist. So such an environment cannot be created and therefore does not exist.
I find Deutsch's proof a bit lacking in mathematical rigorousness, allthough there would be some argument for that, since the book is not only aimed at mathematically sophisticated people. Here are two omissions in Deutsch's proof:
Similarly, an environment defined by being different from environment 1 during the first minute, different from environment 2 during the second minute, etc., might not be a logically possible environment. For Deutsch's proof to continue he must first prove that there is at least one such environment that is logically possible.
These two problems are not of great importance, because they can be easily solved by framing the proof in a different way. So I'll proceed to analyze other aspects of the proof. I do want to note that I think that Deutsch's proof and my analysis are based on the assumption of deterministic virtual reality programs. If we would allow for virtual reality programs to depend on random results, such as quantum experiments, the analysis might be different and more complicated.
A flaw in Deutsch's proof, at least given the way it is stated, is that the definition given for the example of the Cantgotu environment is imprecise. One of the things computer scientist Alan Turing studied was the so-called halting problem. A characteristic of any computer program is whether it will stop at some point, the program being finished, or whether it will go on forever without stopping. Turing proved that there is no generic program that can determine for each possible program whether or not it will stop.
The problem with some of the programs for our virtual reality computer will also be that at some point, after having rendered a virtual reality up to that point, the program doesn't stop during its next time step. That is, it continues to run forever without ever generating the output for the next time instance. If we are considering the n-th program on the list and if the program gets into this infinite cycle before or during the n-th minute, then the environment it creates in the n-th minute is undefined, because it is never rendered. And so the Cantgotu environment we are imagining is undefined in the n-th minute, since it is defined as something different from something which is undefined. The Cantgotu environment we're talking about is therefore not a logically possible environment and doesn't prove what Deutsch seeks to prove, namely that there are logically possible environments which our machine cannot render.
One way out of this might be to build an extra control into our machine, which somehow forces a program to move on to the next time instant if it hasn't been able to come up with any output for that instant in, say, 1 billion computations. If that happens, we assume the program isn't going to halt for that time step, so we move it on and let its output be nothing (sensory deprivation) for that step. The problem with this solution is that it takes away the proof of the diagonal argument. That would still show that our Cantgotu environment is not on the list, but the program for it might still exist, because it might contain a step with more than 1 billion computations in which case it doesn't belong on the list. A program not being on the list is no longer an argument for the program not existing, because the list no longer contains all possible programs.
Another way out for Deutsch may be to adapt the list to include only all programs which do continue over all time steps, or halt permanently at some time, and never get caught in an infinite processing routine. If, say, program 5 does not conform to this, then we leave out program 5 and rename program number 6 as program 5, etc. I'll come back to this scenario later.
Deutsch makes the following claim: "No physically possible virtual reality generator can render an environment in which answers to noncomputable questions are provided to the user on demand. Such environments are of the Cantgotu type." This statement is true if we understand "on demand" to mean that one would be able to ask an infinite number of noncomputable questions. If we are allowed only a finite number of noncomputable questions then the statement becomes false. Deutsch gives the example of the problem of prime pairs. A prime pair is a pair of prime numbers whose difference is 2, such as 3 and 5 or 11 and 13. A mathematical problem, which has not yet been solved, is whether there are an infinite number of such pairs or only a finite number. The answer may be noncomputable. Suppose the answer is noncomputable. And suppose we want to make a very simple virtual reality machine. One that outputs a single 1 if there exists an infinite number of prime pairs and a single 0 if there exist only a finite number of prime pairs.
Let's imagine a virtual reality program A which outputs a single 1 in all cases. And imagine a virtual reality program B which outputs a single 0 in all cases. Such simple programs are surely possible. Now if we run program A and next run program B, then we will have created the Cantgotu environment we attempted to create. We do not know the answer to the prime pair problem. But we can be sure the answer is either that an infinite number of prime pairs exist, in which case program A has correctly rendered the virtual reality we defined, or that only a finite number of prime pairs exist, in which case program B has correctly rendered the virtual reality we defined. The fact that we do not know whether the correct environment is environment A or B only demonstrates a lack of knowledge on our part, the fact remains that the program sought does exist and its virtual reality can be rendered.
What I think this example shows is that there is something odd about Deutsch's method. He takes a certain feature of computability theory, concerning unsolvable problems, and then translates that into a convoluted definition of a virtual reality which he then claims cannot be rendered. And he presents that to us as a limiting quality of virtual reality machines. But I would say these limitations are strictly speaking not genuine features of our virtual reality machine, but rather are aspects of computability theory. The virtual reality machine itself has no special limitation concerning what kind of sequences of outputs are possible. It's only the case that the desired output must be well-defined. And one could question whether definitions based on incomputable questions are valid definitions. Even if they are considered valid we can still generate all logically possible environments as long as the number of noncomputable questions involved is finite, as demonstrated above. We just won't know which specific environment conforms to which specific definition. That tells us something about our limitation of knowledge, it doesn't demonstrate a limitation of the virtual reality machine.
Another contrived aspect of Deutsch's proof is that it relies on considering virtual reality environments of infinite length, but demands each program (like the program of the Cantgotu environment defined by the diagonal) to be of finite length. When Deutsch discusses how it would feel like to be in a Cantgotu environment, he admits that for any given length of time spent in the Cantgotu environment there will be a finite program that could have rendered it. With every further time period spent beyond the length for which the program was made, there is always a larger more complex program that can render the next minute of the Cantgotu environment. If this is true, it seems to me, here is another reason why Deutsch's Cantgotu environments are in fact capable of being rendered. If any environment lasting any given length of time can be rendered by a finite program, then it appears that all conceivable environments can be rendered, and thus there are no Cantgotu environments. The problem only arises when on demands rendering of a certain type of environment lasting an infinite amount of time. But an environment lasting an infinite amount of time rather than a finite time of any given length is a strange type of virtual reality. Perhaps this is as unreasonable to expect as the calculation of an environment over an infinite space, or with infinite precision.
Let's return to the scenario where the list of programs used in the diagonal argument is adapted so that programs which go into infinite processing during any time step are left out, and the remaining programs are renumbered consecutively, starting at 1. In order to calculate the list of programs we would need some algorithm which determines whether a program belongs on the list or not. So we need an algorithm that can determine whether a program halts during each time step, so that it can continue to calculate the response for the next time step. As Turing proved, there is no solution to the halting problem. No general algorithm exists that can determine for every given program whether it will halt at some point or continue forever. Although what we are considering here is not exactly the same, I think we can also assume there is no program that can determine whether or not any given virtual reality program halts during all time steps.
So the characteristics of the virtual reality machine of the diagonal, in this example, depend on which programs belong on the list, which is an unsolvable problem. In other words, the apparent Cantgotu environment we are considering is defined on the basis of the answer to an unanswerable question, as in the example of the prime pairs. Again, the argument could be made that this is an improper definition of a virtual reality environment and demonstrates not that a certain environment cannot be rendered, but rather demonstrates a lack of knowledge on our part.
One could just as well define a virtual reality based on any one of a multitude of facts which we do not know. For example, we could define an environment that at some point should output a 0 if Napoleon said an even number of words on the 1st of January 1800 and a 1 otherwise. Or we could base it on the answer to the question of whether the room in which the virtual reality machine is situated contains an even or an odd number of air molecules, or whatever. Or we could base the environment on future actions of the user, which cannot be known. But these are strange definitions of a virtual reality environment. A proper definition of a virtual reality environment would specify the environment in terms of specific rules which describe some kind of relationship between the input and the output, rather than describing the environment in terms of certain facts of the external world. A proper definition would be to ask for a rendering of Napoleon, for example, so that he is portrayed as the real Napoleon might have behaved, rather than as he actually did behave, which may be unknowable.
Deutsch is right that Cantgotu environments can be said to exist if we allow an environment to be defined on the basis of incomputable answers plus we accept the requirement that an environment is not considered rendered until we build a machine that can render the environment for an infinite amount of time. The latter qualification is necessary, because if we accept an environment as valid if it can be rendered during any given finite length of time, then it is clear that all possible environments can be rendered. If the time specified contains t time steps and each time step contains n possible outputs, then if we build and run nt virtual reality machines, one for each permutation of output possibilities, then we will have rendered all logically possible environments, even those with improper definitions based on incomputable problems or whatever. We just won't know which environment corresponds to which definition, but we can be sure that each logically possible environment has been rendered, for the time period specified.
It may be argued that creating a machine for each possible permutation of outputs does not really create each possible environment. An environment could be viewed as a system which provides the correct outputs for every possible action the individual in the virtual reality takes. A machine which provides a fixed output does not truly interact with the user. But this problem can be solved by building an even larger number of machines. Instead of building a machine for every possible output, we build a machine for every possible output as a function of its input. What is called output here is the sensory input for the user. And the actions of the user form the input of the machine. Let's assume that for every time step there are x possible user actions (for some time steps there may actually be more possible actions than for other time steps, but this doesn't change the thrust of the argument following). The duration of the virtual reality contains t time steps and each time step contains n possible machine outputs. After time step 1 there are nx classes of machines, since for every of the x possible inputs there are n possible outputs. So there are nx ways of defining an output for every possible input. Since the output for time step 2 can depend both on the input for time step 1 and the input for time step 2, each machine now has to define an output for each of the x2 possible combinations of input for time step 1 and time step 2. So after time step 2 each class of machines is further divided into nx2 classes. The number of machines possible after 2 time steps is found by multiplying nx by nx2. Accordingly the number of possible machines M after t time steps is:
M=nx.nx2.nx3......nxt=ntx(1+2+...+t)=ntx(t+1)t/2
Since this is a finite number, my argument that all logically possible virtual reality machines for all possible environments can be created for any given finite duration, still stands.
Actually, if Deutsch's theory of multiple universes is correct, then there is a sense in which even all environments based on improper definitions can be rendered during an infinite period of time. For every next bit of output by the machine, we simply let a quantum experiment determine the outcome. For example, we let a photon collide with a half mirrored plate. If it passes through the next bit will be a 1, if it reflects it will be a 0. According to Deutsch for each such instant two universes exist, one universe where the machine outputs a 1 and one where it outputs a 0. So one can assume that at every time in the future there will be a universe for every permutation of 1's and 0's. So for any given definition of an environment at any time there will exist at least 1 universe which will have correctly rendered the environment up till that time. And, presumably, there will always be one universe in which it is correct to say that its virtual reality machine will correctly render the defined virtual reality for ever.
In conclusion, there is a sense in which there are such things as Cantgotu environments. But only if one accepts rather strange definitions of environments.
I find Deutsch's discussion in chapter 11 on the nature of time very interesting. He presents a good argument that the common perception of time flowing (from the present in the direction of the future) is wrong.
Deutsch also suggests that there is a symmetry of the laws of physics with respect to time and he appears also to suggest thereby that past and future are not fundamentally different. It is notable that when doing the calculations above for the number of possible virtual reality machines M given x possible inputs and n possible outputs for each one of t time steps, the possible output functions at each time step were taken to depend on all possible inputs received up till that time. In other words, the output of a machine was taken to be a function of the past and not of the future. The reason for this is that it is not possible to create a virtual reality machine whose output depends on the future actions of the user. So in virtual reality there is a fundamental difference between past and future. Since it is one of Deutsch's theories that there is a fundamental equivalence between a virtual reality environment and reality itself it would seem logical for him to assume that there is in fact an asymmetry between past an future in reality, just as there is in virtual reality. Or, perhaps we could say there is a direction of time, not in the sense of time flowing in a certain direction, but in the sense of there being a direction in the ordering of what Deutsch calls time snap shots.
When one writes a book review, at least when I write one, there is a tendency to focus on the things one wishes to criticize. One typically has much more to say about things one disagrees with than about things one agrees with. In any case, I want to end this review by noting some things about Deutsch's book which I found good. This is not a complete list of things about the book I thought were good, just as my criticisms are not a complete list of things about the book I thought were bad. Actually, agreeing with a part doesn't necessarily mean I think that part is good. It might still be boring or uninteresting, even if true. Conversely, there are parts of the book which I think were good because they are interesting, thought provoking or well-written, even if I don't agree with them.
I found Deutsch's argument in chapter 1 about the invalidity of reductionism very good. He discusses what he calls emergent phenomena about which there are things to understand which cannot be explained based on lower level theories. He illustrates this very nicely with the example of how to explain how a statue of Winston Churchill came to be built. He shows that to understand that we need to study higher level subjects, such as the subject of history. The lower level laws of physics, while perhaps able to predict and describe the formation of bronze molecules into a statue, cannot provide us with a true understanding of why the statue was made.
Although I think chapter 2 (about the multiple universe theory), together with chapters 10 (The Nature of Mathematics) and 14 (The Ends of the Universe), forms the most scientifically and philosophically confused part of the book, I did find this chapter amusing and interesting to read. There is a kind of interesting and provocative aspect to the way Deutsch attempts to reason his way into the multiverse world, even if I don't buy it. By the way, why is chapter 14 called "The Ends of the Universe" instead of "The Ends of the Multiverse" given Deutsch's multiple universe theory?
In chapters 3, 4 and 7 I found Deutsch's discussion of his philosophy of science, which focuses on explanations, and his critiques of instrumentalism and inductivism quite worthwhile, even though broadly speaking I disagree with his view. I also think his argument against solipsism is quite good. Deutsch's critique in Chapter 13 of Thomas Kuhn's theory on the progress of science is also very good. Kuhn argues that science is much determined by cultural, political and other subjective factors. Deutsch counters this by illustrating from his own experience how science actually progresses based on an open mind for new theories, objective criticism, rational dialogue and the survival of the best theories.
I like Deutsch's argument in chapter 8 where he suggests that intelligent life might have a very grand effect on the universe, and therefore cannot be said to be insignificant in terms of the large scale development of the universe. His argument is that with significant technology intelligent life might be able to do great things, such as change the life cycle of stars, or whatever.
I found Deutsch's demonstration in chapter 12 of how a multiple universe theory can solve the paradox of time travel interesting, even though I don't agree with the multiverse theory.
And last but not least, what I found good about the book was the fact that every chapter is ended with a short overview and description of the most important (technical) terms used plus a short summary. I consider this valuable, especially the summaries, because they provide the reader with a tool to help him assimilate what he's read. Perhaps every book should have summaries at the ends of all chapters.
Book reviewed: David Deutsch, "The Fabric of Reality", First Edition, Penguin Books, London 1997.
Fabric of Reality Discussion List
A review by Bryce DeWitt
A review by regehr.org
A review by Christopher Maloney
A review by 2think.org
A review by David Park
A review by Dr. Zetie and Mr. Adams
Reviews on Amazon.com
Email: henry@sturman.net