Leonard Susskind
Lee Smolin


Recently, I received a copy of an email sent by Leonard Susskind to a group of physicists which included an attached file entitled "Answer to Smolin". This was the opening salvo of an intense email exchange between Susskind and Smolin concerning Smolin's argument that "the Anthropic Principle (AP) cannot yield any falsifiable predictions, and therefore cannot be a part of science".

After reading several postings by each of the physicists, I asked each if (a) they would consider posting the comments on Edge, and (b) if they would write a new, and final "letter".

Both agreed, but only after a negotiation: (1) No more than 1 letter each; (2) Neither sees the other's letter in advance; (3) No changes after the fact. A physics shoot-out.

While this is a conversation written by physicists for physicists, it should nonetheless be of interest for Edge readers as it's in the context of previous Edge features with the authors, it's instructive as to how science is done, and it's a debate that clarifies, not detracts. And finally it's a good example of what Edge is all about, where contributors share the boundaries of their knowledge and experience with each other and respond to challenges, comments, criticisms, and insights. The constant shifting of metaphors, the intensity with which we advance our ideas to each other — this is what intellectuals do. Edge draws attention to the larger context of intellectual life.

Below are the original email pieces, followed by the final letters presented side-by-side.


LEE SMOLIN, a theoretical physicist, is concerned with quantum gravity, "the name we give to the theory that unifies all the physics now under construction." More specifically, he is a co-inventor of an approach called loop quantum gravity. In 2001, he became a founding member and research physicist of the Perimeter Institute for Theoretical Physics, in Waterloo, Canada. He is the author of The Life of The Cosmos and Three Roads to Quantum Gravity. (See Edge Bio Page)

LEONARD SUSSKIND, the discoverer of string theory, is the Felix Bloch Professor in theoretical physics at Stanford University. His contributions to physics include the discovery of string theory, the string theory of black hole entropy, the principle of "black hole complementarity", the holographic principle, the matrix description of M-theory, the introduction of holographic entropy bounds in cosmology, the idea of an anthropic string theory "landscape". (See Edge Bio Page).

Edge Links:
"Loop Quantum Gravity: A Talk With Lee Smolin"
"The Landscape: A Talk With Leonard Susskind"

[13,165 words]

July 26, 2004

Smolin's publication of "Scientific alternatives to the anthropic principle" and his summary of the arguments, written at Susskind's request.

Lee Smolin published a paper hep-th/0407213 ["Scientific alternatives to the anthropic principle"]. He emailed Leonard Susskind asking for a comment, Not having had a chance to read the paper, Susskiind asked Smolin if he would summarize the arguments. Here is Smolin's message:

Dear Lenny,

Thanks. Glad to. I'll start with one of the main arguments I make.

I show that the argument of Weinberg and others [Garriga and Vilenkin] are incorrect. The subtle point is that their arguments have embedded in them correct arguments having to do only with what we observe. To an already correct argument is then added mention of the anthropic principle. As it is added to an already correct argument, the anthropic principle plays no role in the actual scientific argument.

Here is how it goes: We start with a theory of structure formation that tells us

"Too large positive Lanbda interferes with galaxy formation.

We do observe that galaxies have formed. Therefore we predict that the cosmological constant could not have been too large.

This is correct reasoning, and it agrees with observation.

What Weinberg and others have done is to make the error of embedding this argument into one that first mentions some version of the antrhopic principle. The mistake is not to notice that because it has been added to an argument that is already correct, the mention of the anthropic principle [or the principle of mediocrity, or life] plays absolutely no role in the argument.

The logic of their arguments is

A implies B B is observed B, together with theory C implies D.


A is any form of the Anthropic Principle of Principle of Mediocrity, together with assumptions about priors, proabability distrbutions on universes etc, plus our own existence, that leads to the conclusion that we should observe B.

B is that galaxies have formed.

C is the theory of structure formation,

D is that the cosmological constant is not too large.

The fallacy is not to recognize that the first line plays no role in the argument, and the prediction of D is equally strong if it is dropped. One can prove this by noting that if D were not seen, one would have to question the theory C [assuming the observation is correct, as it certainly is here.] One would have no reason to question either A or the assertion that A implies B.

This is the same fallacy involved in Hoyle's argument about carbon. He reasoned simply from an observation that carbon is plentiful in our universe to a prediction that, as it must have been formed in stars, there must be a resonance at a particular energy. This was correct and the resonance was observed. But he fallaciously attributed the argument to the existence of life, which was a non-sequitor.

In the paper I show that every use of the anthropic principle claimed in physics and cosmology is either an example of this fallacy, or is so vague that one can get any conclusion one wants, and match any observation, by manipulating the assumptions made.

I then go on to explain what a multiverse theory would have to do to yield genuine falsifiable predictions that actually depend on the existence of the multiverse. I give conditions for this to work. I then show that there exist real theories that satisfy these conditions, one of them being my old natural selection idea.

Therefore, the anthropic principle cannot help us to do science. But there are ways to do science if we are faced with a multiverse.


July 28, 2004

Susskind's response to Smolin's paper ["Answer to Smolin"], distributed by email to his mail list of physicists

Answer to Smolin

L. Susskind
Department of Physics
Stanford University
Stanford, CA 94305-4060

Smolin's recent criticisms on the anththropic principle are answered.

Yesterday I received an email message from Lee Smolin asking for my comments about a paper he had posted the previous day [1]. Having not had a chance to read the paper I asked if he would summarize the arguments. He was kind enough to do so in an admirably concise and clear summary. (See above).

I took a quick look at the paper just to make sure that there was not more that I might have missed in the summary

I noticed that in the paper Smolin quotes the philosopher, Karl Popper. Personally I don't think these are deep philosophical issues requiring a heavyweight like Popper. Weinberg was just using good old fashioned common sense.

Weinberg's argument, which is clearly stated in his general audience book Dreams of a Final Theory is not that the cosmological constant has to be smaller than the limit from galaxy formation. If it were, then Smolin would be correct: the Anthropic Principle doesn't add much to the observed fact that galaxies exist. What Weinberg argued was that if the Anthropic Principle is correct, then the cosmological constant will probably not be much smaller than the galaxy formation limit. Weinberg was just expressing the common sense opinion that if anthropics is the only reason the cosmological constant is small, then it is unlikely that it will be orders of magnitude smaller than the anthropic limit.

Was it a prediction that could be proved wrong? Most people thought so. Just about everyone I know was certain that the cosmological constant was exactly zero.

Smolin's next argument involves what he calls a scientific alternative to the Anthropic Principle. He argues for a Darwinian natural selection principle. Smolin believes, as I do, that universes can reproduce and give rise to mutated offspring that differ in the values of the constants of nature. He believes the mechanism involves black holes while I believe it involves eternal inflation and Coleman de Luccia bubble nucleation. But either will do.

Smolin further believes that the constants of nature are determined by survival of the fittest: the fittest to reproduce that is. Those properties which lead to the largest rate of reproduction will dominate the population of universes and the overwhelming likelihood is that we live in such a universe. At least that's the argument.

But this logic can lead to ridiculous conclusions. In the case of eternal inflation it would lead to the prediction that our universe has the maximum possible cosmological constant, since the reproduction rate is nothing but the inflation rate.

It can also lead to ridiculous conclusions in a more zoological context. Imagine what a fool Gregor Samsa would have been if he woke up one morning wondering what sort of creature he was: "When I woke this morning, I realized that I must be a bacterium. They are the fastest reproducers and they outnumber everything else by a factor of trillions."

Samsa's error is obvious but I will spell it out anyway. Just the fact that he could even think to ask the question implies that that he is a rare exception and not an average organism. Of course, given that he was very far from average, he might have successfully used a similar logic to conclude that he was unlikely to be the queen of England.

In the same way Smolin incorrectly concludes that we are in the sort of universe which maximizes the rate of reproduction. But the conditions for life are fantastically exceptional and the fact that we are here at all means we are certainly not in an average universe.

Samsa's common sense observation, that he is probably not Queen Victoria, is an example of Vilenkin's Principle of Mediocrity. It doesn't always work but it is good common sense.


[1] Lee Smolin, "Scientific alternatives to the anthropic principle", hep-th/0407213

July 29, 2004

Smolin's response to Susskind's "Smolin's recent criticisms on the anththropic principle are answered", distributed in email to the same list

Dear Lenny and colleagues,

I am grateful to Lenny for taking the time to respond to my paper. I will be as brief as I an in replying, especially as the key points are already presented in detail in my paper hep-th/0407213 ["Scientific alternatives to the anthropic principle"] or in my book, Life of the Cosmos or previous papers on the subject.

For clarity I had in section 5.1.6 identified two arguments in Weinberg's papers. The first is the one I criticized in the summary. Susskind reponds, reasonably, by agreeing, and then raising the second argument. This argument is also criticized in detail in my paper, and it was perhaps a mistake not to include this in the summary I sent to Susskind.

This second argument is based on a version of the AP called the "Principle of Mediocrity" by Garriga and Vilenkin, who have done the most to develop it. Their version states that, "...our civilization is typical in the ensemble of all civilizations in the universe."

This argument is discussed in full in sections 5.1.5 and 5.1.6. There I argue that the mediocrity principle cannot yield falsifiable predictions because it depends on the definition of the ensemble within which our civilization is taken to be typical as well as on assumptions about the probability distribution. I establish this by general argument as well as by reference to specific examples including Weinberg's use of it.

Can this be right if, as Susskind claims, Weinberg's prediction was found to hold? In fact, Weinberg's prediction did not work all that well. In the form that he made it, it led to an expectation of a cosmological constant larger than the observed value. Depending on the ensemble chosen and the assumptions made about the probability distribution, the probability that Lambda be as small as observed ranges between about 10 % and a few parts in ten thousand. In fact, the less probable values are the more reasonable, as they come from an ensemble where Q, the scale of the density fluctuations, is allowed to vary. While I am not an expert here, it appears from a reading of the literature [references in the paper] that to make the probability for the present value as large as 10% one has to assume that Q is frozen and fixed by fundamental theory. It is hard to imagine a theory where the parameters vary but Q does not, as it depends on parameters in the inflaton potential.

But, because there is so much flexibility-and an absence of strict, up or down, falsifiable predictions, anyone who wants to continue to use the AP in this context is free to modify the assumptions about the prior probability distribution to raise the probability for the observed value of the vacuum energy from 10^{-4} to order unity. No one can prove they are wrong to do so-and this is precisely the problem.

I do believe it is important to insist on falsifiability, because it alone prevents we theorists from keeping theories alive indefinitely, by freely adjusting them to match data.

It was worry about the possibility that string theory would lead to the present situation, which Susskind has so ably described in his recent papers, that led me to invent the Cosmological Natural Selection [CNS] idea and to write my first book. My motive, then as now, is to prevent a split in the community of theoretical physicists in which different groups of smart people believe different things, with no recourse to come to consensus by rational argument from the evidence.

The CNS idea was invented, not for itself, but to give an existence proof that shows that the Anthropic Principle can be replaced by a falsifiable theory, that explains everything the AP claims to. The reason I chose the term "landscape" of string theories, is to anticipate the transition to "fitness landscapes", a term that comes from mathematical models that explain why the mechanism of natural selection is falsifiable.

As the theory of CNS is falsifiable it is vulnerable to criticisms of the kind Susskind makes. Let me briefly address them.

The last first. Susskind claims that life is exceptional in the ensemble of universes. This is not true in CNS. The whole point of cosmological natural selection is that it follows the logical schema described in section 5.1.4 and 5.2. There, and in more detail in the book and papers, I show that falsifiable predictions can be gotten from a multiverse theory if the distribution of universes is very different from random. CNS results in a distribution peaked around small regions of the parameter space-so that a typical universe in this distribution is very untypical in any randomly chosen ensemble. I show in detail why falsifiability follows from this. I also show why reproduction through black holes leads to a multiverse in which the conditions for life are common-essentially because some of the conditions life requires-such as plentiful carbon- also boost the formation of stars massive enough to become black holes.

Next, there is a big difference between the explanatory power of the reproduction mechanisms of eternal inflation and black holes bouncing. This stems from the fact that any selection mechanism can only operate to tune parameters that strongly affect the rate of reproduction. Given that standard inflation acts on grand unified scales, the differential reproduction rate due to eternal inflation is only sensitive to the parameters that govern GUT scale physics plus the vacuum energy. Thus, eternal inflation cannot explain the values of any of the low energy parameters such as the masses of the light quarks and leptons. This means it cannot explain why there are long-lived stars or many stable nuclear bound states leading to a complex chemistry.

Reproduction through black holes explains all the puzzles and coincidences of low energy physics because carbon chemistry, long lived stars etc, are essential for the mechanisms that lead to the formation of massive stars-those that become black holes. There is a long list of observed facts this turns out to explain, and a few genuine predictions. These are summarized in the paper and discussed at length in the book.

With regard to cosmological natural selection and the cosmological constant; if both mechanisms of reproduction exist, there is a competition between them that determines Lambda. Eternal inflation favours a larger cosmological contant, as Lenny says. But-by Weinberg's first argument, if Lambda is too big there are no galaxies, hence many fewer black holes. To my knowledge, nobody has attempted to do a detailed analysis including both mechanisms. Vilenkin has pointed out that too small Lambda can hurt black hole production in a single universe, by making mergers of spirals more common. This is briefly discussed in section 6.2, where I propose that the observed value may maximize the production of black holes-but this has not been analyzed in any detail.

Certainly, if the only mechanism of reproduction is eternal inflation, cosmological natural selection is wrong. But we know that there are black holes and we have reasonable theoretical evidence that black hole singularities bounce. I expect that in the next year we may have reliable quantum gravity calculations that will settle the issue, building on the methods Bojowald and collaborators have used to study cosmological bounces. So the consequences of reproduction through black holes seem reasonable to explore. Not only that, we are on firm ground when we do so because star and black hole formation are observed and controlled by known physics and chemistry.

We know much less about eternal inflation. We cannot observe whether it takes place or not, and there is little near term chance to check the theories that lead to it independently, as there are alternative early universe theories-inflationary and not-that agree with all the cosmological data and do not yield reproduction through eternal inflation.

I do not know if CNS is the only way to get a falsifiable multiverse theory, it is just the only way I've been able to think of. As I have been saying for more than ten years, if CNS can be proved wrong, and if people are stimulated to invent more falsifiable theories to explain the observed parameters, this would be all to the good. So far CNS has not been falsified, but I read astro-ph every day looking for the discovery of a very massive neutron star that will disprove it.

I am very glad that Susskind has been able to give these issues much more visibility. But it would be very unfortunate if string theorists finally accept there is an issue with predictability, only to fall for the easy temptation of adopting a strategy towards it that cannot yield falsifiable theories. The problem with non-falsifiable theories is nothing other than that they cannot be proven wrong. If a large body of our colleagues feels comfortable believing a theory that cannot be proved wrong, then the progress of science could get stuck, leading to a situation in which false, but unfalsifiable theories dominate the attention of our field.



July 29, 2004

Susskind paper ["Cosmic natural selection"] on Smolin's theory of Cosmic Natural Selection

Cosmic Natural Selection

L. Susskind
Department of Physics
Stanford University
Stanford, CA 94305-4060

I make a number of comments about Smolin's theory of Cosmic Natural Selection.

In an unpublished note I criticized Smolin's theory of cosmological natural selection [1]which argues that we live in the fittest of all possible universes. By fitness, Smolin means the ability to reproduce. In my criticism I used the example of eternal inflation which is an extremely efficient reproduction mechanism. If Smolin's logic is applied to that example it would lead to the prediction that we live in the universe with the maximum cosmological constant. This is clearly not so.

Smolin proposes that the true mechanism for reproduction is a bouncing black hole ingularity that leads to a new universe behind the horizon of every black hole. Thus Smolin suggests that the laws of nature are determined by maximizing the number of black holes in a universe.

Smolin also argues that it is not obviously wrong that our physical parameters, includ- ng the smallness of the cosmological constant, maximize the black hole formation. To make sense of this idea, one must assume that there is a very dense discretuum of possi- bilities, in other words a rich landscape of the kind that string theory suggests [4][5][6][7].

The detailed astrophysics that goes into Smolin's estimates in extremely complicated– oo complicated for me–but the basic theoretical assumptions that go into the theory can be evaluated, especially in light of what string theory has taught us about the landscape and about black holes.

As I said, there are two mechanisms, eternal inflation and black hole production that can contribute to reproduction, and it is important for Smolin's scenario that black holes dominate. Considering the low density of black holes in our universe and the incredible efficiency of exponential inflation, it seems very hard to believe that black holes win unless eternal inflation is not possible for some reason.

Smolin's argues that we know almost nothing about eternal inflation but we know a great deal about black holes including the fact that they really exist. This is a bit disingenuous. Despite a great deal of serious effort [8] [9], the thing we understand least s the resolution of black hole and cosmic singularities. By contrast, eternal inflation in a alse vacuum is based only on classical gravity and semiclassical Coleman de Luccia bubble nucleation [2][3].

The issue here is not whether the usual phenomenological inflation was of the eternal kind although that is relevant. Eternal inflation taking place in any false vacuum minimum on the landscape would favor [in Smolin's sense] the maximum cosmological constant. But for the sake of argument I will agree to ignore eternal inflation as a reproduction mechanism.

The question of how many black holes are formed is somewhat ambiguous. What if two black holes coalesce to form a single one. Does that count as one black hole or two? Strictly speaking, given that black holes are defined by the global geometry, it is only one black hole. What happens if all the stars in the galaxy eventually fall into the central black hole? That severely diminishes the counting. So we better assume that the bigger the black hole, the more babies it will have. Perhaps one huge black hole spawns more offspring that 1022 stellar black holes.

That raises the question of what exactly is a black hole? One of the deepest lessons that we have learned over the past decade is that there is no fundamental difference between elementary particles and black holes. As repeatedly emphasized by 't Hooft [10][11][12], black holes are the natural extension of the elementary particle spectrum. This is especially clear in string theory where black holes are simply highly excited string states. Does that mean that we should count every particle as a black hole?

Smolin's theory requires not only that black hole singularities bounce but that the parameters such as the cosmological constant suffer only very small changes at the bounce. This I find not credible for a number of reasons. The discretuum of string theory does indeed allow a very dense spectrum of cosmological constants but neighboring vacua on the landscape do not generally have close values of the vacuum energy. A valley is typically surrounded by high mountains, and neighboring valleys are not expected to have similar energies.

Next–the energy density at the bounce is presumably Planckian. Supposing that a bounce makes sense, the new universe starts with Planckian energy density. On the other hand Smolin wants the final value of the vacuum energy density to be very close to the original. It sounds to me like rolling a bowling ball up to the top of a very high mountain and expecting it to roll down, not to the original valley, but to one out of 10120 with almost identical energy. I find that unlikely.

Finally, we have learned some things about black holes over the last decade that even Stephen Hawking agrees with [13]. Black holes do not lose information. The implication [14] is that if there is any kind of universe creation in the interior of the black hole, the quantum state of the offspring is completely unique and can have no memory of the initial state. That would preclude the kind of slow mutation rate envisioned by Smolin.

Smolin seems to think that there is significant evidence that singularity resolution [by bounce] is imminent. Loop quantum gravity, according to him, is on the threshold of accomplishing this. Perhaps it will. But either it will be consistent with information conservation in which case the baby can have no memory of the parent, or it will not. If not it probably means that Loop gravity is inconsistent.


[1] Lee Smolin, Scientific alternatives to the anthropic principle, hep-th/0407213

[2] S. R. Coleman and F. De Luccia, Phys. Rev. D 21, 3305 [1980].

[3] S. K. Blau, E. I. Guendelman and A. H. Guth, "The Dynamics Of False Vacuum Bubbles," Phys. Rev. D 35, 1747 [1987].

[4] Raphael Bousso, Joseph Polchinski, Quantization of Four-form Fluxes and Dynami- cal Neutralization of the Cosmological Constant, hep-th/0004134, JHEP 0006 [2000] 006

[5] Shamit Kachru, Renata Kallosh, Andrei Linde, Sandip P. Trivedi, de Sitter Vacua in String Theory, hep-th/0301240

[6] Leonard Susskind, The Anthropic Landscape of String Theory, hep-th/0302219

[7] Michael R. Douglas, The statistics of string/M vacua, hep-th/0303194, JHEP 0305 [2003] 046 Sujay Ashok, Michael R. Douglas, Counting Flux Vacua, hep-th/0307049 Michael R. Douglas, Bernard Shiffman, Steve Zelditch, Critical points and super- symmetric vacua math.CV/0402326

[8] G. T. Horowitz and J. Polchinski, Phys. Rev. D 66, 103512 [2002] [arXiv:hep-th/0206228].

[9] L. Fidkowski, V. Hubeny, M. Kleban and S. Shenker, "The black hole singularity in AdS/CFT," JHEP 0402, 014 [2004] [arXiv:hep-th/0306170].

[10] G. 't Hooft, "The unification of black holes with ordinary matter," Prepared for Les Houches Summer School on Gravitation and Quantizations, Session 57, Les Houches, France, 5 Jul - 1 Aug 1992

[11] L. Susskind, "Some speculations about black hole entropy in string theory," arXiv:hep-th/9309145.

[12] G. T. Horowitz and J. Polchinski, "A correspondence principle for black holes and strings," Phys. Rev. D 55, 6189 [1997] [arXiv:hep-th/9612146].

[13] New York Times 7/22/04

[14] G. T. Horowitz and J. Maldacena, The black hole final state," JHEP 0402, 008 [2004] [arXiv:hep-th/0310281].

A. Guth and L. Susskind, To be published

Final Letters

Lee Smolin

I am very pleased that Lenny Susskind has taken the time to respond to my paper on the Anthropic Principle (AP) ["Scientific alternatives to the anthropic principle"] and to discuss cosmological natural selection (CNS). Susskind is for me the most inspiring figure of his generation of elementary particle physicists. Indeed, the initial ideas that became loop quantum gravity came from applying to quantum gravity some of what I had learned from his work on gauge theories. And when in the late 1990's I began to work again on string theory, it was because of papers of his describing how special relativity was compatible with string theory.

I was thus extremely pleased when Susskind began arguing for a view of string theory I came to some time ago — that there is not one theory, but a "landscape" of many theories. But I was equally disturbed when he and other string theorists embraced versions of the Anthropic Principle that I had, after a lot of thought, concluded could not be the basis for a successful scientific theory. To see if we could do better, I formulated conditions that would allow a theory based on a landscape to be a real scientific theory. As an example I had invented the CNS idea. This was all described in my book, The Life of the Cosmos.

Susskind's papers on these issues led me to revisit them, to see if anything that had happened since might change my mind. So I undertook a carefully argued paper on the AP and alternatives to it [a]. The dialogue with Lenny began when I sent a note to him, asking whether he might have any response to the arguments in that paper. At first there were some misunderstandings, because Susskind responded only to a summary, and not the full paper. Nevertheless, some important points were raised, although nothing that requires modification of my original paper. This letter is my response to a paper Susskind put out in the course of our dialogue, making certain criticisms of cosmological natural selection (CNS) [b], and is mostly devoted to answering them.

We agree on several important things, among them that fundamental physics likely gives us a landscape of possible theories, while cosmology may give a multiverse containing a vast number of regions like our own universe. We disagree here mainly on one thing: the mechanism of reproduction we believe has been most important in populating the multiverse.

My main point is that string theory will have much more explanatory power if the dominant mode of reproduction is through black holes, as is the case in the original version of CNS. This is the key point I would hope to convince Susskind and his colleagues about, because I am sure that the case they want to make is very much weakened if they rely on the Anthropic Principle (AP) and eternal inflation.

Susskind believes instead that eternal inflation is the mode of reproduction. But suppose that everything Susskind wants to be true about both eternal inflation and the string theory landscape turns out to be true. What is the best thing that could reasonably be expected to happen?

Weinberg, Vilenkin, Linde and others proposed that in this case we might be able to explain the value of the vacuum energy, both during and after inflation. This is because it is the vacuum energy that determines how many universes are made in eternal inflation, and how large each one is.

However, a careful examination exposes two problems. The first is that the methods so far proposed to make predictions in this scenario are either logically flawed or ambiguous, so that the assumptions can be manipulated to get different predictions. This is explained in detail in section 5.1 of my paper. A second piece of bad news is that, even if this can somehow be made to work, you can't expect to explain much more than the vacuum energy. The reason, as I explain in some detail in section 5.1.4, is that a statistical selection mechanism can only act to tune those parameters that strongly influence how many universes get created. As the selection mechanism in eternal inflation involves inflation, which happens at the grand unified scale, the low energy parameters such as the masses of the light quarks and leptons are not going to have much of an effect on how many universes get created.

In order to tune the low energy parameters, there must be a selection mechanism that is differentially sensitive to the parameters of low energy physics. So we can ask, what possible mechanisms are there for production of universes within a multiverse, such that the number of universes made is sensitive to the values of light quark and lepton masses? I asked myself this question when I realized there would be a landscape of string theories.

The only answer I could come up with is reproduction through black holes. It works because a lot of low energy physics and chemistry goes into the astrophysics that determines how many black holes get made.

Susskind complains that this is complicated, but it has to be complicated. The reason is that we are trying to understand a very curious fact, which is that, as noted by the people who invented the anthropic principle, the low energy parameters seem tuned to produce carbon chemistry and long lived stars. This is explained if CNS is true, because the formation of stars massive enough to become black holes depend on there being both carbon and a large hierarchy of stellar lifetimes.

Thus, if you like eternal inflation because it has a chance of explaining the tuning the vacuum energy, you should like cosmological natural selection much more — because it has potentially much more explanatory power. It offers the only chance so far proposed to actually explain from string theory the parameters that govern low energy physics. Also, as I argued in detail in my paper, the selection mechanism in CNS is falsifiable, whereas those proposed for eternal inflation so far are too ambiguous to lead to clean predictions.

Moreover, because the selection mechanism is dominated by known low energy physics and chemistry, we really do know much more about it than about eternal inflation. We know the dynamics, we know the parameters, and we can use relatively well tested astrophysical models to ask what the effect on the number of universes is of small changes in the parameters. None of this is true for inflation, where unfortunately there are a large variety of models which all are in agreement with observation, but which give different predictions concerning eternal inflation.

Of course it is possible that both mechanisms play a role. It might be useful to study this, so far no one has. It is premature to conclude, as Susskind does, that the production of universes by eternal inflation will dominate. Our universe has "only" 1018 black holes, but the total number of universes in CNS is vastly bigger than this, as there must have been a very large number of previous generations for the mechanism to work.

Susskind made a few direct criticisms of CNS, which are easy to answer, as they have been considered earlier.

He raises the question of how many new universes are created per astrophysical black hole. In the initial formulation of CNS I presumed one, but some approximate calculations have suggested that the number could be variable. I discussed this in detail on page 320 of Life of the Cosmos. The reader can see the details there, what I concluded is that if theory predicts that the number of new universes created increases with the mass, by at least the first power of the mass, the theory can easily be disproved. This hasn't happened, but it could, and it is one of the ways CNS could be falsified. This is of course good not bad, for the more vulnerable a theory is to falsification, the better science it is, and the more likely we are to take it seriously if it nonetheless survives.

One of the assumptions of CNS is that the average change in the low energy parameters when a new universe is created is small. Susskind says he doubts this is true in string theory. If Susskind is right then CNS and string theory could not both be true. But I don't share his intuitions about this. I would have to invoke technicalities to explain why, but all that need be said here is that so far there are no calculations detailed enough to decide the issue. But there could be soon, as I mentioned before, using methods developed recently in loop quantum gravity. These methods may help us study what happens to singularities in string theory and may also provide a better framework to understand eternal inflation.

The rest of this note concerns Susskind's comments about black holes. He says, "...we have learned some things about black holes over the last decade that even Stephen Hawking agrees with [13]. Black holes do not lose information." From this he draws the conclusion that "the quantum state of the offspring is completely unique and can have no memory of the initial state. That would preclude the kind of slow mutation rate envisioned by Smolin."

This is the central point, as Susskind is asserting that black holes cannot play the role postulated in CNS, without contradicting the principles of quantum theory and results from string theory. I am sure he is wrong about this. I would like to carefully explain why. This question turns out to rest on key issues in the quantum theory of gravity, which many string theorists, coming from a particle physics background, have insufficiently appreciated.

The discussion about black holes "losing information" concerns processes in which a black hole forms and then evaporates. Hawking had conjectured in 1974 that information about the initial state of the universe is lost when this happens. Susskind and others have long argued that this cannot be true, otherwise the basic laws of quantum physics would break down.

As Hawking initially formulated the problem, the black hole would evaporate completely, leaving a universe identical to the initial one, but with less information. This could indeed be a problem, but this is not the situation now under discussion. The present discussion is about cases in which a black hole singularity has bounced, leading to the creation of a new region of spacetime to the future of where the black hole singularity would have been. In the future there are two big regions of space, the initial one and the new one. If this occurs then some of the information that went into the black hole could end up in the new region of space. It would be "lost" from the point of view of an observer in the original universe, but not "destroyed", for it resides in the new universe or in correlations between measurements in the two universes.

The first point to make is that if this happens it does not contradict the laws of quantum mechanics. Nothing we know about quantum theory forbids a situation in which individual observers do not have access to complete information about the quantum state. Much of quantum information theory and quantum cryptography is about such situations. Generalizations of quantum theory that apply to such situations have been developed and basic properties such as conservation of energy and probability are maintained. Using methods related to those developed in quantum information theory, Markopoulou and collaborators have shown how to formulate quantum cosmology so that it is sensible even if the causal structure is non-trivial so that no observer can have access to all the information necessary to reconstruct the quantum state [c]. Information is never lost — but it is not always accessible to every observer.

So there is nothing to worry about: nothing important from quantum physics [d] is lost if baby universes are created in black holes and some information about the initial state of the universe ends up there.

A second point is that there is good reason to believe that in quantum gravity information accessible to local observers decoheres in any case, because of the lack of an ideal clock. In particle physics time is treated in an ideal manner and the clock is assumed to be outside of the quantum system studied. But when we apply quantum physics to the universe as a whole we cannot assume this: the clock must be part of the system studied. As pointed out independently by Milburn [e] and by Gambini, Porto and Pullin [f], this has consequences for the issue of loss of information. The reason is that quantum mechanical uncertainties come into the reading of the clock — so we cannot know exactly how much physical time is associated with the motion of the clock's hands. So if we ask what the quantum state is when the clock reads a certain time, there will be additional statistical uncertainties which grow with time. (In spite of this, energy and probability are both conserved.) But, as shown by Gambini, Porto and Pullin, even using the best possible clock, these uncertainties will dominate over any loss of information trapped in a black hole. This means that even if information is lost in black hole evaporation, no one could do an experiment with a real physical clock that could show it.

I believe this answers the worries about quantum theory, but I haven't yet addressed Susskind's assertion that "we have learned some things about black holes over the last decadeŠ.Black holes do not lose information."

I've found that to think clearly and objectively about issues in string theory it is necessary to first carefully distinguish conjectures from the actual results. Thus, over the last few years I've taken the time to carefully read the literature and keep track of what has actually been shown about the key conjectures of string theory. The results are described in two papers [g].

In this case, I am afraid it is simply not true that the actual results in string theory — as opposed to so far unproven conjectures — support Susskind's assertions [h].

There are two classes of results relevant for quantum black holes in string theory. One concerns the entropy of very special black holes, which have close to the maximal possible charge or angular momenta for black holes. For this limited class of back holes the results are impressive, but it has not, almost ten years later, been possible to extend them to typical black holes. The black holes that were successfully described by string theory have a property that typical astrophysical black holes do not have — they have positive specific heat. This means that when you put in energy the temperature goes up. But most gravitationally bound systems, and most black holes have the opposite property — you put in energy and they get colder. It appears that the methods used so far in string theory only apply to systems with positive specific heat, therefore no conclusions can be drawn for typical astrophysical black holes.

A second set of results concerns a conjecture by Maldacena. According to it, string theory in a spacetime with negative cosmological constant is conjectured to be equivalent to a certain ordinary quantum system, with no gravity. (That ordinary system is a certain version of what is called a gauge theory, which is a kind of generalization of electromagnetism).

Even if Maldacena's conjecture is true, that is no reason to assume there could not be baby universes where information was kept apart from an observer in the initial universe for a very long, but not infinite, time. This can be accomplished so long as all the different regions eventually come into causal contact so that, if one waits an infinite time, it becomes possible to receive the information that has gone into the baby universes.

But in any case, Maldacena's conjecture has so far not been proven. There is quite a lot of evidence showing there is some relation between the two theories, but all of the results so far are consistent with a far weaker relationship holding between the two theories than the full equivalence Maldacena conjectured. This weaker relationship was originally formulated in a paper by Witten, shortly after the one of Maldacena. Except for a few special cases, which can be explained by special symmetry arguments, all the evidence is consistent with Witten's weaker conjecture. We should here recall a basic principle of logic that when a collection of evidence is explained by two hypotheses, one stronger and one weaker, only the weaker one can be taken to be supported by the evidence.

But Witten's conjecture requires only that there be a partial and approximate correspondence between the two theories. It does not forbid either baby universes or the loss of information by black holes. For example, Witten shows how some black holes can be studied using results in the other theory, but again it turns out that these are atypical black holes with positive specific heat.

This discussion is related to a conjecture called the Holographic Principle (HP), an idea proposed by 't Hooft (and a bit earlier Crane) that Susskind brought into string theory. Susskind proposes a strong form of the HP, which holds that a complete description of a system resides in the degrees of freedom on its boundary. He takes Maldacena's conjecture as a demonstration of it. I believe here also the evidence better supports a weaker form (proposed with Markopoulou) according to which there is a relation between area and information, but no necessity that the boundary has a complete description of its interior [i].

I would urge a similar caution with respect to Susskind's claim, "As repeatedly emphasized by 't HooftŠ black holes are the natural extension of the elementary particle spectrum. This is especially clear in string theory where black holes are simply highly excited string states. Does that mean that we should count every particle as a black hole?"

As I mentioned, the only results in string theory that describe black holes in any detail describe only very atypical black holes. In those cases, they are related — at least by an indirect argument — to states described by string theory, but they are not in fact excitations of strings. They involve instead objects called D-branes. So Susskind must mean by "a highly excited string state" any state of string theory. But in this case the argument has no force as stars, planets and people must also be "highly excited string states". In any case, until there are detailed descriptions of typical black holes in string theory, it is premature to judge whether Susskind and 't Hooft have conjectured correctly.

Susskind attempts to invoke Hawking's authority here, and it is true that Hawking has announced that he has changed his view on this subject. But he has not yet put out a paper, and the transcript of the talk he gave recently doesn't provide enough details to judge how seriously we should take his change of opinion.

Next Susskind refers to a paper by Horowitz and Maldacena, of which he says that "The implication [14] is that if there is any kind of universe creation in the interior of the black hole, the quantum state of the offspring is completely unique and can have no memory of the initial state. That would preclude the kind of slow mutation rate envisioned by Smolin."

I read that paper and had some correspondence with its authors about it; unfortunately Susskind misstates its implications. In fact, that paper does not show that there is no loss of information, it merely assume it and proposes a mechanism — which the authors acknowledge is speculative and not derived from theory — that might explain how it is that information is not lost. They do not show that information going into baby universes is precluded, in fact Maldacena wrote to me that "If black hole singularities really bounce into a second large region, I also think our proposal is false [j]."

Finally, Susskind suggests that loop quantum gravity will be inconsistent unless it agrees with his conjectures about black holes. I should then mention that there are by now sufficient rigorous results (reviewed in [k]) to establish the consistency of the description of quantum geometry given by loop quantum gravity . Whether it applies to nature is an open question, as is what it has to say about black hole singularities, but progress in both directions is steady.

Let me close with something Susskind and I agree about — which I learned from him back in graduate school: an idea called string/gauge duality according to which gauge fields, like those in electromagnetism and QCD, have an equivalent description in terms of extended objects. For Susskind, those extended objects are strings. I believe that may be true at some level of approximation, but the problem is we only know how to make sense of string theory in a context in which the geometry of spacetime is kept classical — giving a background in which the strings move.

But general relativity teaches us that spacetime cannot be fixed, it is as dynamical as any other field. So a quantum theory of gravity must be background independent. We should then ask if there is a version of this duality in which there is no fixed, classical background, so that the geometry of spacetime can be treated completely quantum mechanically? Indeed there is, it is loop quantum gravity. Moreover, a recent uniqueness theorem [l] shows essentially that any consistent background independent version of this duality will be equivalent to loop quantum gravity. For this reason, I believe it is likely that, if string theory is not altogether wrong, sooner or later it will find a more fundamental formulation in the language of loop quantum gravity.

Indeed, what separates us on all these issues is the question of whether the quantum theory of gravity is to be background independent or not. Most string theorists have yet to fully take on board the lesson from Einstein's general theory of relativity; their intuitions about physics are still expressed in terms of things moving in fixed background spacetimes. For example, the view of time evolution that Susskind wants to preserve is tied to the existence of a fixed background. This leads him to propose a version of the holographic principle which can only be formulated in terms of a fixed background. The strong form of Maldacena's conjecture posits that quantum gravity is equivalent to physics on a fixed background. The approaches string theory takes to black holes only succeed partially, because they describe black holes in terms of objects in a fixed background. Eternal inflation is also a background dependent theory, indeed, some of its proponents have seen it as a return to an eternal, static universe.

On the other hand, those who have concentrated on quantum gravity have learned, from loop quantum gravity and other approaches, how to do quantum spacetime physics in a background independent way. After the many successful calculations which have been done, we have gained a new and different intuition about physics, and it leads to different expectations for each of the issues we have been discussing. There is still more to do, but it is clear there need be — and can be — no going back to a pre-Einsteinian view of space and time. Anyone who still wants to approach the problems of physics by discussing how things move in classical background spacetimes — whether those things are strings, branes or whatever — are addressing the past rather than the future of our science.


[a] Lee smolin, Scientific Alternatives to the Anthropic Principle, hep-th/0407213.

[b] Leonard Susskind, Cosmic natural selection, hep-th/0407266

[c] E. Hawkins, F. Markopoulou, H. Sahlmann, Evolution in Quantum Causal Histories,

[d] In particular, global unitarity is automatically present whenever there is a global time coordinate, but need not be if that condition is not met. Quantum information accessible to local observables is propagated in terms of density matrices following rules that conserve energy and probability, because a weaker property, described in terms of completely positive maps, is maintained.

[e] G.J. Milburn, Phys. Rev A44, 5401 (1991).

[f] Rodolfo Gambini, Rafael Porto, Jorge Pullin, Realistic clocks, universal decoherence and the black hole information paradox hep-th/0406260, gr-qc/0402118 and references cited there.

[g] L. Smolin, How far are we from the quantum theory of gravity? , hep-th/0303185; M. Arnsdorf and L. Smolin, The Maldacena conjecture and Rehren duality, hep-th/0106073.

[h] This is one of several key cases in which conjectures, widely believed by string theorists, have not so far been proven by the actual results on the table. Another key unproven conjecture concerns the finiteness of the theory.

[i] F. Markopoulou and L. Smolin, Holography in a quantum spacetime, hepth/9910146; L. Smolin, The strong and the weak holographic principles, hep-th/0003056.

[j] Juan Maldacena, email to me, 1 November 2003, used with permission.

[k] L. Smolin, An invitation to Loop Quantum Gravity, hep-th/0408048.

[l] By Lewandowski, Okolow, Sahlmann and Thiemann, see p. 20 of the previous endnote.

Leonard Susskind

When I was asked if I would be willing to continue a debate with Lee Smolin on the Edge website my initial reaction was to say no. The problem is that the easiest ideas to explain, which sound convincing to a general audience, are not always the best ideas. The unwary layman says to himself, "Yeah, I understand that. Why is this other guy making it so complicated?" Well the answer is that those simple ideas, that sound like you understand them, often have deep technical flaws and the correct ideas can be very difficult to explain. All a person like myself can do is to say, "Trust me. I know what I'm doing and he doesn't. And besides, so-and-so agrees with me." That doesn't make a good impression. It can be a no win situation.

Why did I agree to do it? Partly because I love explaining physics. Mostly—I don't know why. But here goes nothing as they say.

In a nutshell, here is the view of physics and cosmology that Smolin is attacking:

In the remote past the universe inflated to an enormous size, many orders of magnitude bigger than the observed portion that we can see. Most of the universe is behind the cosmic horizon and cannot be directly detected.

The mechanism of inflation leads to a diverse universe; filled with what Alan Guth calls pocket universes (PU's ). We live in one such PU. Some people call this super-universe the "Multiverse." I like the term "Megaverse". This growth and continuous spawning of pocket U's is called, in the trade, eternal inflation.

String theory leads to a stupendously large "Landscape" of possibilities for the local laws of nature in a given pocket. I will call these possibilities "environments". Most environments are very different from our own, and would not permit life: at least, life as we know it.

Combining 1,2 and 3—the universe is a megaverse filled with a tremendously large number of local environments. Most of the volume of the megaverse is absolutely lethal to life. Some small fraction is more hospitable. We live somewhere in that fraction.

That's it.

There are good reasons for believing 1—4 based on a combination of theoretical and experimental physics. In fact I don't know anyone that disagrees with 1. Assumption 2 is not quite a consequence of 1 but its difficult to avoid 2 in conventional inflation theories.

The physics that goes into it is a very familiar application of trustworthy methods in quantum field theory and general relativity. It's called Coleman de Luccia semi-classical tunneling by instantons, based on a very famous paper by the incomparable Sidney Coleman and his collaborator Frank deLuccia. It is the same physics that has been used from the 1930's to explain the decay of radioactive nuclei.

String theorists are split on whether 3 is a good thing or a bad thing but not about whether it is correct. Only one string theorist seriously challenged the technical arguments, and he was wrong. In any case Smolin and I agree about 3. I think we also agree that most of the Landscape is totally lethal to life, at least life of our kind. Finally 4. There's the rub. As far as I am concerned 4 is simply 1+2+3. But Smolin has other ideas and 4 just gets in the way.

Let's suppose for the moment that these 4 points are correct. What then determines our own environment? In other words why do we find ourselves in one kind of PU rather than another? To get an idea of what the issues are, in a more familiar context, lets replace 1—4 with analogous points regarding the ordinary known universe.

1'. The universe is big—about 15 billion light years in radius.

2'. The expansion of the universe led to a huge number of condensed astronomical objects — at minimum 10[23] solar systems.

3'. The laws of gravity, nuclear physics, atomic physics, chemistry thermodynamics allow a very diverse set of possible environments, from the frozen cold of interstellar space to the ferocious heat of stellar interiors, with planets, moons, asteroids and comets somewhere in between. Even among planets the diversity is huge—from Mercury to Pluto.

4'. The universe is filled with these diverse environments, most of which are lethal. But the universe is so big, that statistically speaking, it is very likely that one or more habitable planets exists.

I don't think anyone questions these points. But what is it that decides which kind of environment we live in—the temperature, chemistry and so on? In particular what determines the fact that the temperature of our planet is between freezing and boiling? The answer is that nothing does. There are environments with temperatures ranging from almost absolute zero to trillions of degrees. Nothing, determines the nature of our environment—except for the fact that we are here to ask the question! The temperature is between freezing and boiling because life (at least our kind) requires liquid water. That's it. That's all. There is no other explanation. [1]

This rather pedestrian, common sense logic is sometimes called "The Anthropic Principle." Note that I mean something relatively modest by the A.P. I certainly don't mean that everything about the laws of physics can be determined from the condition that life exists— just those things that turn out to be features of the local environment and are needed to support life.

Let's imagine that the earth was totally cloud bound or that we lived at the bottom of the sea. Some philosopher who didn't like these ideas, might object that our hypotheses 1'—4' are un-falsifiable. He might say that since there is no way to observe these other regions with their hostile environments—not without penetrating the impenetrable veil of clouds—the theory is un-falsifiable. That, according to him, is the worst sin a scientist can commit. He will say, "Science means falsifiability. If a hypothesis can't be proved false it is not science." He might even quote Karl Popper as an authority.

From our perspective we would probably laugh at the poor deluded fellow. The correctness of the idea is obvious and who cares if they can falsify it.

Even worse, he wouldn't even be correct about the falsifiability. Here is a way that the anthropic reasoning might be proved false without penetrating the veil of clouds: Suppose an incredibly accurate measurement of the average temperature of the earth gave the answer (in centigrade) T=50.0000000000000000000000000000
degrees. In other words the temperature was found to be exactly midway between freezing and boiling, to an accuracy of one hundred decimal places. I think we would be justified in thinking that there is something beyond the anthropic principle at work. There is no reason, based on the existence of life, for the temperature to be so symmetrically located between boiling and freezing. So discovering such a temperature would pretty convincingly mean that the existence of life is not the real reason why the temperature is between 0 and 100 degrees.

Smolin's chief criticism of 1—4 is that they are un-falsifiable. But it is not hard to think of ways of falsifying the Anthropic Principle. In particular Weinberg's prediction that if the anthropic principle is true, then the cosmological constant should not be exactly zero, is very similar to the example I just invented. Weinberg attempted to falsify the anthropic principle. He failed. The Anthropic Principle survived. You can read about the details in Weinberg's book Dreams of a Final Theory.

By un-falsifiable Smolin probably means that other pocket universes can never be directly observed because they are behind an impenetrable veil, i.e. the cosmic event horizon. Throughout my long experience as a scientist I have heard un-falsifiability hurled at so many important ideas that I am inclined to think that no idea can have great merit unless it has drawn this criticism. I'll give some examples:

From psychology: You would think that everybody would agree that humans have a hidden emotional life. B.F. Skinner didn't. He was the guru of a scientific movement called behaviorism that dismissed anything that couldn't be directly observed as unscientific. The only valid subject for psychology according to the behaviorist is external behavior. Statements about the emotions or the state of mind of a patient were dismissed as un-falsifiable. Most of us, today, would say that this is a foolish extreme.

From physics: In its early days of the quark theory, it's many opponents dismissed it as un-falsifiable. Quarks are permanently bound together into protons, neutrons and mesons. They can never be separated and examined individually. They are, so to speak, hidden behind a different kind of veil. Most of the physicists who made these claims had their own agendas, and quarks just didn't fit in. But by now, although no single quark has ever been seen in isolation, there is no one who seriously questions the correctness of the quark theory. It is part of the bedrock foundation of modern physics.

Another example is Allan Guth's inflationary theory. In 1980 it seemed impossible to look back to the inflationary era and see direct evidence for the phenomenon. Another impenetrable veil called the "surface of last scattering" prevented any observation of the inflationary process. A lot of us did worry that there might be no good way to test inflation. Some—usually people with competing ideas—claimed that inflation was un-falsifiable and therefore not scientific.

I can imagine the partisans of Lamark criticizing Darwin, "Your theory is un-falsifiable, Charles. You can't go backward in time, through the millions of years over which natural selection acted. All you will ever have is circumstantial evidence and an un-falsifiable hypothesis. By contrast, our Lamarkian theory is scientific because it is falsifiable. All we have to do is create a population that lifts weights in the gym every day for a few hours. After a few generations, their children's muscles will bulge at birth." The Lamarkists were right. The theory is easily falsified—too easily. But that didn't make it better than Darwin's theory.

There are people who argue that the world was created 6000 years ago with all the geological formations, isotope abundances, dinosaur bones, in place. Almost all scientists will point the accusing finger and say "Not falsifiable!" I'm sure that Smolin would agree with them and so would I. But so is the opposite—that the universe was not created this way—un-falsifiable. In fact that is exactly what creationists do say. By the rigid criterion of falsifiability "creation-science" and science-science are equally unscientific. The absurdity of this position will, I hope not be lost on the reader.

Good scientific methodology is not an abstract set of rules dictated by philosophers. It is conditioned by, and determined by, the science itself and the scientists who create the science. What may have constituted scientific proof for a particle physicist of the 1960's—namely the detection of an isolated particle—is inappropriate for a modern quark physicist who can never hope to remove and isolate a quark. Let's not put the cart before the horse. Science is the horse that pulls the cart of philosophy.

In each case that I described—quarks, inflation, Darwinian evolution—the accusers were making the mistake of underestimating human ingenuity. It only took a few years to indirectly test the quark theory with great precision. It took 20 years to do the experiments that confirmed inflation. And it took 100 years or more to decisively test Darwin (Some would even say that it has yet to be tested). The powerful methods that biologists would discover a century later were unimaginable to Darwin and his contemporaries. What people usually mean when they make the accusation of un-falsifiability is that they, themselves, don't have the imagination to figure out how to test the idea. Will it be possible to test eternal inflation and the Landscape? I certainly think so although it may be, as in the case of quarks, that the tests will be less direct, and involve more theory, than some would like.

Finally, I would point out that the accusation of un-falsifiability is being thrown by someone with his own agenda. Smolin has his own theory based on ideas about the interior of black holes. There is of course, absolutely nothing wrong with that, and Smolin is completely candid about it.

Smolin believes (as I and most cosmologists do) that there is a sense, in which the universe, or perhaps I should say a universe, can reproduce, parent universes spawning baby universes. Perhaps, here is a good time to talk about a linguistic point. The word universe was obviously intended to refer to all that exists. It was not a word that was intended to have a plural. But by now, physicists and cosmologists have gotten used to the linguistic discord. Sometimes we mean all that exists, but sometimes we mean an expanding region of space with particular properties. For example we might say that in our universe the electron is lighter than the proton. In some other distant universe the electron is heavier than the proton. Guth's term—pocket universe—may be a better term but it tends to ruin the prose.

Although we agree that some form of universe-reproduction can occur, Smolin and I disagree about the mechanism. Just the ordinary expansion of the universe is a form of reproduction. For example, if the radius of the universe doubles, you can either picture each cubic meter stretching to 8 cubic meters, or you can say that the original cubic meter gave birth to seven children. Inflation is the exponential expansion of space. It can be understood as an exponentially increasing population of regions. Moreover, according to absolutely standard principles, some of the offspring can be environmentally different than the parent. In this sense a population of PU's exponentially reproduces as the universe inflates. The modern idea of eternal inflation is that the universe eternally inflates, endlessly spawning PU's such as our own. The analogy with the tree of life is apt. Any species eventually becomes extinct but the tree keeps on growing by shooting off new branches and twigs. In the same way a given PU will eventually end but eternal inflation goes on. Just as the population of organisms will be numerically dominated by the fastest reproducers (bacteria) the volume of space will be dominated by the most rapidly inflating environment: an environment I might add, that is totally lethal. If eternal inflation is part of the story of the universe, we can conclude that our local environment is by no means typical. The typical region of space will be one with the largest possible cosmological constant. For Smolin's alternative reproduction mechanism to be relevant, eternal inflation must not occur for some unknown reason.

Smolin's picture of reproduction is that it takes place in the interior of black holes. He believes that in the deep interior of every black hole, the dreaded singularity is a source of a new universe that arises out of the infinitely compressed and heated matter as it contracts and (according to Smolin) subsequently rebounds and expands. By this hypothetical mechanism, a new infant universe is created inside the black hole. The idea is that the child can expand and grow into a genuine adult universe, all hidden from the parent, behind the horizon. Moreover the child universe must have different properties than the parent for Smolin's cosmological natural selection to work. Random mutation is a necessary ingredient to natural selection.

Smolin adds one more assumption that follows the biological paradigm. In ordinary biology the child inherits information about the parents' traits through the genetic code, which may be altered by mutation, but only a tiny bit. Smolin must assume that the offspring only differ by very small amounts from the parent. More precisely he assumes that the constants of physics in the offspring universe are almost the same as in the parent universe. Without this assumption natural selection wouldn't work.

And what does this setup select for? As in life, evolution selects for maximal ability to reproduce. This, according to Smolin, means that PU's whose properties maximize the tendency to produce black holes, will dominate the population. So Smolin argues that our laws and constants of nature are tuned to values that maximize black hole production. According to Smolin, no anthropic reasoning is needed and that makes his theory "scientific".

This is an extremely clever idea. You can read about it in one of Smolin's papers that you can find on the net. Open your web browser to That's where physicists publish their work these days. On the General Relativity and Quantum Cosmology archive, look up gr-qc/9404011. That is one of the first papers that Smolin wrote on "Cosmological Natural Selection." The paper, from 1994, is clear and enjoyable to read. But for some reason it hasn't caught on with either physicists or cosmologists. In fact when I went to track down subsequent papers on the subject, to see if new developments had taken place, I found that there were only 11 citations to the paper. Four of them were by Smolin and two others were critical of the idea—one, incorrectly so, in my judgment.

I'm not sure why Smolin's idea didn't attract much attention. I actually think it deserved far more than it got. But I do know why I was skeptical. Two details, one very technical and one not so technical, seem to me to undermine the idea.

The first, not so technical objection: Frankly, I very much doubt that our laws maximize the number of black holes in the universe. In fact the meaning of the number of black holes in a given universe is unclear. Suppose that in our pocket universe, every star collapses to a black hole eventually. Then in the part of the universe that we presently observe, the number of black holes will eventually be about 10[22]. But suppose that as time goes on, all the black holes in a given galaxy eventually fall into a central black hole at the galactic center. Then the final number will be more like the number of galaxies—about 10[11]. Smolin of course prefers the larger number since he wants to argue that our universe has more black holes than any other possible universe. But strictly speaking, according to the rigorous definition of a black hole, the smaller number is the correct one. But let me be generous and use a looser definition of black hole, so that anything that temporarily looked like a black hole, is counted. But with this rule, it is easy to change the laws of physics so that many more black holes would have been present in the past.

If for example, the minute density contrasts in the early universe, which had the un-naturally small numerical value of 10[-5] were not so weak the universe would have been dominated by small black holes. Those black holes might have coalesced into larger black holes, but I said I would be generous and count them all.

Combine the increase of density contrast with an increase in the strength of gravity and a rapid inflation prehistory and you can make stupendous numbers of black holes. In fact if gravity were made as strong as it could reasonably be, every elementary particle (except photons and gravitons) would be a black hole!

I have exactly the opposite opinion from Smolin's. If the universe were dominated by black holes all matter would be sucked in, and life would be completely impossible. It seems clear to me that we live in a surprisingly smooth world remarkably free of the ravenous monsters that would devour life. I take the lack of black holes to be a sign of some anthropic selection.

Now I come to one of those technical objections, which I think is quite damning but which may mean very little to a layman. Smolin's idea is tied to Hawking's old claim that information can fall into a black hole and get trapped behind the horizon. Smolin requires a great deal of information to be transferred from the parent universe to the infant at the bouncing singularity. But the last decade of black hole physics and string theory have told us that NO information can be transferred in this way!

Some readers may recognize the issue that I am talking about. Anyone who has read the recent New York Times article by Dennis Overbye knows that the ultimate fate of information falling into a black hole was the subject of an long debate involving Stephen Hawking, myself, the famous Dutch physicist Gerard 't Hooft and many other well known physicists. Hawking believed that information does disappear behind the horizon, perhaps into a baby universe. This would be consistent with Smolin's idea that offspring universes, inside the black hole, remember at least some of the details of the mother universe. My own view and 't Hooft's was that nothing can be lost from the outside world—not a single bit. Curiously the cosmological debate about Cosmological Natural Selection revolves around the same issues that came to the attention of the press a week or two ago. The occasion for the press coverage was Hawking's recantation. He has reversed his position.

Over the last decade, since Smolin put forward his clever idea, the black hole controversy has largely been resolved. The consensus is that black holes do not lose any information. I'll cite some of the most influential papers that you can look up yourself: HEP-TH 9309145 , HEP-TH 9306069, HEP-TH 9409089, HEP-TH 9610043, HEP-TH 9805114, HEP-TH 9711200. Incidentally, the combined total number of citations for these six papers is close to 6,000. Another paper, co-authored very recently, by the author of one of these classics, directly attacks Smolin's assumption. In fact it was one of the 11 papers that I found citing Smolin's paper. If you want to look it up, here is the archive reference: HEP-TH 0310281. I warned you that I would say "And besides, so-and-so agrees with me." I apologize, but at least you can go check for yourself.

The implication of these papers is that no information about the parent can survive the infinitely violent singularity at the center of a black hole. If such a thing as a baby universe makes any sense at all, the baby will have no special resemblance to the mother. Given that, the idea of an evolutionary history that led, by natural selection, to our universe, makes no sense.

I'm sure there are physicists that are unconvinced by the arguments of the abovementioned papers, despite the number of citations. They have all the right in the world to be skeptical but the average reader of this page should know that these people are swimming against the tide.

Finally let me quote a remark of Smolin's that I find revealing. He says "It was worry about the possibility that string theory would lead to the present situation, which Susskind has so ably described in his recent papers, that led me to invent the Cosmological Natural Selection (CNS) idea and to write my first book. My motive, then as now, is to prevent a split in the community of theoretical physicists in which different groups of smart people believe different things, with no recourse to come to consensus by rational argument from the evidence."

First of all, preventing a "split in the community of theoretical physicists" is an absurdly ridiculous reason for putting forward a scientific hypothesis.

But what I find especially mystifying is Smolin's tendency to set himself up as an arbiter of good and bad science. Among the people who feel that the anthropic principle deserves to be taken seriously, are some very famous physicists and cosmologists with extraordinary histories of scientific accomplishment. They include Steven Weinberg [2], Joseph Polchinski [3], Andrei Linde [4], and Sir Martin Rees [5]. These people are not fools, nor do they need to be told what constitutes good science.


[1] Of course you might say that the distance to the sun determines the temperature. But that just replaces the question with another, "Why is our planet at the precise distance that it is?"

[2] Professor of Physics, University of Texas and Nobel Prize winner 1979.

[3] Professor of Physics, Kavli Institute for Theoretical Phyiscs.

[4] Professor of Physics, Stanford University, Winner of many awards and prizes including the Dirac Medal and Franklin Medal.

[5] Astronomer Royal of Great Britain.


"An engrossing treat of a book...crammed with hugely enjoyable'll have a wonderful time reading these reminiscences." New Scientist

" ...inspiring glimpses into the minds of today’s scientific intelligentsia" Publishers' Weekly

Curious Minds:
How a Child Becomes a Scientist

U.S. (August)

Original essays vy Nicholas Humphrey • David M. Buss • Robert M. Sapolsky • Mihaly Csikszentmihalyi • Murray Gell-Mann • Alison Gopnik • Paul C. W. Davies • Freeman Dyson • Lee Smolin • Steven Pinker • Mary Catherine Bateson • Lynn Margulis • Jaron Lanier • Richard Dawkins • Howard Gardner • Joseph LeDoux • Sherry Turkle • Marc D. Hauser • Ray Kurzweil • Janna Levin • Rodney Brooks • J. Doyne Farmer • Steven Strogatz • Tim White • V. S. Ramachandran • Daniel C. Dennett • Judith Rich Harris • edited, with an introduction by John Brockman

[Click here for table of contents and contributors]

When We Were Kids:
How a Child Becomes a Scientist


"A stellar cast of thinkers tackles the really big questions facing scientists in a book developed from pieces that first appeared on the web forum Edge." — The Guardian

Edge ... "Brilliant!" The Sunday Times "Deep" New Scientist "Thoughtful" Bill Gates: New York Times Column "Stimulating" The Sunday Times (London)"Fascinating" American Scientist "Terrific" Guardian Unlimited "Heady" Kevin Kelly (Editor-At-Large, Wired) "Influential" The Times "Lively" The Independent "A-List" Wired "Inspired" The Sunday Times"Awesome" Wired"Stunning" Arts & Letters Daily"Entertaining" San Jose Mercury News"Wonderful" The Times (London) •"Interesting" The New York Times"Prestigious" Wired News "Marvellous" Prospect

The New Humanists:
Science at the Edge


Jared Diamond • Steven Pinker • Helena Cronin • Andy Clark • Marc D. Hauser • Richard Wrangham • Daniel C. Dennett • Stephen M. Kosslyn • Jordan B. Pollack • David Gelernter • Rodney Brooks • Hans Moravec • David Deutsch • Marvin Minsky • Ray Kurzweil • Jaron Lanier • Seth Lloyd • Alan Guth • Paul Steinhardt • Lisa Randall • Lee Smolin • Martin Rees • edited, with an introduction by John Brockman

[Click here for table of contents and contributors]

Science at the Edge

"A compact, if bumpy, tour through the minds of some of the world's preeminent players in science and technology."
Philadelphia Inquirer

"[Brockman] wisely exits the stage after a brief monologue and turns the rest of the show over to his guests. What a show they put on!" San Jose Mercury News

"Entertaining" New Scientist "Provocative" Daily Telegraph "Inspired" Wired "Mind-stretching" Times Higher Education Supplement "Captivating The Futurist "Great fun" Physics Web "Fascinating" Dallas Morning News "Dazzling" Washington Post Book World "Bedazzled" Publisher's Weekly 

Next Fifty Years: Science in the First Half of the Twenty-First Century


Original essays by Peter Atkins, • Samuel Barondes • Paul Bloom • Rodney Brooks • Mihalyi Csikszentmihalyi • Paul C. W. Davies • Richard Dawkins • Nancy Etcoff • Paul W. Ewald • David Gelernter • Brian Goodwin • Alison Gopnik • Judith Rich Harris • Marc D. Hauser • John H. Holland • Stuart Kauffman • Jaron Lanier • Joseph LeDoux • Geoffrey Miller • Martin Rees • Robert Sapolsky • Roger C. Schank • Lee Smolin • Ian Stewar • , Steven Strogatz • edited, with an introduction by John Brockman

[Click here for table of contents and contributors]

Next Fifty Years: Science in the First Half of the Twenty-First Century