The Trouble With Physics: The Rise of String Theory, The Fall of a Science, and What Comes Next

by Lee Smolin

  I find it unfortunate that Susskind and others have embraced the anthropic principle, because it has been understood for some time that it is a very poor basis for doing science. Since every possible theory governs some part of the multiverse, we can make very few predictions. It is not hard to see why.

  To make a prediction in a theory that posits a vast population of universes satisfying randomly chosen laws, we would first have to write down all the things we know about our own universe. These things would apply to some number of other universes as well, and we can refer to the subset of universes where these facts are true as possibly true universes.

  All we know is that our universe is one of the possible universes. Given that the population of universes was produced by randomly distributing the fundamental laws of nature among them, we can know little else. We can make a new prediction only if every, or almost every, possibly true universe has a property that is not on the list of the properties we’ve already observed in our own universe.

  For example, suppose that in almost every possibly true universe the most resonant oscillation is a low C. Then it is highly probable that a universe picked randomly from the possibly true universes will be resonating at low C. Since we can know nothing about our own universe except that it is a possibly true universe, we can predict with a high probability that our universe is singing a low C, too.

  The problem is that since the distribution of theories over all the universes is assumed to be random, there are very few properties like this. Most likely, once we have specified the properties we observe in our own universe, the remaining properties that any universe might have will be distributed randomly among the other possibly true universes. Thus we can make no predictions.

  What I have been describing is what cosmologists call the weak anthropic principle. As the name indicates, the one thing we know about our universe is that it supports intelligent life; therefore, every possibly true universe must be a place where intelligent life could live. Susskind and others argue that this principle is nothing new. For example, how do we explain the fact that we find ourselves on a planet situated so that the temperature is in the range where water is liquid? If we believed that there was only one planet in the universe, we would find this fact puzzling. We would be tempted to believe in the necessity of an intelligent designer. But once we know that there are a vast number of stars and many planets, we understand that just by chance there will be many planets friendly to life. Therefore we are not surprised to find ourselves on one of them.

  There is, however, a big difference between the planet analogy and the cosmological situation, which is that we do not know of any universes except our own. The existence of a population of other universes is a hypothesis that cannot be confirmed by direct observation; hence, it cannot be used in an explanatory fashion. It is true that if there were a population of universes with random laws, we would not be surprised to find ourselves in one where we can live. But the fact that we are in a biofriendly universe cannot be used as a confirmation of a theory that there is a vast population of universes.

  There is a counterargument, which we can illustrate with the planet example. Let us suppose that it was impossible to observe any other planets. If we deduced from this that there was in fact only one planet, we would be forced to believe in something very improbable, which is that the single planet that exists is biofriendly. On the other hand, if we assume that there are many planets with random properties, even if we can never observe them, then the probability that a few are biofriendly is greatly increased—in fact, it approaches 1. Therefore, it is argued, it is much more probable that there are many planets than that there is only one.

  But this apparently strong argument is fallacious.* To see why, let us compare it to another argument that might be made from the same evidence. Someone who believed in intelligent design could argue that if there is only one planet and it is biofriendly, there is a high probability of an intelligent designer at work. Given a choice between two theories—(1) the unique planet is biofriendly just by extreme luck, and (2) there was an intelligent designer who made the unique planet and made it biofriendly—the same logic leads us to conclude that it is more rational to choose the second alternative.

  The scenario of many unobserved universes plays the same logical role as the scenario of an intelligent designer. Each provides an untestable hypothesis that, if true, makes something improbable seem quite probable.

  Part of the reason these arguments are fallacious is that they rely on an unstated assumption—that we have in hand the complete list of alternatives. For (to return to the planet analogy) we cannot preclude the possibility that a genuine explanation for our planet’s biofriendliness will emerge sometime in the future. The fallacy in the two arguments is that they both compare a single possible—but untestable—explanation with the statement that there is no possible explanation. Of course, given only those two choices, an explanation appears more rational than any unexplained improbability.

  For centuries we have had good reason to believe there are lots of planets, because there are lots of stars—and recently we have confirmed directly the existence of extrasolar planets. So we believe the many-planet explanation for the biofriendliness of our planet. But when it comes to the biofriendliness of our universe, we have at least three possibilities:

  Ours is one of a vast collection of universes with random laws.

  There was an intelligent designer.

  There is a so-far-unknown mechanism that will both explain the biofriendliness of our universe and make testable predictions by which it can be confirmed or falsified.

  Given that the first two possibilities are untestable in principle, it is most rational to hold out for the third possibility. Indeed, that is the only possibility we should consider as scientists, because accepting either of the first two would mean the end of our field.

  Some physicists claim that the weak anthropic principle must be taken seriously because it has led in the past to genuine predictions. I’m talking here about some of the people I most admire—not only Susskind but also Steven Weinberg, the physicist who, you’ll recall from chapter 4, together with Abdus Salam unified the electromagnetic and weak nuclear forces. It then pains me to conclude that in every case I’ve looked into, the claims have proved erroneous.

  Consider, for example, the following argument about the properties of the carbon nuclei, based on investigations made in the 1950s by the great British astrophysicist Fred Hoyle. This argument is often taken as indicating that real physical predictions can be based on the anthropic principle. The argument begins with the observation that for life to exist, there must be carbon. Indeed, carbon is plentiful. We know that it could not have been made in the Big Bang, hence we know that it must have been made in stars. Hoyle noticed that carbon could have been formed in stars only if there was a certain resonant state of carbon nuclei. He communicated this prediction to a group of experimentalists, who found it.

  The success of Hoyle’s prediction is sometimes used as support for the effectiveness of anthropic principle. But the argument from life in the preceding paragraph has no logical relation to the rest of the paragraph. What Hoyle did was to reason from the observation that the universe is full of carbon to a conclusion based on the necessity of there being some process whereby all that carbon got made. The fact that we and other living things are made of carbon is unnecessary to the argument.

  Another example often cited in support of the anthropic principle is a prediction about the cosmological constant made in a celebrated 1987 paper by Steven Weinberg. In it, he pointed out that the cosmological constant must be less than a certain value, otherwise the universe would have expanded too rapidly for galaxies to form.2 Since we observe that the universe is full of galaxies, the cosmological constant must be less than that value. And it is, as it must be. This is perfectly good science. But Weinberg took this valid scientific argument further. Suppose there is a multiverse, he said, and suppose the val
ues of the cosmological constant are randomly distributed among its member universes. Then, among the possibly true universes, a typical value of the cosmological constant would be of the order of magnitude of the largest that is consistent with galaxy formation. Hence, if the multiverse scenario is true, we should expect the cosmological constant to be as large as it can be, while still allowing galaxies to form.

  When Weinberg published this prediction, the cosmological constant was generally believed to be zero. Thus it was impressive that his prediction came true to within roughly a factor of 10. However, when the new results forced Weinberg’s statements to be examined more carefully, some problems emerged. Weinberg had considered a population of universes in which only the cosmological constant is randomly distributed, while all the other parameters are held fixed. Instead, he should have averaged over all the members of the multiverse consistent with galaxy formation, allowing all the parameters to vary. Done this way, the prediction of the cosmological constant’s value turns out to be much further off.

  This illustrates a persistent problem with reasoning of this type. If your scenario invokes randomly distributed parameters, of which you can observe only one set, you can get a wide range of possible predictions, depending on the precise assumptions you make about that unknown, unobservable population of other sets. For example, each of us is a member of many communities. In many of these we will be typical members, but in many others we will be untypical. Suppose, in my author’s bio on the book jacket, all I write is that I am a typical person. How much about me will you be able to deduce?

  There are many other cases in which some version of the weak anthropic principle may be tested. Within the standard model of elementary-particle physics, there are constants that simply don’t have the values we would expect them to have if they were chosen by random distribution among a population of possibly true universes. We would expect that the quark and lepton masses, except for the first generation, would be randomly distributed, but relations between them are seen. We would expect that some symmetries of the elementary particles would be violated by the strong nuclear force much more than they are. We would expect the proton to decay at a much faster rate than present experimental limits allow. In fact, I know of no successful predictions that have been made by reasoning from a multiverse with a random distribution of laws.

  But what about the third possibility, which is an explanation for the biofriendliness of our universe based on testable hypotheses? In 1992 I put a proposal of just this kind on the table. To get testable predictions from a multiverse theory, the population of universes must be far from random. It must be intricately structured so that there are properties that all or most universes have that have nothing to do with our existence. We can then predict that our universe has these properties.

  One way to get such a theory is to mimic the way natural selection works in biology. I invented such a scenario in the late 1980s, when it became clear that string theory would come in a very large number of versions. From books by evolutionary biologists Richard Dawkins and Lynn Margulis, I learned that biologists had models of evolution that were based on a space of possible phenotypes they called fitness landscapes. I adopted the idea and the term and invented a scenario in which universes are born from the interiors of black holes. In The Life of the Cosmos (1997), I reflected at length on the implications of this idea, so I will not go into it in detail here, except to say that that theory, which I called cosmological natural selection, made genuine predictions. In 1992 I published two of them and they have since held up, although they could have been proved false by many observations made since then. These are (1) that there should be no neutron stars more massive than 1.6 times the mass of the sun, and (2) that the spectrum of fluctuations generated by inflation—and, plausibly, observed in the cosmic microwave background—should be consistent with the simplest possible version of inflation, with one parameter and one inflaton field.3

  Susskind, Linde, and others have attacked the idea of cosmological natural selection, because they claim that the multitudes of universes created in eternal inflation will overwhelm any numbers made through black holes. To address this objection, it is important to know how reliable the prediction of eternal inflation is. The case is sometimes made that it is hard to have inflation at all without eternal inflation. The fact that some of the predictions of inflationary cosmology have been confirmed is taken as evidence for it. However, moving from inflation to eternal inflation assumes that there is no barrier to extending conclusions that hold on our present cosmological scale to vastly bigger scales. There are two problems with this: The first is that the extrapolation to bigger scales at the present time implies, in some models of inflation, an extrapolation to much smaller scales in the early universe. (I will not explain this here, but it is true of several inflation models.) This means that to get an inflated universe vastly bigger than our present universe, we must extend the description of the early universe to times vastly smaller than the Planck time, before which quantum gravity effects dominate the evolution of the universe. This is problematic, because the usual description of inflation assumes that spacetime is classical and there are no effects of quantum gravity; moreover, several theories of quantum gravity predict that there is no interval of time shorter than the Planck time. Second, there are indications that the predictions of inflation are not satisfied on the largest scales we can currently observe (see chapter 13). Hence, the extrapolation from inflation to eternal inflation runs into trouble both theoretically and observationally, so this does not seem a strong objection to cosmological natural selection.

  In spite of the fact that the anthropic principle has not led to any real predictions and is not likely to, Susskind, Weinberg, and other leading theorists have embraced it as signaling a revolution not just in physics but in our conception of what a physical theory is. Weinberg asserts in a recent essay,

  Most advances in the history of science have been marked by discoveries about nature, but at certain turning points we have made discoveries about science itself. . . . Now we may be at a new turning point, a radical change in what we accept as a legitimate foundation for a physical theory. . . . The larger the number of possible values of physical parameters provided by the string landscape, the more string theory legitimates anthropic reasoning as a new basis for physical theories: Any scientists who study nature must live in a part of the landscape where physical parameters take values suitable for the appearance of life and its evolution into scientists.4

  Steven Weinberg is justly honored for his contributions to the standard model, and his writings usually advance a compelling and sober rationality. But, simply put, once you reason like this, you lose the ability to subject your theory to the kind of test that the history of science shows over and over again is required to winnow correct theories from beautiful but wrong ones. To do this, a theory must make specific and precise predictions that can either be confirmed or refuted. If there is a high risk of disconfirmation, then confirmation counts for a lot. If there is no risk of either, then there is no way to continue to do science.

  The debate about how science is to confront the newly vast string landscape seems to me to come down to three possibilities:

  String theory is right, and the random multiverse is right, so to accommodate them we must change the rules that govern how science works, because according to the usual scientific ethic, we would not allow ourselves to believe in a theory that made no unique predictions by which it could be confirmed or falsified.

  Some way will eventually be found to deduce genuine unique and testable predictions from string theory. This might be done either by showing that there really is a unique theory or by a different, nonrandom multiverse theory that leads to genuine testable predictions.

  String theory is not the right theory of nature. Nature is best described by another theory, yet to be discovered or yet to be accepted, which does lead to genuine predictions, which experiments will eventually confirm.
<
br />   What is remarkable to me is the number of distinguished scientists who seem unable to accept the possibility either that string theory or the hypothesis of a random multiverse is wrong. Here is a collection of pertinent comments:

  “Anthropic reasoning runs so much against the historic goals of theoretical physics that I resisted it long after realizing its likely necessity. But now I have come out.”

  —JOSEPH POLCHINSKI

  “Those who dislike the anthropic principle are simply in denial.”

  —ANDREI LINDE

  “The possible existence of a huge landscape is a fascinating development in theoretical physics that forces a radical rethinking of many of our assumptions. My gut feeling is that it may well be right.”

  —NIMA ARKANI-HAMED (Harvard University)

  “I think it’s quite plausible that the landscape is real.”

  —MAX TEGMARK (MIT)

  Even Edward Witten seems stumped: “I just don’t have anything incisive to say. I hope we will learn more.”5

  There is not a person quoted here whom I do not deeply admire. Nevertheless, it seems to me that any fair-minded person not irrationally committed to a belief in string theory would see this situation clearly. A theory has failed to make any predictions by which it can be tested, and some of its proponents, rather than admitting that, are seeking leave to change the rules so that their theory will not need to pass the usual tests we impose on scientific ideas.