Defending the multiverse

Abstract

From a historical perspective, the multiverse is just one more step in our progress from geocentric to heliocentric to galactocentric to cosmocentric worldview. Indeed, several lessons of relevance to the multiverse debate can be gleaned from considering the history of this progression. To the ancient Greeks, the heavenly spheres were the unchanging domain of the divine and therefore outside science by definition. It required Tycho de Brahe's observation of a supernova in 1572 and the realization that its apparent position did not change as the Earth moved around the Sun to dash that view. Because this contradicted the Aristotelian view that the heavens cannot change, the claim was at first received sceptically. Frustrated by those who had eyes but would not see, Brahe wrote: “O crassa ingenia. O coecos coeli spectators.” (Oh thick wits. Oh blind watchers of the sky.) Lesson 1: Theoretical prejudice should not blind one to the evidence. Of course, we will never see the other universes themselves — in that sense we are necessarily blind, so this point might seem irrelevant to the multiverse. However, I would claim that the analogue of Tycho's supernova is the fine-tunings. Long after Galileo had speculated that the Milky Way consists of stars like the Sun and Newton had shown the laws of Nature could be extended beyond the solar system, there was still a prejudice that the investigation of this region was beyond the domain of science. In 1842 August Comte said of the study of stars: “Never, by any means, will we be able to study their chemical compositions. The field of positive philosophy lies entirely within the Solar System, the study of the universe being inaccessible in any possible science.” Comte had not foreseen the advent of spectroscopy, which identified absorption features in stellar spectra with chemical elements. Lesson 2: New observational developments are hard to anticipate. Perhaps we will find extra dimensions at the Large Hadron Collider or even create baby universes in the laboratory one day. Cosmology attained the status of a proper science in 1915, when the advent of general relativity gave it a secure mathematical basis. Nevertheless, for a further decade there was resistance to the idea that science could be extended beyond our galaxy. Indeed many astronomers refused to believe that there was anything beyond. Although Kant had speculated as early as 1755 that some nebulae are “island universes” similar to the Milky Way, most astronomers continued to adopt a galactocentric view until the 1920s. Indeed, the most popular model of the galaxy at the start of the 20th century — Kapteyn's Universe — even had the Sun at its centre! The controversy came to a head in 1920 when Heber Curtis defended the island universe theory in a famous debate with Harlow Shapley. The issue was finally resolved in 1924, when Edwin Hubble measured the distance to M31 using Cepheid variable stars. In many ways this parallels the current debate about whether anything exists beyond our horizon. Lesson 3: More conservative cosmologists might prefer to maintain the cosmocentric view but perhaps the tide of history is against them. The evidence for other universes can never be as decisive as that for extragalactic nebulae but the transformation of worldview required may be just as necessary. A few years later Hubble obtained radial velocities and distance estimates for several dozen nearby galaxies, thereby discovering that all galaxies are moving away from us with a speed proportional to their distance. The most natural interpretation of this is that space itself is expanding, as indeed had been predicted by Alexander Friedmann in 1920 on the basis of general relativity. Einstein rejected this model at the time because he believed the universe (i.e. the Milky Way) was static and he even introduced an extra repulsive term into his equations — the cosmological constant — to allow this possibility. After Hubble's discovery, he described this as his “biggest blunder”. Lesson 4: One should not necessarily reject theoretical predictions because they have no observational support. In fact, Einstein continued to uphold the static model even after the evidence was against it — he only accepted the Friedmann model in 1931, several years after Hubble published his data — so knowing how much weight to attach to theory and observation can be tricky. Let me now address George's specific issues. There are plausibly galaxies just beyond the visual horizon, where we cannot see them, so we can extend this argument, step by step, to way beyond the horizon and infer there are many different universes that we cannot see. Even though we can never prove what happens outside our visual horizon, the standard FRW model has been well tested within it, so there is surely some probabilistic sense in which one can extrapolate models at least some way beyond it. Also the smooth dependence of the CMB fluctuations on angular separation (whatever the source of those fluctuations) gives no reason to suppose that anything strange happens just beyond the horizon. George himself seems to accept this, which illustrates the problem of regarding speculations as non-scientific just because they involve the unobservable. Admittedly one's confidence in any proposed model must decrease as one extrapolates ever further beyond the horizon, but one should beware of using Rees's slippery slope argument in reverse: we cannot extrapolate to scales much larger than the horizon, so we should not extrapolate to scales only slightly outside it. The problem comes when one makes the jump from the Level I to Level II multiverse (which is where George's argument that the FRW solution extends everywhere must fail). In fact, the inflationary scenario does provide an answer to this. For if the amplitude of the density fluctuations increases slightly with scale (as appears to be the case), one can predict the scale at which the FRW approximation breaks down. Current data suggest that this happens at around 10100 horizon scales. The existence of a multiverse is implied by inflation, which is verified by the CMB anisotropy observations. In particular, known physics leads to chaotic inflation and this implies a multiverse. There are two distinct issues here: does one believe in inflation and does inflation lead to a multiverse? Inflation is attractive because it resolves several cosmological conundra. Quantum fluctuations of the scalar field can also generate the small density perturbations that eventually give rise to galaxies and large-scale structure and it is impressive that the predicted dependence of the CMB fluctuations on angular separation is almost exactly as observed by the WMAP satellite (Spergel et al. 2003). Of course, the evidence for inflation is not conclusive — there is still no evidence for any scalar field in Nature!— but the Level I multiverse is still a good bet. As regards the second issue, I agree with George that the evidence for the sort of chaotic inflation that leads to a Level II multiverse is more equivocal, and certainly one cannot infer this from the form of the CMB anisotropies. There are now around 100 models of inflation and, while Linde (1990) claims that the existence of other domains with different coupling constants is generic, this is debatable. The multiverse idea is testable, because it can be disproved if we determine there are closed spatial sections in the universe (for example, if the curvature is positive). This is really a straw man argument because we have seen that inflation is only one of several multiverse proposals — for example, quantum cosmology models give closed spatial sections — and not all inflationary models require that the spatial sections be open anyway. However, George is surely right to stress the importance of looking for circles in the CMB. The idea of small universes is not mainstream but it has the advantage that it can be tested. The existence of a multiverse is the only physical explanation for the fine-tuning of parameters that leads to our existence. In the absence of direct evidence for other universes, I regard the anthropic fine-tunings as the best indirect evidence. (A multiverse in which the constants were the same everywhere would have no explanatory value.) I agree with George that the fine-tunings do not constitute proof, but they still carry weight. One can argue about how impressive the fine-tunings are (could we really exclude life if the constants changed a lot?), but I still think the number and precision of the tunings is remarkable. Nearly 30 years ago I wrote a review with Martin Rees about these fine-tunings (Carr and Rees 1979). In the intervening period a few of them have gone away (e.g. inflation may explain the value of the cosmological density parameter) but most of them have got stronger. Without a multiverse one may be forced to adopt a non-physical explanation like a fine-tuner, which is why Neil Manson (2003) claims that “the multiverse is the last resort of the desperate atheist”. This is not necessarily true — Paul Davies (2006) advocates a “third way” in which the laws of Nature evolve in a single universe in such a way that life can arise — but if you reject the multiverse, you certainly lower the scientific status of the anthropic arguments. I agree with George's argument against physical infinities. However, we do not need an infinity to validate the anthropic principle — just a large number. The existence of a multiverse is implied by a probability argument: the universe is no more special than it need be to create life. In particular, the small value of the cosmological constant shows that other universes exist. George argues that multiverse theories are not useful because they cannot be disproved: if all possibilities exist somewhere, then they can explain all conceivable observations. However, the fact that we only observe one sample of the multiverse still allows the proposal to be refuted at a given confidence level. Statistical predictions still qualify as science and that is why Rees has stressed the importance of calculating the probability distribution for various parameters across the universes. Indeed, a core difference between the Bayesian and frequentist views is the former's willingness to make inferences from single, and possibly unrepeatable, pieces of data. George rejects the Λ argument but there is no doubt that this has been very influential in attracting many physicists to the multiverse cause. It used to be thought that Λ was exactly zero and it was then plausible that there might be some physical (non-anthropic) explanation for this. However, the fact that Λ is non-zero but very tiny is a profound mystery that completely changes the situation. Critics say that we cannot know what distribution for Λ is predicted across the multiverse and that is correct. It may be simplistic to assume that the distribution is uniform, but postulating that there is a spike in precisely the observed region is just as improbable as what we are trying to explain. Even if one does not accept inflation, multi-verses are predicted by many theories of particle physics. It is still legitimate to invoke the existence of other universes for which there can be no direct evidence if one has a theory (like M-theory) that predicts this. It is not necessary to check all predictions of the theory for it to be considered scientific (e.g. we cannot probe inside black holes and we cannot see quarks but we still regard these as subjects for scientific discourse); it is only necessary to test some of them. Does M-theory qualify in this respect? George claims no; it does not come under the purview of science because our confidence in it is based on faith and aesthetic considerations (mathematical beauty etc) rather than experimental data. Certainly he is not alone in this attitude. For example, Woit (2006) and Smolin (2007) dismiss M-theory as mathematics rather than physics because it has not made contact with observations after 20 years. However, I feel this rejection is premature. It may take 200 years to solve the equations of M-theory and test them, but the definition of what constitutes a scientific question should not depend on how difficult it is. The nature of science changes, so what is illegitimate science today may be legitimate tomorrow. The fundamental issue in the dispute between myself and George concerns which features of science are to be regarded as sacrosanct. Experimentation used to be regarded as sacrosanct but by that criterion all of astronomy would be excluded since one cannot experiment with stars and galaxies. Fortunately, one can still make observations and — since there are billions of these objects — Nature effectively performs experiments for us. Cosmologists are in worse shape because there is only one universe to observe and speculations about processes at very early and very late times have to be viewed as ultra-speculative. For this reason, more conservative physicists regard even relatively standard cosmological speculations as trespassing into metaphysics. George places a lot of emphasis on falsifiability, but not everybody in the philosophy of science agrees with Popper on this and it is surely dangerous to impose a philosophical prescription that prevents scientists changing the border of their field. As Susskind cautions, it would be a pity to miss out on some fundamental truth because of an over-restrictive definition of science. Of course, one needs some degree of falsifiability, but the question is, how much? It is certainly not fair to put M-theory in the same class as astrology. On the other hand, I share George's scepticism of the Level IV multiverse, which corresponds to universes governed by different mathematical structures. The view that any mathematically possible universe must exist somewhere seems untestable in a deeper sense than Levels I to III. The notion of a multiverse entails a new perspective of the nature of science and it is not surprising that this causes intellectual discomfort. But this situation has often occurred before and one should not be surprised if it happens again. The Cosmic Uroborus in figure 2 shows that the history of physics might be regarded as the extension of knowledge into ever smaller and ever larger scales. The ideas encountered at the two frontiers have often been viewed as part of philosophy rather than science, so in a sense the debate is nothing new. However, there is another sense in which the current situation is very special. This is because — for the first time — the boundaries at the largest and smallest scales have connected, as indicated by the top of the Cosmic Uroborus, so the two science/philosophy frontiers have merged. Does this merging represent the completion of science or merely the sort of transformation in the perceived nature of science that accompanies every paradigm shift? This is a contentious issue and clearly we do not yet know the answer. I accept that there may eventually be a limit to the sort of questions that science can address; George and I merely disagree on whether we have reached that limit with the multiverse. In any case, we are surely behoven to try to take science as far as possible. I will end with a comment by Steven Weinberg (2007) in his contribution to Universe or Multiverse?: “We usually mark advances in the history of science by what we learn about Nature, but at certain critical moments the most important thing is what we discover about science itself. These discoveries lead to changes in how we score our work, in what we consider to be an acceptable theory.”