The most ambitious idea in the history of science is not a single theory but a family of related proposals, all united by a common purpose: to explain the extraordinary fine-tuning of our universe without invoking anything outside the universe itself — no creator, no pre-existing intelligence, nothing beyond the physical world. It goes by several names — the multiverse, the many-worlds interpretation, the string theory landscape, eternal inflation — and in its most expansive versions it proposes something that would have seemed like science fiction to any physicist working before the 1980s.

It proposes that our universe is not THE universe. It is one universe among an almost incomprehensible number of others — perhaps infinitely many — each with its own physical constants, its own laws, its own version of the dials we examined in Chapter Two set to its own values. In this vast collection of universes, every possible combination of constants exists somewhere. Most of those universes are sterile — their constants fall in ranges where no stars form, no chemistry develops, no complexity emerges. But some, simply by the sheer number of universes in the ensemble, fall in life-permitting ranges. And we, inevitably, find ourselves in one of those, because we could not find ourselves anywhere else.

On this picture, the fine-tuning is not really fine-tuning at all — it is selection. We observe what we observe because we are the kind of thing that could only develop in a universe like this one. The apparent precision of the constants isn't evidence of design; it's simply the expected result of our own existence, filtered through a collection of universes large enough to contain every possibility.

It is currently the leading naturalistic alternative to design.1 I am not going to dismiss it. I am going to follow it carefully to the place where it stops working. Chapter Two noted that the multiverse relocates rather than resolves the design question. This chapter examines that relocation in full.

―――――――――――――――――

The multiverse hypothesis does not come from nowhere. It emerges, or appears to emerge, from two of the most successful and well-tested theoretical frameworks in modern physics: inflationary cosmology and string theory.

Inflationary cosmology proposes that in the first tiny fraction of a second after the Big Bang, the universe went through a period of extraordinarily rapid expansion, far faster than the speed of light, driven by an energy field called the inflation. The inflationary model is well-supported by observations of the cosmic microwave background radiation — the faint afterglow of the Big Bang that fills the entire universe — and is accepted by most cosmologists as the best available description of the universe's earliest moments.2

The connection to the multiverse arises from a feature of many inflationary models called eternal inflation. In these models, the rapid expansion does not simply stop everywhere simultaneously. Instead, different regions of the expanding space stop at different times, each settling into what physicists call a bubble universe — a separate, causally disconnected region of space and time, with its own physical properties. The expanding space between the bubbles continues to expand, and more bubbles are produced as a natural consequence of the expansion. If this picture is correct, the total number of bubble universes is effectively infinite, and our observable universe is just one bubble among the vast ensemble.3

String theory independently provides a possible explanation for why different bubble universes would have different physical constants. String theory is a branch of physics that attempts to unify all the forces of nature into a single framework. One of its predictions is that there are an almost incomprehensible number of possible ways a universe could be configured — estimates reach as high as ten to the power of five hundred different configurations.4 Think of each configuration as a different set of physical rules — a different version of the dials from Chapter Two, each set to different values. Each bubble universe that forms in eternal inflation randomly settles into one of these configurations — meaning at the moment of each bubble's formation, its constants are genuinely free and independently selected, consistent with the T=0 argument developed in Chapter Two (T=0 meaning the very first instant of the universe's existence). If eternal inflation keeps producing new bubble universes, and each bubble randomly settles into one of these configurations, then given enough time and enough bubbles, every possible combination of physical constants would eventually be realized somewhere in the vast ensemble. Our universe — with its particular set of constants — would simply be one of the configurations that happened to permit life.

This is the string theory landscape, and it is the most developed and scientifically grounded version of the multiverse proposal. What makes it serious is this: it was not invented purely to explain fine-tuning. It emerges, or seems to emerge, from theoretical frameworks developed for independent reasons. If string theory is correct and if eternal inflation is correct, something like the landscape multiverse appears to follow, and the fine-tuning of our universe would be, in principle, explained.

But it does not resolve the problem it was built to solve.

―――――――――――――――――

The first and most fundamental difficulty with the multiverse hypothesis is that it is, at present, empirically untestable — meaning we cannot design any experiment or observation that could prove it right or wrong.

The other universes in the multiverse — if they exist — are causally disconnected from ours. They are not merely distant. They are outside our observable universe in a deep physical sense: no signal, no particle, no gravitational wave could ever travel from them to us, because the space between us is expanding faster than light and has been since the moment of their formation. We cannot observe them, measure them, or detect their presence by any means currently known or even theoretically proposed.

This matters enormously for the status of the hypothesis. Science, as a method of inquiry, rests on the principle of falsifiability — the idea that a hypothesis must make predictions that could, in principle, be shown to be wrong. A hypothesis that cannot be tested cannot be confirmed, but more importantly it cannot be falsified — shown to be false. It exists in a category beyond the reach of scientific method as normally understood.

Some physicists have proposed ways that bubble universe collisions might leave faint circular patterns in the cosmic microwave background, which would be visible if our bubble had collided with a neighboring one in the distant past.5 These would be, in principle, testable predictions of some multiverse models. So far, no such patterns have been found. The absence of evidence is not conclusive evidence of absence, and the search continues, but it is worth noting that the most specific testable prediction of some multiverse models has not been confirmed.

The point of raising the testability problem isn't to dismiss the multiverse — it's that it changes the nature of the comparison. Weighed honestly against the design hypothesis, both are, at present, empirically untestable. The designer cannot be observed or measured directly. Neither can the other universes. We are comparing two frameworks for explaining what we observe that both go beyond what we can currently verify — and we should apply the same standards of scrutiny to each.

―――――――――――――――――

The second difficulty is subtler and, I think, more important.

The multiverse hypothesis does not actually explain the fine-tuning. It relocates it.

The multiverse is not a random process. It is a structured process — a mechanism that generates universes with varying constants. That mechanism must itself be described by physical laws. Those laws must themselves be capable of producing the kind of universe-generating process we are describing. They must, in particular, be capable of producing a universe like ours — a universe where inflation occurs, where the laws of physics permit the kind of structure that generates further universes, where the whole mechanism works as required.

The multiverse works like a slot machine. It randomly produces different combinations on every spin — that is the random part. But the slot machine itself had to be built correctly. If it had been built differently — wrong gears, wrong mechanisms, wrong internal rules — it would produce nothing at all, just broken parts going nowhere. The multiverse is the slot machine. The individual bubble universes are the random combinations it spins out. But the machine itself — the underlying laws of string theory and eternal inflation that make the whole process run — had to be built in a very specific way. It could easily have been otherwise. A slightly different set of meta-laws might have produced a multiverse that generates nothing but empty, lifeless bubbles, or no bubbles at all. The fact that it produces a working, bubble-generating multiverse from which life-permitting universes can emerge is itself something that requires explanation. That explanation is not provided by the multiverse. It is the question the multiverse was supposed to answer, appearing again one level up.

In other words: the laws that govern the multiverse generator must themselves be life-permitting, in the sense of being capable of producing at least some universes that contain life.

This is harder to achieve than it sounds. Most possible sets of physical rules would produce nothing at all — no stars, no chemistry, no complexity, nothing that could think or ask questions. Yes, the individual bubble universes randomly receive their constants from the enormous range of string theory configurations — that part is random. But the mechanism that generates the bubbles in the first place — eternal inflation, the laws of string theory, the rules that make the whole multiverse work — is not random. Those underlying rules simply have to exist and function correctly for any of this to happen. They must be the kind of rules that produce something rather than nothing. But that means the multiverse's own governing rules still require explanation — why do those particular rules exist? Why are they structured in a way that generates a working, bubble-producing multiverse rather than nothing? We have not explained fine-tuning. We have simply pushed it back one step.

The philosopher John Leslie, who has written extensively on fine-tuning, put this well: if the multiverse generator itself requires explanation, then explaining our universe by pointing to the multiverse does not dissolve the problem of design. It generates a larger and more complex version of the same problem.6

The same logic applies to a Theory of Everything — the hope that a future physics will show the constants are mathematically necessary. A theory is still a framework, and frameworks describe what values are possible, not which value was selected at T=0. The relocation problem applies here just as it does to the multiverse: the design question is pushed back one level, not dissolved.

―――――――――――――――――

The third difficulty is the one I find most telling, because it does not come from philosophy or from the limitations of testability. It comes from biology.

Consider what the multiverse hypothesis actually predicts about us.

The multiverse predicts that life should have emerged from a random fluctuation in a smaller, simpler universe whenever possible — because smaller, simpler universes are more statistically likely in a collection of universes that takes all possible sizes and shapes.7 What we actually observe is a universe vastly more generous than that minimum: two trillion galaxies,8 roughly fourteen billion years old, with the physical structure of a genuinely old, genuinely large, genuinely complex cosmos — not the bare minimum that would suffice.

But the deepest problem comes when we look at what this universe produced within itself. It did not produce life once, in one lineage, by one evolutionary pathway. It produced consciousness — the capacity for conscious experience — repeatedly and independently, across multiple lineages: in mammals, in birds, in fish, in cephalopods (animals like octopuses and squid), across hundreds of millions of years of separate evolutionary history. Minds keep appearing in this universe by different routes and from different starting points.

The multiverse has nothing to say about either of these observations. It predicts a minimal universe; ours is far more generous than minimal.

―――――――――――――――――

Here is a scorecard — not a rhetorical victory but an honest accounting of where the evidence stands after three chapters.

We have three hypotheses before us. The first is chance in a single universe: the constants fell in life-permitting ranges by accident, with no mechanism and no selection. The second is the multiverse: a collection of universes realizes all possible constants, and we occupy one of the life-permitting members of that collection by anthropic selection. The third is design: the constants were set by an intelligence with the ability and intention to produce a universe of this kind.

Chance in a single universe faces numbers that break ordinary probability frameworks. The combined probability of all relevant constants simultaneously landing in life-permitting ranges, by random assignment with no mechanism, is not merely small — it is a number so small that all the resources of the observable universe could not represent it in any format. At that scale, the word "probability" stops meaning anything useful.

The multiverse is intellectually serious, empirically untestable at present, and faces the two problems I have described: it relocates rather than resolves the fine-tuning of the underlying rules that govern it, and it has nothing to say about the repeated, independent emergence of consciousness that we observe within our universe. It remains a live possibility — I will not pretend otherwise — but it is not an escape from the question. It is a sophisticated deferral of it.

Design has genuine explanatory power: a universe produced by an intelligence that values complexity and mind would be expected to have fine-tuned constants capable of producing those things. It faces the regress problem — what explains the designer? — which is a genuine philosophical challenge but not one unique to design.

Here it is worth pausing on something we established in Chapter Two. We showed that not everything that exists requires a cause — mathematical truths like one plus two equals three were never created, have always been true, and cannot be destroyed. This tells us that it is possible for something to exist without being created — without needing a cause. An intelligence of that kind — one that exists necessarily, the way mathematical truths exist necessarily — would not face the regress problem at all. This does not prove that such an intelligence exists. But it shows that the regress objection, while genuine, does not automatically close the door on design.

It also faces the problem of how to describe the nature of the designing intelligence, which takes us beyond physics into territory we will enter carefully and with appropriate humility in the chapters ahead.

After two chapters of building, this is where we stand on probability.

Chance in a single universe: approximately two percent. This is charitable. The honest number might be orders of magnitude smaller, but two percent acknowledges the genuine philosophical uncertainty about the range of values from which constants could have been drawn.

The multiverse with random origin: approximately thirty-two percent. A serious hypothesis that explains a great deal but leaves significant questions unanswered, particularly about the underlying rules that govern it and about consciousness.

Intelligent design of a single universe: approximately twenty-eight percent. Strong explanatory power, genuine philosophical challenges, honest uncertainty about the nature of the designer.

The multiverse with intelligent origin — that is, a universe-generating mechanism that was itself designed: approximately thirty-eight percent. This scenario, strange as it may initially sound, is internally coherent: an intelligence that chose to set in motion a universe-generating process, knowing that it would explore the space of possible universes and that at least some of those universes would contain minds capable of the kind of appreciation and inquiry we are engaged in right now.

Combined, the intelligent origin scenarios account for approximately sixty-six percent of the probability space. Random origin — whether in a single universe or a multiverse — accounts for approximately thirty-four percent.

These numbers will change as we add more evidence. They are not the final word. They are where the physics alone leaves us.11

Physics, it turns out, has brought us more than halfway to a conclusion that physics alone cannot complete.

―――――――――――――――――

It is worth closing this chapter with a reflection on what it means to take the multiverse seriously, because how one holds this hypothesis reveals something important about the underlying commitments of the inquiry.

The multiverse hypothesis is motivated, at least in part, by a prior commitment: the commitment to find a purely natural explanation for fine-tuning at any cost. This is not a criticism — prior commitments drive all inquiry, and the commitment to natural explanation has been enormously productive in the history of science. But it is worth naming the commitment for what it is, because it shapes the way the evidence is read.

The physicist Paul Davies, who is not a theist, has observed that the multiverse represents a departure from the usual standards of scientific evidence in one crucial respect: it multiplies unobservable entities without limit in order to explain one observed feature of one observable universe.9 This violates, or at least severely strains, what philosophers call Ockham's razor — the principle that explanations should not add more entities than are necessary to explain what we observe. In other words: when two explanations fit the facts equally well, the simpler one is generally preferred.

A designer — a single intelligence with the ability to bring a universe into existence — is, in terms of simplicity of explanation, considerably more economical than an infinite or near-infinite collection of unobservable universes. This does not prove design. Preferring simpler explanations is a methodological preference, not a logical proof. But it means that on at least one standard way of evaluating competing explanations, the design hypothesis is not at a disadvantage relative to the multiverse. It is, if anything, at an advantage.

Davies concludes, after a careful assessment of both hypotheses, that the fine-tuning of the universe for life and complexity "cries out for explanation" and that neither hypothesis currently available — multiverse or design — provides a fully satisfying one.10 He is right. The honest position is not confident theism or confident atheism. It is the position of a careful investigator who has followed the physics to the edge of what physics alone can say, and is now asking what comes next.

What comes next is biology. And what biology has to tell us about the distribution of minds in this universe will, I think, surprise you.

ENDNOTES — CHAPTER THREE

1. The physicists cited are among the leading advocates of multiverse hypotheses in the scientific literature. Steven Weinberg was among the first to argue that anthropic reasoning in a multiverse context could explain the cosmological constant; see: Steven Weinberg, "Anthropic Bound on the Cosmological Constant," Physical Review Letters 59 (1987): 2607–2610. Leonard Susskind developed the string theory landscape and its multiverse implications in: Leonard Susskind, The Cosmic Landscape: String Theory and the Illusion of Intelligent Design (New York: Little, Brown, 2005). Martin Rees surveys the evidence for fine-tuning and multiverse arguments in: Martin Rees, Just Six Numbers: The Deep Forces That Shape the Universe (New York: Basic Books, 2000). For Stephen Hawking's multiverse-related later work, see note 8 below.

2. The cosmic microwave background (CMB) radiation was first detected by Arno Penzias and Robert Wilson in 1965, earning them the 1978 Nobel Prize in Physics. The detailed structure of the CMB — particularly the pattern of temperature fluctuations consistent with inflationary predictions — was first mapped precisely by the COBE satellite (1992) and subsequently by WMAP (2003–2012) and the Planck satellite (2013–2018). For the CMB as support for inflation, see: G. Hinshaw et al. (WMAP Collaboration), "Nine-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations," The Astrophysical Journal Supplement Series 208 (2013): 19. DOI: 10.1088/0067-0049/208/2/19. For Planck results, see: Planck Collaboration, "Planck 2018 Results: X. Constraints on Inflation," Astronomy & Astrophysics 641 (2020): A10.

3. The concept of eternal inflation — in which the inflating space continuously produces new bubble universes, never coming to a complete stop — was developed primarily by: Andrei Linde, "Eternally Existing Self-Reproducing Chaotic Inflationary Universe," Physics Letters B 175 (1986): 395–400; and Alexander Vilenkin, "Birth of Inflationary Universes," Physical Review D 27 (1983): 2848. For a comprehensive review, see: Andrei Linde, "A Brief History of the Multiverse," Reports on Progress in Physics 80 (2017): 022001 (arXiv:1512.01203).

4. The estimate of approximately 10⁵⁰⁰ possible vacuum states in string theory was introduced in: Raphael Bousso and Joseph Polchinski, "Quantization of Four-Form Fluxes and Dynamical Neutralization of the Cosmological Constant," Journal of High Energy Physics 6 (2000): 6. The term "landscape" for this space of vacua was popularized by Leonard Susskind in: Leonard Susskind, "The Anthropic Landscape of String Theory," arXiv:hep-th/0302219 (2003). For a technical review, see: Michael R. Douglas, "The Statistics of String/M Theory Vacua," Journal of High Energy Physics 5 (2003): 46.

5. The proposal that bubble universe collisions might leave detectable signatures in the CMB was developed in: S. Chang, M. Kleban, and T.S. Levi, "When Worlds Collide," Journal of Cosmology and Astroparticle Physics 0904 (2009): 025 (arXiv:0712.2261). The first systematic observational search using WMAP data found no confirmed signatures: S.M. Feeney et al., "First Observational Tests of Eternal Inflation," Physical Review Letters 107 (2011): 071301; and S.M. Feeney et al., "First Observational Tests of Eternal Inflation: Analysis Methods and WMAP 7-Year Results," Physical Review D 84 (2011): 043507. The search continues with Planck data, with no confirmed detections as of 2024.

6. John Leslie, Universes (London: Routledge, 1989). Leslie is Professor Emeritus of Philosophy at the University of Guelph and has written extensively on fine-tuning and the design argument. The argument that multiverse explanations merely relocate rather than resolve the fine-tuning problem is developed throughout his Universes, particularly chapters 1–3. Anthony Flew described it as "the best book about the anthropic principle." For related arguments, see also: Robin Collins, "The Teleological Argument: An Exploration of the Fine-Tuning of the Universe," in The Blackwell Companion to Natural Theology, ed. William Lane Craig and J.P. Moreland (Oxford: Wiley-Blackwell, 2009).

7. Roger Penrose's argument that the multiverse predicts a much simpler universe than we observe — because simpler universes generating conscious observers by random fluctuation are vastly more probable in any large ensemble — is developed in: Roger Penrose, The Road to Reality: A Complete Guide to the Laws of the Universe (London: Jonathan Cape, 2004), pp. 762–765; and Roger Penrose, Cycles of Time: An Extraordinary New View of the Universe (New York: Alfred A. Knopf, 2010), chapters 2–3. This problem is related to the "Boltzmann Brain" problem in cosmology: in a sufficiently large or old universe, random thermal fluctuations are more likely to produce a single conscious observer (a "Boltzmann Brain") than the orderly universe we observe. See also: Sean Carroll, "Why Boltzmann Brains Are Bad," arXiv:1702.00850 (2017).

8. The estimate of approximately two trillion galaxies in the observable universe is from: Christopher J. Conselice et al., "The Evolution of Galaxy Number Density at z < 8 and Its Implications," The Astrophysical Journal 830 (2016): 83. DOI: 10.3847/0004-637X/830/2/83. This figure represents a tenfold upward revision from the previously accepted estimate of approximately 200 billion galaxies, based on analysis of deep-field observations.

9. Paul Davies, The Goldilocks Enigma: Why Is the Universe Just Right for Life? (London: Allen Lane, 2006; U.S. edition published as Cosmic Jackpot, Boston: Houghton Mifflin, 2007). Davies's critique of the multiverse as invoking "an overabundance of entities, most of which could never be observed" and requiring "a lot of unexplained and very 'convenient' physics to make it work" appears throughout chapters 8–9. Davies is a physicist and cosmologist at Arizona State University and recipient of the 1995 Templeton Prize; he is not a theist, which makes his critical assessment of the multiverse particularly notable. The reference to Ockham's razor in this context is standard in philosophical discussions of the multiverse; see also: Richard Swinburne, "The Argument from the Fine-Tuning of the Universe," in Physical Cosmology and Philosophy, ed. John Leslie (New York: Macmillan, 1990).

10. Paul Davies, The Goldilocks Enigma, op. cit., p. 3: "On the face of it, the universe does look as if it had been designed by an intelligent creator expressly for the purpose of spawning sentient beings. Yet most scientists are deeply uneasy with this explanation. They prefer to believe... that our existence is not the result of design but of chance." Davies concludes that neither the multiverse nor the design hypothesis fully satisfies: "In the end, both theological and multiverse explanations encounter the problem of what, if anything, lies beyond." This conclusion that fine-tuning "cries out for explanation" is consistent with Davies's longer treatment in: Paul Davies, The Mind of God: The Scientific Basis for a Rational World (New York: Simon & Schuster, 1992).

11. BASIS FOR THE PROBABILITY ESTIMATES — BAYESIAN METHODOLOGY

The probability estimates in the scorecard above are not the output of a precise mathematical formula. They are calibrated philosophical judgments produced by applying Bayesian inference — the standard scientific and philosophical tool for updating probabilities in the light of evidence. The following explains the reasoning behind each figure as transparently as possible.

The Framework: Bayesian Inference

Bayesian inference works in three steps. First, assign a starting probability — called a prior — to each hypothesis before examining the evidence. Second, assess how well each hypothesis predicts the evidence we actually observe — called the likelihood. Third, update the prior in proportion to the likelihood to produce a revised probability — called the posterior.

The formula is: P(H|E) = P(E|H) × P(H) / P(E)

In plain English: the probability of a hypothesis being true given the evidence equals the probability of observing the evidence if the hypothesis is true, multiplied by the prior probability of the hypothesis, divided by the total probability of the evidence across all hypotheses.

Step 1 — Prior Probabilities

Before examining any evidence, we assign equal probability to all four hypotheses — twenty-five percent each. This is called a uniform prior and is the standard starting point when genuine uncertainty exists and no hypothesis has a strong advantage before the evidence is considered. Some philosophers argue the prior for design should be lower because naturalistic explanations have historically succeeded in science. Others argue it should be higher because the existence of anything at all already requires explanation. The uniform prior is the most intellectually honest neutral starting point.

Step 2 — Likelihood Ratios

For each hypothesis, we ask: how well does this hypothesis predict the fine-tuning evidence we observe?

Chance in a single universe:

The cosmological constant is fine-tuned to approximately 1 part in 10¹²⁰. Even being extremely generous, the probability of this occurring by chance is approximately 1 in 10⁶⁰ at best. Its likelihood ratio relative to the other hypotheses is approximately 1 in 10⁶⁰.

The multiverse with random origin:

If the string theory landscape produces approximately 10⁵⁰⁰ universes and the life-permitting fraction is approximately 1 in 10¹²⁰, then the expected number of life-permitting universes is 10⁵⁰⁰ ÷ 10¹²⁰ = 10³⁸⁰. On this hypothesis our observation is not surprising. Likelihood ratio: approximately 1. However it is partially penalized for the second-order fine-tuning problem — the meta-laws governing the multiverse must themselves be life-permitting, which the hypothesis does not explain.

Intelligent design of a single universe:

A designer who intends to produce life would be expected to set the constants in life-permitting ranges. Likelihood ratio: approximately 1.

The multiverse with intelligent origin:

This hypothesis predicts fine-tuning at both the level of individual constants and the structured meta-laws. Likelihood ratio: approximately 1.2 — slightly higher than single-universe design.

Step 3 — Updating the Probabilities

Applying the likelihood ratios to the uniform priors, the chance hypothesis collapses to approximately two percent. The remaining probability is redistributed across the three surviving hypotheses in proportion to their likelihood ratios, adjusted for the partial penalty on the random multiverse: