by Brian Tomasik
First written: 9 Jan. 2015; last update: 9 Oct. 2016

Summary

Why does physics exist? What makes logical rules the "right" ones to govern physical possibilities? And what does "exist" even mean? Puzzles like these seem to be the hardest in philosophy, not amenable to the sorts of explanations that can account for higher-level problems built on top of these assumptions. I have no satisfying answers, and it's not clear that humans will ever produce valid answers, assuming the questions are even coherent at all. Confronting these questions helps remind us of how much our fundamental assumptions about reality could change with further insights.

See also "Why does the universe exist? – A bibliography".

Introduction

"The question of being is the darkest in all philosophy." --William James, 1911

It seems that most puzzles of philosophy can be resolved or dissolved by science. Understanding cognitive science and picturing ourselves as biological computers helps make sense of free will, consciousness, time, sorites paradox, the nature of mathematics, and more.

But one set of puzzles seems to be completely mysterious, in the sense not only that we lack a solid explanation but that we can't imagine what any explanation could look like. These are the questions like

  • Why does physics exist?
  • Why does it obey the laws of logic and mathematics?
  • What does "exist" even mean? (How would you taboo "exist"?)

It's hard to fully capture the oddity of these questions in words. There are times when I think about them and then feel as though I'm standing on the edge of the roof of the Empire State Building. It feels just so strange that there's a universe that contains stuff, including this glob of stuff here that I call "myself". What's up with that? How is that even possible? What does it even look like for that to be true? At some point the questions descend into incoherence because we lack the concepts to even talk about the weirdness of the situation.

I find it interesting that some people don't seem to consider this question much or maybe even at all, instead just taking reality for granted. Of course, this isn't surprising, since evolved creatures that spend most of their thoughts puzzling about why anything exists aren't likely to last very long. I myself only ponder it on occasion, and most of the time I'm too much engrossed in everyday life to feel the weirdness of thinking that there even is such a thing as "everyday life".

Is there something rather than nothing?

Many philosophical disputes can be dissolved by explaining how the postulated entity doesn't actually exist, at least not in some ontologically fundamental sense (libertarian free will, property-dualist consciousness, a fundamental "flow" to time, a "real" boundary of a heap, platonist mathematical objects, etc.). It's certainly possible -- I would say probable -- that all human concepts of fundamental reality are at least somewhat confused and that this confusion infects our questions like why there's something rather than nothing.

That said, it seems like we must be pointing to something when we ask why and how physics exists. Descartes got many things wrong, but cogito ergo sum seems to have been one of his more convincing arguments, although I agree with critics that it should be revised along the lines of: "Thinking is occurring. Therefore, something exists." Even this may be on uncertain footing because at bottom it's not clear how to explain concepts like what it means to "exist". Indeed, if we could do that, maybe the question of why anything exists would be less puzzling. But even upon ontological and conceptual overhaul, it seems plausible that something like the cogito argument would still remain sound. In other words, it seems that there really is something rather than nothing, and even if we're thinking about it completely wrong, the existence of physics is not entirely illusory, though see this later section for a contrary view.

Insufficient explanations

God

Theism is the classic explanation of why the universe exists, but the problem with it is clear: It just pushes the mystery one step back to "Why does God exist?". The only effect is to make the hypothesis more complex. Jonathan Wallace and Eliezer Yudkowsky helpfully call non-explanations like these "semantic stopsigns".

"God" could be explanatorily helpful if it was a very simple mechanism to generate the universe from elegant initial conditions. Even then God wouldn't fully answer the question but would only reduce the size of what needs to be explained.

Physics itself

Lawrence Krauss's A Universe from Nothing argues that physics can explain the origins of the universe without God. While physical explanations of cosmogony are very interesting and plausibly relevant to the ultimate question, they don't yet answer the full puzzle.

David Albert and others have pointed this out. Sean Carroll summed up the debate:

Very roughly, there are two different kinds of questions lurking around the issue of “Why is there something rather than nothing?” One question is, within some framework of physical laws that is flexible enough to allow for the possible existence of either “stuff” or “no stuff” (where “stuff” might include space and time itself), Elliptical Galaxy NGC 4150why does the actual manifestation of reality seem to feature all this stuff? The other is, why do we have this particular framework of physical law, or even something called “physical law” at all? Lawrence (again, roughly) addresses the first question, and David cares about the second, and both sides expend a lot of energy insisting that their question is the “right” one rather than just admitting they are different questions. Nothing about modern physics explains why we have these laws rather than some totally different laws, although physicists sometimes talk that way — a mistake they might be able to avoid if they took philosophers more seriously.

Universe has always existed

Some accept the weirdness of the universe's being created. After all, nothing comes from nothing, and everything must have an explanation. From this starting point one might conclude that the universe was never created, which must mean it always existed. This is a "turtles all the way down" account.

In this case, it may be true that for any particular event, we can point to its cause. But the spirit of the question why there's something rather than nothing remains unaddressed. The fundamental puzzlingness of infinite regress has just as much force as if we postulate that time began at a definite endpoint (the big bang) of a timeless mathematical object.

Trilemma

The above positions resemble those of the Münchhausen trilemma:

  1. God is an axiomatic approach -- an unanalyzable bedrock answer.
  2. Physics creating itself is a circular approach.
  3. Universe always existing is an infinite regress.

None is actually satisfying.

Modal realism would get us only part way

Carroll's article mentions the modal-realist explanation: Our universe exists because all possible universes exist. This seems to help a bit because at least it reduces the arbitrariness of the question why this universe exists and not some other. It seems that the two simplest scenarios are that nothing exists or that everything exists.

Max Tegmark's mathematical universe hypothesis can be seen as answering the question of what makes a mathematical abstraction concretely actualized by saying there is only the abstract.

But lurking underneath, there remain puzzles, such as the following:

  • Jürgen Schmidhuber and Alexander Vilenkin note that it remains unclear how to apportion measure among all possible universes. Whatever the measure is, it would seem to violate the simplicity of the explanation, since then we have to ask "Why that measure and not a different one?" (That said, if we reject Nick Bostrom's anthropics, we no longer need a notion of measure over universes, which would render this point a non-problem. See the appendix of this piece for further discussion.)
  • If physics is only mathematically abstract rather than concrete, what does "abstract" mean?
  • What are the limits of "possible"? Which parts of logic and mathematics define the boundaries there? What about paraconsistent logic and dialetheism?
  • Why these rules of logic and not others? Why does any logic "exist", so to speak? What does that even mean? How do you explain logic in simpler terms? And if you can't, what is it?

Is the problem soluble?

Following are some off-the-cuff, ungrounded guesses of mine about whether the question of why physics exists can be answered:

Answer Probability
A solution exists, and humans could figure it out 15%
A solution exists, it's too deep for human minds, but superintelligences could figure it out using physically realistic computational resources 10%
A solution exists and could be computed by a theoretical Turing machine, but the resources of the actual universe aren't big enough to allow for that 2.5%
A solution exists, but no Turing machine could figure it out 5%
The question is meaningful but has no answer 5%
The question is based on so much confusion that it's not even meaningful, even when interpreted charitably 47.5%
Something else I can't imagine 15%

Maybe it wouldn't even make sense for there to be an explanation, since explanations start from initial conditions and move to later conditions, but in this case we're trying to explain the first initial conditions. That my brain gets dizzy thinking about this doesn't mean my brain is tracking a question that can actually be answered. My brain can also get dizzy if I spin myself around too fast, but that dizziness doesn't mean I'm thereby picking up on some fundamental mystery to the universe.

Why the question matters

This question is not just an idle way to pass the time but may concretely affect how we approach altruism. In general, areas where you're confused are often areas most ripe for what Robin Hanson calls "viewquakes" -- insights that disrupt your previous understanding of reality and thereby either make you more effective at accomplishing existing goals or show you that your goals need adjustment because they were previously based on a confusion.

I doubt that detailed investigation of why physics exists is a top altruistic priority because it doesn't seem tractable; humans all over the world have been pondering the mystery basically forever. But I think we should cogitate upon the problem and use it to update our intuitions about the powers and limits of human understanding. Religions sometimes speak about the finitude of human knowledge compared with God's unlimited insight. While I wouldn't bring God into the picture, that religious thought does capture a useful reminder about how much we don't know.

I think the main take-home message is to remember how different the ontologies of post-humans might be from present-day physics. Maybe there's vastly (perhaps infinitely) more suffering (or something that should roughly be construed as such) than we realize, existing in other realms that we can't even comprehend. Even if this is true, it's not clear that our descendents would be able to determine this either. But at least they might develop vastly more sophisticated approaches for handling ontological uncertainty of this type. (Alas, note that this doesn't imply that it's good for our post-human descendents to exist, since I find it more likely they would take actions that increase suffering rather than decrease it, such as by expanding the size of the multiverse if doing so is possible.)

Maybe there's no such thing as "existence"?

Scott Aaronson:

what difference does it make if the lookup table physically exists—why isn’t its abstract mathematical existence enough? (Of course, all the way at the bottom of this slippery slope is Max Tegmark, ready to welcome you to his mathematical multiverse!)

A natural extension of rejecting the hard problem of consciousness is to reject the hard problem of existence. Maybe there is no "reality fluid" that distinguishes the real from the merely possible. This seems like a reasonable and consistent stance to take, but as discussed below, its implications are extremely weird. While I'm quite comfortable with eliminativism on consciousness, eliminativism on "existence" is a bridge I'm not yet ready to cross. (Maybe a future version of myself will look back on my hesitance about this with derision.)

Marvin Minsky appears to hold this view:

it seems to me you shouldn't be asking "What made everything?" You should be saying, "What could this word 'exist' mean [...]?" And I think the best answer is that it's probably meaningless. There are probably a lot of possible universes, and there's nothing special about any of them.

He continues: "There's no way you could tell whether you're in one [universe] that exists or not [...]."

I agree that the logical structure of a universe containing an organism reflecting on itself is identical whether or not the universe is endowed with "reality fluid", but it still seems to me a different question whether there's such a thing as "reality fluid". This looks like a non-empirical, ultimately metaphysical question, because Minsky is right that there's no way to "observe" whether we exist or not. Rather, we just need to sort out some fundamental concepts about what we're even talking about, although our intuitions about metaphysics will certainly be influenced by empirical observation.

Appendix: Tegmark's ultimate ensemble without objective measure

Much of this section is inspired by tidbits I've read on LessWrong over the years, especially from Wei Dai. It took a long time for me to actually understand why these ideas make sense.

My previous understanding

Prior to ~2015, I understood Tegmark's Level IV multiverse in the following way: There are tons of universes (i.e., mathematical structures), but there's some anthropic measure over universes determining which we should expect to find ourselves in. For instance, simpler universes have higher measure, which explains why our universe is relatively simple. We can do science by taking the measure over universes as our prior and then updating based on what we observe about our particular universe. That simpler universes have higher measure justifies Occam's razor.

And even though all possibilities are realized somewhere with some measure, we can make a difference to the multiverse by (timelessly) affecting the measure with which outcomes are realized. For example, suppose I decide to do some seemingly good action. A person who takes a Level IV multiverse as implying futility would complain that in one universe, the good action has good consequences, but in another universe, it has bad consequences, so there's nothing we can do to make the multiverse better. In reply, someone who believes in measure over universes can explain that we should aim to make the good outcomes happen in the higher-measure universes. For instance, you should attempt the good action because in the higher-measure (simpler) universe, doing so has good consequences, while the fluke bad consequences only happen in the lower-measure universe. This is better in terms of measure-weighted utility than doing a bad action that, in some fluke universes, has good side effects.

But the problem with this view, as noted in the main text, is that it requires there to be some measure over universes in the ultimate ensemble. But what could that measure be? And why this measure rather than some other one? Postulating a measure introduces free parameters, which we tried to eliminate by going to a Level IV multiverse.

My new understanding

What if we could get by without an objective, God-given measure? This isn't possible with Bostrom's anthropic framework, since if we think of ourselves as observers drawn from the set of observers according to measure, then there must be some measure that determines how likely we are to be this observer rather than that one. But in fact, this is all a fantasy. There is no drawing of observers from a distribution according to measure. There is no such thing as a discrete observer at all. There is just the multiverse, with all possibilities laid out in logic-space, being what they are, without needing to conform to some measure. (Indeed, it seems a hopeless mathematical task to assign a measure to all mathematical structures, unless the set of them is restricted to computable bitstrings or something more tractable.)

But doesn't this render action futile? Without measure, there's no longer a sense in which the good outcomes of my well-intentioned action have more weight than the bad outcomes. To solve this problem, let's take a detour into epistemology.

Besides rescuing action from futility, the other function of measure in the Level IV multiverse was for carrying out science: Measure served as a prior over universes, so that we could assume we were in simpler universes and then update our probabilities based on evidence. If there is no measure over universes, we have no prior probabilities, and we also seemingly can't update based on evidence because the freak-observer problem is maximally vexing in a Level IV multiverse.

The way around this is to replace epistemology with prudence. Rather than assuming you're not a Boltzmann brain because the probability of your being that observer is very low, you just don't act like a Boltzmann brain, because there are many more copies of you that aren't Boltzmann brains, and controlling them in good ways has much vastly more potential for altruistic impact. But what do we mean by "many more" copies within a Level IV multiverse where the infinities of structures at play are too large to handle? (Indeed, cosmologists have trouble with this question even in a Level ~II multiverse.)

My guess is that there's no objective way to count copies in a Level IV multiverse. Maybe there is one natural way that mathematicians of the future would converge on, but I'm doubtful because the problem seems so hard.a However, we still seem to have intuitive notions of "more common" and "less common". We could try to formalize these into an invented measure that we attribute to the multiverse. Within the multiverse of bitstrings, a common measure that people attribute to it is, up to a proportionality constant, 2-(Kolmogorov complexity). In fact, this is the typical Occam prior for Solomonoff induction. So in a sense, Occam's razor in epistemology -- which we all thought was a question of getting closer to truth -- is actually a process of constructing how much weight we want to give to different possible worlds. (The postmodernists are vindicated! There is no truth about which priors are better, only social constructions.)

The same idea applies to our actions. We invent a measure over the Level IV multiverse to say how much we care about different universes, and then we choose actions with higher measure-weighted utility. We could inherit this measure from epistemology if we so choose, and in that case, we might want to resolve tensions between epistemology-implied measure and our moral convictions. Paul Christiano: "But I do think that to the extent that we dislike solomonoff induction as an allocator of moral value we should also reject it as a prior over experiences". Personally I find the Solomonoff prior to be fine for epistemology but morally ghastly, since it suggests that a universe generated by a string one bit longer matters only half as much. My ethical impulse is that all universes should matter equally, insofar as this can be made coherent given the insanely huge infinity of universes at play.

So there are issues to work out, but at least it's now clear what's going on. Measure isn't an objective feature of the multiverse but is something we attribute, to express our "degree of (prudentially implied) belief"b or "degree of caring" about some possibilities versus others. Basically what we're doing is moving the free parameters that would have been required to explain an objective measure over universes into the realm of ethics and social construction, where the arbitrariness is then about our choices rather than about reality itself.

Maybe this approach can also make sense of measure in quantum mechanics, where the Born rule says that we care about quantum branches in proportion to their modulus squared.

A final question is: What's a "copy" of "you"? By "you" I mean one particular spatiotemporally connected piece of matter. By "copy" I have in mind another chunk of physics that's arranged in a sufficiently similar way to your present body and environment such that it tends to make basically the same choices, tends to have basically the same experiences, etc. For example, borrowing a motivating illustration from Eliezer Yudkowsky, if you have a calculator on Earth and an identical calculator on another planet, and if you press "8 * 8" into both of them, both return "64" because they're copies of each other. In a classical universe, we can count on roughly identical physical systems producing roughly identical outputs. But in a quantum world, a given physical system produces all possible outputs with some measure. So it's no longer true that all copies are guaranteed to behave the same way. We can resolve this conundrum similarly as we resolved what it means to count copies: By noting that if we pick a simplicity-favoring measure, then it should tend to be the case that similar input systems produce similar output choices, and it's only in the fluke worlds where that's not true. So unless you're a fluke copy, most of your measure-weighted copies do behave the same as you, and hence the idea that all your copies make the same choice is roughly true in a measure-weighted sense. Here's an oversimplified example to make this point clear. Suppose that 99% of the time (with 99% measure), your body+environment behave classically, and 1% of the time (which is unrealistically high), quantum effects interfere. If there are 1 million copies of your body+environment, then if you're in the 99% of copies that behave classically, then about 99% * (1 million) = 990,000 other copies behave the same way you do.

The view described here implies a radical degree of epistemic relativism. For instance, fundamentalist Christians are not wrong (except insofar as their beliefs are logically contradictory); they just care more than non-Christians about the subset of universes in the multiverse in which copies of us exist on a planet created and overseen by the God described in the Bible. Science-minded people reject Christianity because the universes consistent with it have high Kolmogorov complexity, but this only reflects a prejudice by science-minded people to care more about simpler universes. The presuppositionalists are right: "there can be no set of neutral assumptions from which to reason with a non-Christian." The two sides in the debate just hold different preferences about which universes in the modal-realist multiverse are more important. By the same token, Muslims, Raëlians, and Flying Spaghetti Monster adherents are not wrong either -- they just care most about their own favored types of universes. All logically consistent religions -- or beliefs of any kind -- are correct somewhere (and infinitely often) in the modal-realist multiverse.

In "Understanding Deutsch’s Probability in a Deterministic Multiverse", Hilary Greaves offers what I think(??) is a similar conclusion, just in the realm of the quantum multiverse rather than the larger modal-realist multiverse:

I argue that subjective uncertainty is not to be had, but I offer an alternative interpretation that enables the Everettian to live without uncertainty: we can justify Everettian decision theory on the basis that an Everettian should care about all her future branches. The probabilities appearing in the decision-theoretic representation theorem can then be interpreted as the degrees to which the rational agent cares about each future branch.
[...]
I have suggested that we should accept the Born rule itself as something of a primitive; this suggestion is supported by the observation that there are no remotely plausible alternatives on offer.

Anyway, this is all very mind-bending and non-mainstream, so I'm not sure how seriously to take it. But it does seem to be a consistent framework for understanding reality.

(See also a discussion of this topic on Facebook.)

Footnotes

  1. It's quite plausible to me that the multiverse is just too big for us to be able to ever do sensible utilitarian calculations over it or assign sensible, normalized measures to its constituents. Religious ideas about God (i.e., the multiverse) being beyond human understanding (or perhaps the understanding of any physical agent that can be built in our universe) may apply here:

    God is infinitely better and greater than man. Thus we can build all the little theoretical molds we want, and we can try to force God into these molds, but in the end God will not “fit.” He will always be beyond our grasp. He is too high for us to scale and too deep for us to fathom. We cannot get God in a box. The finite span of the human mind will never encompass the infinite God of [ Max Tegmark's ;) ] Scripture.

      (back)

  2. Universes with more measure generally have more measure-weighted copies of us, which implies we can generally have more impact in those universes, which means we should generally "act as if" those universes are true, and hence we have more prudentially implied "belief" in those universes.  (back)