[loup-vaillant]

There's a difference between the epicycles and the multiverse (assuming we're talking about quantum decoherence, I don't really know about the others). See, the epicycles really were additional entities, and therefore a problem, with respect to Occam's razor.

Quantum decoherence is different: it merely follows the equations, and do not posit any additional entity on top of them (such as a collapse). The consequence of removing that entity is multiple universes, but so what?

The problem with science is that it tends to favour the first theory that fits the fact. Instead, it should favour the simplest theory that fits the fact. http://lesswrong.com/lw/qa/the_dilemma_science_or_bayes/

[sriku]

I thought the multiverse "theory" the article talked about was not the many worlds interpretation of quantum mechanics, but the proposal that multiple universes are exploding like bubbles with the boundary of each expanding at the speed of light so no communication between them is possible. No?

You do bring up a valid point about occam's razor - i.e. It is a valid effort to come up with a strictly simpler theory that can explain already known observations than what is currently accepted. "Simpler" in the sense of fewer assumptions being needed. The difficulty is that we're often blind to the assumptions we're making in our theories and it can take a while for those to surface.

[nerd_herder]

> I thought the multiverse "theory" the article talked about was not the many worlds interpretation of quantum mechanics, but the proposal that multiple universes are exploding like bubbles with the boundary of each expanding at the speed of light so no communication between them is possible. No?

Yes, OP's article is talking about bubble-universes, the lesswrong article is talking about the Many Worlds Interpretation. I think parent comment is talking about MWI as well.

[JoeAltmaier]

I wonder if that isn't essentially 'overfitting the data'. Any bounded data set can be modeled simply. But it doesn't predict anything. A useful theory would beget testable hypotheses.

[sriku]

They're not quite the same when it comes to physics, where you have the additional characteristic of precision for theories. For a trivial example, "anything can happen anywhere at any time" certainly covers all observations and is simpler, but has no precision in what it forbids.

Any new theory in physics must have at least the same precision as the known theories. The problem in current physics-at-the-edge (raised by the article) is that these newer theories are coming up with so many ways in which universes can be formed that even though we're moving in the "simpler" direction by relinquishing fundamental some assumptions, we're losing "precision" in this sense. This is now happening to the extent that the very notion that the job of a scientific theory is to only explain the observable universe is being called into question and theories that consider our universe to be just one of many universes with different parameters are also included in the play. At that point, the working assumption that all these universes that our mathematics invents actually exist though we cannot observe them, is a simpler view of the cosmos ... until we know better. However, the practical utility of such an expansive theory ends up being limited relative to the theory that it is trying to replace, since our ability to predict things in our universe is not being improved on.

edit: What I'm trying to say is, I guess, that "precision" and "simplicity" are opposing forces that keep a check on overfitting.

[T-A]

Quantum decoherence != multiverse, and the multiverse interpretation of quantum mechanics != the multiverse of cosmology mentioned in this essay.

The latter is a rather more concrete affair, basically saying that in regions of the universe too far away to be observable, the parameters of low energy physics (kinds of particles, strengths of interactions between them) can be different from what we see in our region; rather than being fundamental properties of physics, they were picked randomly (by physical processes) at the big bang. If they are not fundamental properties of the theory, there is no need to come up with a theory which explains their values (the fine-tuning problem mentioned in the essay); we live in a region where the values are such that our existence is possible, because that is the only possibility. In regions where the values do not allow the existence of observers capable of asking "why are the values such that I can exist?", there are no observers asking "why are the values such that I can exist?".

Some call this idea (the anthropic principle) neat, others call it a cop-out and point out that since the postulated regions with different physical parameters can not be observed, the whole construction falls outside the scope of empiricism, and so can not be science (as traditionally understood). Hence essays such as this...

[loup-vaillant]

(Note: by "entity", I mean an assumption in a theory. Whatever is a consequence of the theory itself is not an entity, from an Occam's razor point of view)

I did imply I did not know what multiverse was being considered. As far as I can tell, the article did not say which multiverse it was talking about. There are many reasons to believe in an… unbelievably large universe:

1) Beyond the observable universe. Beyond a certain limit, the universe expands so fast that even light from there can't reach us, ever. Beyond that limit, it might as well be another universe. But whatever lies beyond that limit aren't additional entities. They're just the consequence of known laws of physics. Positing that they somehow don't exist would form an additional assumption in the theory, and therefore not good from an Occam's razor perspective.

2) Inflation. Would apparently create a number of "bubbles" or something, that are sufficiently far a part not to observe each other. I can't judge this one.

3) Macroscopic decoherence. The particle is in in a superposition of being destroyed/intact, the cat is in a superposition of being dead/alive, the scientist that observe the cat is in a superposition of mourning/petting the cat… Well, the Many World Interpretation of quantum mechanics. Well, that's what we call the Many World Interpretation of quantum mechanics. Again, no additional entity here: the other universes are just a natural consequence of long known equations. If anything, we remove an entity, compared to the Copenhagen interpretation: that pesky collapse.

4) Tegmark's level IV multiverse, where every possible mathematical construct "exists" in some sense, and our universe is just one of them (which also happen to support sentient life). Right now, I don't know what to think of it. Though it would be incredibly convenient, from an anthropic principle stand point.

> since the postulated regions with different physical parameters can not be observed,

Are we postulating the regions, or are we postulating a large universe with changing parameters? This is not the same thing. The former is obviously incredibly complex, and therefore a priori impossibly improbable. The latter doesn't involve that many entities, and may even be simpler than current mainstream theories. Or it may not. I'm not a physicist.

[z92]

I still find basically both epicycle and multiverse theories on the same level. That multiverse theory replaced other theories (like collapse) doesn't make a meaningful difference, as probably epicycle explanation also replaced other more complex theories that were floating around at that time trying to explain the backward motion of the planets.

[loup-vaillant]

Ellipses are more counter-intuitive, compared to circles, but epicycles are still more complex than ellipses. Not to mention, ellipses are explained by something even simpler, namely Newton mechanics.

Many World Interpretation is not like the epicycles. It is like the ellipses. It's the Copenhagen interpretation that is like the epicycles, by postulating a collapse that the equations don't mention at all. The MWI is just taking the equations at face values. Postulating a collapse on top of that makes a more complex theory.

If many world came first, the collapse postulate would just be laughed at. "You're postulating a collapse in just the parts of our universe we can't observe? That reeks of "if I can't see it, it doesn't exists. If you're going to push that theory, you'd better produce empirical evidence."

[jules]

Note that the multiverse that this article is talking about is very different than the many worlds multiverse. The many worlds multiverse is just that the whole universe is in a quantum state. The multiverse that this article is talking about is different. Due to the expansion of space we can only see some finite region around us. If you go far enough away, then the expansion of space between that point and the earth is faster than the speed of light, so we can never reach that point, and nothing from that point can reach us. Effectively we are in a different universe than that point. The multiverse theory is that some things that appear constant in our part of the universe, such as the fine structure constant, may well be not exactly constant. Then it could be the case that the fine structure constant in that far away part of space that we can never reach is different than our fine structure constant. Then because the universe continues to expand, regions that were connected become disconnected. So one region with fine structure constant x may split into two regions one with fine structure constant x + 0.0002 and another with fine structure constant x - 0.0001. That process creates an infinite tree of regions. If a region happens to have physical "constants" that create deflation rather than inflation, that branch of the tree dies. By the anthropic principle we live in a region with properties that produce humans.

By the way, I don't get the obsession with quantum mechanics interpretations. Copenhagen and many worlds give exactly the same predictions, so if one is correct then so is the other. Who cares how you choose to interpret it. By analogy to probability theory, Copenhagen is like conditioning on an observation, many worlds is like looking at the whole probability distribution. These are just two ways of looking at the same thing. It's as silly as arguing whether the Hamiltonian or Lagrangian formulation of classical mechanics is correct. Both are.

[loup-vaillant]

> By the way, I don't get the obsession with quantum mechanics interpretations. Copenhagen and many worlds give exactly the same predictions, so if one is correct then so is the other.

Not quite. While they predict the same observations, they certainly don't predict the same universe. Under the Copenhagen interpretation, there is only one cat, who is either dead or alive. The possibility you don't see doesn't even exist, the collapse has seen to that.

We can make an analogy with the expansion of space being faster than light. Let's say you send a life ship far away into deep space to do some colonisation. Let that ship travel beyond our observable bubble (it's a very high tech ship).

So, once your ship is so out of reach that it can't even send any signal back (not even in theory), does it still exist? If you take the current laws of physics at face value, it's still out there. The colonists are on their own, but they should be fine. On the other hand, if there is some kind of "collapse" where anything that goes beyond our observable bubble just disappears, then you have sent the colonists to their death. Oops.

For the record, I must say I am very uncomfortable about having the fundamental constants of the universe change as we go beyond our observable bubble. That sounds like an additional assumption, and I don't like it at all. I'd sooner believe in a Tegmark level IV multiverse.

[jules]

It's not entirely clear what happens when a spaceship goes outside our observable bubble due to issues with conservation of information (similar to how it's not entirely clear what happens when you drop a spaceship in a black hole). Let's leave that aside and assume that the spaceship and all its information is truly lost when it leaves our bubble. Then the question of whether that spaceship still exists is not a sensible question in physics, because there is no experiment that can confirm or deny it. It's a metaphysical/philosophical question. My point is that BOTH the claim that it disappears AND the claim that it does not disappear aren't sensible. Many worlds says that if you observe the alive cat, there still exists some other version of you that observes the dead cat. That is a metaphysical claim. Copenhagen says that there is no other version of you that observes the dead cat. This is also a metaphysical claim. I say that both claims are silly if you truly believe in the predictions of quantum mechanics. Whether the other version of the cat still exists is not a question worth worrying about, like it's not worth worrying about whether an invisible god exists or not. The simplest theory is to make no claim either way.

By the way, many worlds by itself is not actually a complete theory. It simply says that the wavefunction of the entire universe evolves according to the rules of quantum mechanics. To actually get predictions out of it you have to say something extra about observers within that universe. We certainly don't experience multiple simultaneous versions of ourselves, but we do experience multiple simultaneous versions of the things around us (e.g. double slit experiment). You need additional rules about what an observer in the universe will see and with which probabilities.

[loup-vaillant]

> It's not entirely clear what happens when a spaceship goes outside our observable bubble due to issues with conservation of information (similar to how it's not entirely clear what happens when you drop a spaceship in a black hole).

Crap. Okay, I'll keep that in mind.

> Then the question of whether that spaceship still exists is not a sensible question in physics, because there is no experiment that can confirm or deny it.

We could say it's not a sensible question in science (no experimental difference), but I think this is still a very important ethical question: I would still care about whether the colonists live or die.

Long term, this could be a very practical question: how should we expand? Must we stay within reach, or can we safely go as far away as possible? At this point, I don't really care if it's a metaphysical question. From the look of it, there's a definite answer, and one which will influence expansion policy a great deal.

Though to be fair, the point is kinda moot until we have a theory of everything.

[chriswarbo]

When theories get 'large' then the location of our observations within that theory becomes just as important as the rules governing them. We need to take both into account when using Occam's razor.

For example, explaining observations using multiple-worlds QM is very simple in terms of the rules used, but has a very complex "location" within that theory (ie. pin-pointing which Universe those observations are being made in).

There's a nice overview of this idea in http://www.mdpi.com/1999-4893/3/4/329

[jkot]

Epicycles provided better predictions than heliocentric system until Newton.