I'm not a subscriber to the many-worlds interpretation so I'm not the right person to defend it, partly for this very reason: the question of when one universe "splits off" doesn't actually have an answer, for the same reason that the question of "when does a measurement actually happen?" doesn't actually have an answer. Many-worlds and Copenhagen both have this rhetorical problem: they want to draw a sharp line when something "happens" (collapse, universe-split) and there is no such sharp line. The transition from quantum to classical is gradual, not abrupt. It typically happens very fast (picosecods or femtoseconds to get to the point where the state of the system is no longer distinguishable from a classical state) which is why it appears to be an abrupt transition, but it's not. This is one of the many reasons that I find the QIT/zero-worlds interpretation to be the most satisfactory. It and decoherence are the only ones that don't have this problem. But decoherence is too mathy :-)

IANAP, but I think part of the problem with MWI is that popularizers of it wanted to sound cool by calling it the Many-Worlds Interpretation in the first place.

The idea was originally known as the Theory of the Universal Wave Function, which makes a lot more sense as long as you're unafraid of mathematics.

There is no "splitting off" of universes; I agree with you that that wouldn't make much sense. Instead, the observer simply becomes entangled with the observed physical system during the measurement process. This entanglement is a gradual process, though it certainly happens very quickly.

This basically pushes all the weirdness out of quantum mechanics and into the fact that we just don't understand consciousness very well. Why don't we perceive the full linear combination of quantum states? And can we somehow map our subjective experience of probability and statistics to what happens in the Universal Wave Function? It's much easier to ignore that weirdness, since we're used to ignoring it in our daily lives anyway. It's also a more parsimonious interpretation than the others, because it doesn't postulate anything special about us (the "observers").

(This does not mean that consciousness is a quantum or even meta-physical phenomenon. Quite the opposite, actually: I find that the mystery lies more in why consciousness is unable to perceive quantum states despite existing in a universe that has quantum physics.)

> Why don't we perceive the full linear combination of quantum states?

My personal belief is that it is because consciousness is a classical information-processing phenomenon. In other words, we can only directly perceive things that can be described as real numbers rather than complex numbers because we are Turing machines, and Turing machines are classical.

That's an interesting thought, especially with the addendum of the no-cloning theorem.

This fits with a Hofstadter-like perspective that consciousness is about "strange loops", where we somehow repeatedly evaluate simplified models of the world including yourself. Doing such repeated evaluations requires (at least partial) "cloning" of the state of the world outside for the purpose of evaluation, and cloning quantum states is impossible, hence consciousness must be a classical phenomenon.

I like that line of argument.

Note: I wouldn't state it as being unable to perceive things that can be described as complex numbers, but rather complex linear combinations.

Then wouldn't an appropriately constructed sensor be able to record the full linear combination of quantum states?

No, because records are necessarily classical. This is because of the no-cloning theorem. You cannot copy quantum information, only classical information.

Wait a minute, since when is there any difference between Many Worlds and Decoherence? I always assumed they referred to the same interpretation, do they not?

There are actually three different ideas in play here: decoherence as described by Zurek, the relative-state interpretation as described by Everett, and the popular account of the universe "splitting" when a measurement is made. All three of these are different, at least rhetorically, even though at a high enough conceptual level they all amount to the same thing.

I have never understood what the classic Many World Interpretation actually is. Splitting of universes has the same problem as collapse and probabilities become rather nebulous since presumably all possibilities do happen, after all.

But there is a version of it which does make sense: https://arxiv.org/abs/0903.2211

The basic question for all theories is, "what is fundamentally real?" and the wave function of quantum mechanics just doesn't work in that role. This is why Bohmian mechanics adds particles; that is what is real in that theory.

But to make a MW theory, one can integrate out the wave function to create a mass density function on 3-space. In any given instance, this is a mess and useless. But if you watch it evolve, then you could see multiple different stories evolving. And those are the different "worlds".

There is no splitting, just regimes that are no longer relevant.

> I have never understood what the classic Many World Interpretation actually is.

There's little to understand: at its core, Many Worlds is real close to "shut up and calculate". The only significant assumption is that the wave function as described by the math is real. And when the math says there are 2 non-interacting blobs, well, this translates to 2 universes.

That's about it.

As for why people seems to reject it instinctively, it's probably because our subjective experience is linear —or mono-threaded. However, linear histories aren't contradictory with trees. Imagine a Git History that never merges: each leaf is the product of a linear set of modifications, a perfectly clean history. From that final commit's perspective, other branches might as well not exist. You'd need merges to break that illusion, and our universe doesn't have macro-merges —we only observe them at a very small scale, for instance with photon interference.

So the wave function, a function say on configuration space, is real. And that's it?

But what does that mean? What is the mapping to my experience? Is it that if the wave function is non-zero at a configuration point, then the configuration is realized as an actual universe? One big problem with this is that wave functions can be non-zero at all configuration points and the dispersive nature of the Laplacian tends to make that happen.

So then it becomes one of size of |psi|^2? Is there a cutoff for considering that to be real? Having a probability of something is needed to deal with this, but it is not clear to me what the probability would correspond to here.

In BM, the probability is that of finding particles in some region. In collapse theories, it is the probability of collapsing to a specific state.

As for the Git History, my above concern could be phrased as a history in which all possible streams of texts are possible though maybe some have a large font-size. This is library of babel kind of stuff. How is the evolution of the text that I consider myself to be a part of handled? How are nearby configurations connected by the evolution of a wave function?

Also, macro-mergers are not theoretically ruled out; it is practically ruled out.

Let's say that I convinced you that 2 non-interacting blobs did not objectively evolve to occur, but rather a smearing over all. Would this break your interpretation?

> Let's say that I convinced you that 2 non-interacting blobs did not objectively evolve to occur, but rather a smearing over all. Would this break your interpretation?

If I understand correctly, what you speak of should actually be observable —you could make falsifiable predictions about it. That would break more than my interpretation, I think.

I didn't say that well. I mean, non-zero everywhere. There can be bumps with most of the support, but integrating |psi|^2 over any region gives a non-zero result, if mostly below the order of, say 10^-100 outside of your two bumps. That is, from a probabilistic point of view, it can be discounted without a thought, but from a "this is a real world" it does not seem to me to be so easy to discard.

That is to say, where is the magnitude of the wave function being used in this interpretation?

> Imagine a Git History that never merges

That is a really excellent analogy. It is also important to understand that there is a reason we don't have "macro merges", and that is because some irreversible process is required to create classical information, which is what we (the entities having this conversation) are actually made of. See:

http://blog.rongarret.info/2014/10/parallel-universes-and-ar...

Doesn't that also require at least one additional dimension of spacetime to separate the two universes?