It's more like, right now we have two rather bad explanations for basic quantum mechanics. There is the "particles change when a human measures them" explanation, and there is the "there are many worlds and we can usually only observe one" explanation. They both explain the same set of formulas that have been verified many times by experiment. Depending on what you want to do with the theory, it may be more intuitive to use one formulation than the other. Hopefully at some point in the future it becomes obvious that one of the formulations lacks the explanatory power of the other, and then scientists can cohere on one.

Like epicycles as a scientific explanation of how the sun revolves around the earth. Somewhat off topic but interesting reading - http://en.wikipedia.org/wiki/Deferent_and_epicycle

You're thinking about it wrong. Start from Bell's theorem - it tells us we have to sacrifice locality or realism. MWI weasels out of it by making worlds be observer dependent, basically splitting the observer up. Bell's theorem is the root. It's not an interpretation & all interpretations come from it.

I don't think it's testable -- that's why I'd call it an interpretation rather than a theory.

Then how is it different from counting angels on pinheads?

Many worlds is testable, though I wouldn't recommend it:

http://en.wikipedia.org/wiki/Quantum_suicide

Edit: :)

I don't buy quantum immortality as a necessary consequence of many-worlds. There are some rather dodgy steps in the logic -- in particular the one that says that "you" can't find your personal experiences going down a branch where your existence gets terminated. Surely somebody has to be there to subjectively experience having a bullet blast a hole in their skull?

Personally I think if you tried that, even if many-worlds is true, you'd almost certainly wind up dead.

That "you" can be you in other people's mind as described by Hofstadter in http://en.wikipedia.org/wiki/I_Am_a_Strange_Loop

Can't this experiment just be reduced to a succession of observations that are extraordinarily improbable? Why sex it up with the suicide attempts?

The key is to link up observer with the experiment outcomes so you can only observe unlikely possibilities (and be dead in the rest).

Understood, but you could also set up an apparatus that simply observes a large region for a long period of time without killing anyone, watching for highly improbable events. Since each observer also lives their total lifespan, you get a much higher probability of a given observer observing a succession of improbable occurences.

Just create a machine that reads bits from a random data source and stops forever upon reading a 0 bit. Will the device run forever in some universe?

If you built the device, and it didn't run forever, what could you learn from this? Nothing, except maybe how to build cool counting machines. :)

That what reporters are for :) If you read the news you will see an endless stream of highly unusual events, but unfortunately that doesn't prove anything.

Touche. :) But in the same vein, what would a series of failed suicide attempts prove that the mentioned apparatus could not?

Any machine that you build will produce expected results, up to 10^-15 precision with current hardware.

Failed suicide attempt is unlikely and provides evidence towards many worlds as a theory and not just interpretation.

Sorry, by apparatus I mean any method of observing a large body of matter or space that exhibit quantum behaviour (i.e. anything), not necessarily a physical machine. We are just looking for a series of highly improbable quantum state transitions.

Doesn't matter, any method of observation you choose will produce expected results.

Unless you observe yourself, where you will not be able to observe expected results because you're dead.

Think of each observer attempting to push a rock into their palm and having it appear on the other side, without enough force to push it through. Think of everyone on the planet attempting this simultaneously, thousands of times. If the rock disappears and reappears on the other side for one observer, doesn't this accomplish the same thing, at least for that observer, that the suicide attempt does?