Here's a funny reformulation of your argument: if you live in a quantum world where deaths are swift and painless, it makes sense to bet a lot of money on the assertion that you will stay alive. This incentivizes many other people to bet on your death and try hard to kill you. The market takes this into account, so your payoff for staying alive grows very high. Sounds like a win-win situation all around!

That said, at root it's just a vanilla application of quantum immortality which people may or may not believe in (the MWI doesn't seem to logically imply it). For a really mindblowing quantum trick see the Elitzur-Vaidman bomb tester. For a deeper exploration of the immortality issue, see Quantum suicide reality editing.

However there is a way to convince a lot of people in a few Everett branches: You make a one-time big announcement in the Internet, TV etc. and say that there is a well tested quantum coin-flipper, examined by a community consisting of the most honest and trusted members of the society. You take some random 20 bit number and say that you will flip the equipment 20 times and if the outcome is the same as the predetermined number, then you will take it as a one to million evidence that the Multiple World theory works as expected. Of course, only people in the right branch will be convinced. Nevertheless, they could be convinced enough to make serious thoughts about the viability of quantum Russian roulette type games.

I vaguely remember this being discussed here before, and people deciding it wouldn't work. Before the coin-flipper is run, you have a 1/2^20 chance of seeing your number come up, whether many worlds is true or false. That means that seeing the number come up doesn't tell you anything about whether MW is true or not. It just tells you you're extremely lucky: either lucky enough that the coin-flipper got a very specific number, or lucky enough to have ended up in the very specific universe where the flipper got that number.

You take some random 20 bit number and say that you will flip the equipment 20 times and if the outcome is the same as the predetermined number, then you will take it as a one to million evidence that the Multiple World theory works as expected.

That doesn't convince anyone else; from their perspective, in Bayesian terms, the experiment has the same million-to-one improbability of producing this result, regardless of whether QTI is true, since they're not dying in the other worlds. From your perspective, you've ended up in a world where you experienced what feels like strong evidence of QTI being true, but you can never communicate this evidence to anyone else. If we hook up a doomsday device to the coinflipper, then in worlds where we survive, we can never convince aliens.

Do you really need Many Worlds for quantum suicide/immortality? How's a Single Big World (e.g. Tegmark's Level 1 universe) different? Since a Single Big World is infinite, there will be an infinite number of people performing quantum suicide experiments, and some of them will seem to be immortal.

It seems to me that there is no practical difference between living in Many Worlds vs a Single Big World.

If you assume 3, average utilitarianism says you should kill anyone who has below average utility, since that raises the average. So in the end you kill everyone except the one person who has the highest utility. There is no need for assumption 2 at all.

BTW, are you aware of any of the previous literature and discussion on quantum suicide and immortality?

See also http://www.acceleratingfuture.com/steven/?p=215

If the worlds in your MWI experiment are considered independent, you might as well do the same in a single deterministic world. Compare the expected utility calculations for one world and many-worlds: they'll look the same, you just exchange "many-worlds" with "possible worlds" and averaging with expectation. MWI is morally uninteresting, unless you do nontrivial quantum computation. Just flip a logical coin from pi and kill the other guys.

More specifically: when you are saying "everyone survives in one of the worlds", this statement gets intuitive approval (as opposed to doing the experiment in a deterministic world where all participants but one "die completely"), but there is no term in the expected utility calculation that corresponds to the sentiment.

I think there will probably have to be some set of worlds in which the losers of the game are alive but near death (it really is necessary to specify the means by which they die). So this really is a gamble since participating means there is a very slight chance you will wake up, 50,000 dollars poorer and needing an ambulance. To figure out the overall average utility of the game one would need to include the possible worlds in which the killing mechanism fails. Average utility over the universal wave function would probably still go up, but there would be a few branches where average utility would go down, dramatically. So the answer it would depend on whether you were quantifying over the entire wave function or doing individual worlds.

Or were you ignoring that as part of the thought experiment?

EDIT: I just thought it all out and I think the probability of experiencing surviving the killing mechanism might be 5 in 6. See here. Basically, when calculating your future experience the worlds in which you survive take over the probability space of the worlds in which you don't survive such that, given about a 5 in 6 chance of your death there is actually about a 5 in 6 chance you experience losing and surviving since in the set of worlds in which you lose there is a 100% chance of experiencing one of the very few worlds in which you survive. This would make playing Quantum Russian roulette considerably stupider then playing regular Russian roulette unless you have a killing mechanism that can only fail by failing to do any damage at all.

With a similar technique you can solve any NP-Complete problem. Actually, you can solve much harder problems. For instance, you can minimize any function you have enough computing power to compute. You could apply this, for instance, to genetic algorithms, and arrive at the globally fittest solution. You could likewise solve for the "best" AI given some restraints, such as: find the best program less than 10000 characters long that performs best on a Turing test.

like it.

I'm developing a web game called 'quantum roulette' for the website of my upcoming feature Third Contact, and I happened on this thread.

the game will be at www.thirdcontactmovie.com at some point. You might find the film of interest too. Love the discussion.