carado.moe

tammy's blog about
AI alignment, utopia, anthropics, and more;

also on twitter, lesswrong, rss, and at

a safer experiment than quantum suicide

in tegmark's page about multiverses, he mentions that you could experimentally test the validity of the many-worlds interpretation of quantum mechanics by running a machine that uses a quantum random number generator a decide whether to kill you, with overwhelmingly large odds of doing so.

in my perspective, the way this works is that if the many-worlds interpretation of quantum mechanics is true and anthropic juice reallocates itself to your surviving instances at any point in time (which is in my opinion the only way for quantum immortality to work, and might make sense if there is an objective fundamental arrow of time, which i think is plausible), then you still have a reasonable probability of observing existing after the experiment. if not, then almost all your anthropic juice is located before you can the experiment. based on that, a given local observer is able to gain bits of evidence about how the world works, in a manner similar to what the doomsday argument says about our chance of extinction.

however, if instances of anthropic reasoning are the things which are able to deduce information from their existence — as opposed to "observers" — as i suspect in anthropic reasoning coordination, then a much safer experiment could be done: instead of making it likely that the machine kills you, simply make it likely that the machine brainwashes you into being committed to never doing anthropic reasoning ever again — or simply have the machine tell you to make such a commitment, and then do it yourself, if you're able to stick to it. then, it is not you but you undertaking anthropic reasoning which is undergoing quantum suicide, and that's theoretically enough to gain the same bits of evidence as proper full quantum suicide.

(not that, in the grand scheme of things, an observer is expected to be able to do anything with that evidence — a reasonable understanding of how ethics juice relates to anthropics juice as well as weird causality in decision theory will probly function in a way that makes such evidence entirely unusable with regards to what actions to take to maximize something like expected utility in the one way that makes sense, from what i understand. at most, one can use that evidence to know how to feel about how things might be in the grander scheme of what's real, but not to impact what exists unless strange acausal shenanigans are at play)