[ExI] any exact copy of you is you + universe is infinite = you are guaranteed immortality

Eliezer S. Yudkowsky sentience at pobox.com
Fri Jun 15 01:58:07 UTC 2007


Suppose I want to win the lottery.  I write a small Python program, 
buy a ticket, and then suspend myself to disk.  After the lottery 
drawing, the Python program checks whether the ticket won.  If not, 
I'm woken up.  If the ticket did win, the Python program creates one 
trillion copies of me with minor perturbations (this requires only 40 
binary variables).  These trillion copies are all woken up and 
informed, in exactly the same voice, that they have won the lottery. 
Then - this requires a few more lines of Python - the trillion copies 
are subtly merged, so that the said binary variables and their 
consequences are converged along each clock tick toward their 
statistical averages.  At the end of, say, ten seconds, there's only 
one copy of me again.  This prevents any permanent expenditure of 
computing power or division of resources - we only have one bank 
account, after all; but a trillion momentary copies isn't a lot of 
computing power if it only has to last for ten seconds.  At least, 
it's not a lot of computing power relative to winning the lottery, and 
I only have to pay for the extra crunch if I win.

What's the point of all this?  Well, after I suspend myself to disk, I 
expect that a trillion copies of me will be informed that they won the 
lottery, whereas only a hundred million copies will be informed that 
they lost the lottery.  Thus I should expect overwhelmingly to win the 
lottery.  None of the extra created selves die - they're just 
gradually merged together, which shouldn't be too much trouble - and 
afterward, I walk away with the lottery winnings, at over 99% 
subjective probability.

Of course, using this trick, *everyone* could expect to almost 
certainly win the lottery.

I mention this to show that the question of what it feels like to have 
a lot of copies of yourself - what kind of subjective outcome to 
predict when you, yourself, run the experiment - is not at all 
obvious.  And the difficulty of imagining an experiment that would 
definitively settle the issue, especially if observed from the 
outside, or what kind of state of reality could correspond to 
different subjective experimental results, is such as to suggest that 
I am just deeply confused about the whole issue.

It is a very important lesson in life to never stake your existence, 
let alone anyone else's, on any issue which deeply confuses you - *no 
matter how logical* your arguments seem.  This has tripped me up in 
the past, and I sometimes wonder whether nothing short of dreadful 
personal experience is capable of conveying this lesson.  That which 
confuses you is a null area; you can't do anything with it by 
philosophical arguments until you stop being confused.  Period. 
Confusion yields only confusion.  It may be important to argue 
philosophically in order to progress toward resolving the confusion, 
but until everything clicks into place, in real life you're just screwed.

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence



More information about the extropy-chat mailing list