Friday, July 06, 2018

How to Create Immensely Valuable New Worlds by Donning Your Sunglasses

[A satisfied REI customer, creating whole gobs of new worlds.]

One world is good, so two is better, right?

(If you think one world is bad and two are worse, just invert the reflections below.)

Here's one way to look at it: Run a universe, or a world, from beginning to end, sum up all the good stuff, subtract all the bad stuff, and note the (hopefully positive) total. Now consider, from your end-of-the-universe God's-eye point of view: Should you launch another world similar to the previous one? Well, of course you should! There would be even more good stuff, and a higher positive total, after that second world has been run. Similar considerations suggest that two good worlds running in parallel would also be better than a single good world.

(To avoid problems with summing infinitudes, let's assume finite worlds with finite value. For some complexities regarding the value or disvalue of repetition specifically see this earlier post.)

Now on some interpretations of quantum mechanics, every time there's a quantum event with different possible outcomes, all of the outcomes occur, each in a different world. Such "many world" interpretations often describe the world as "splitting" into two worlds -- one world in which Outcome A occurs and one in which Outcome B occurs. You too, the observer, will split: One copy of you goes into World A, observing Outcome A, and the other goes into World B, observing Outcome B.

In a classic article on the many-worlds interpretation of quantum mechanics, Bryce S. DeWitt writes:

This universe is constantly splitting into a stupendous number of branches, all resulting from the measurementlike interactions between its myriads of components. Moreover, every quantum transition taking place on every star, in every galaxy, in every remote corner of the universe is splitting our local world on earth into myriads of copies of itself.... I still recall vividly the shock I experienced on first encountering this multiworld concept. The idea of 10^100+ slightly imperfect copies of oneself all constantly splitting into further copies, which ultimately become unrecognizable.... (DeWitt 1970, p. 33).

Notice that on DeWitt's portrayal, there are a finite but large number of worlds: 10^100+. Now here's the normative question I want to consider: Is there positive value -- would it be ethically or prudentially or aesthetically good -- to increase the amount of splitting, so that there are more worlds rather than fewer?

On the face of it, it seems that, yes, it would be good if there were more worlds. If each world independently considered is good, then plausibly more worlds is better. Suppose World W has positive value V. Suppose now that it splits into two worlds that are very similar but not identical: W1 with value V1 and W2 with value V2. If we assume V1 ~= V2 ~= V and that the value of worlds is additive, then after the split the whole W1+W2 is approximately twice the value of W. We have doubled the amount of value that the cosmos as a whole contains!

Now you might object in one of three ways:

(1.) You might reject the whole splitting-worlds interpretation. Fair enough! But then you're not really playing the game. I'm interested in thinking about the normative consequences assuming that the world does split as DeWitt describes.

(2.) You might reject the assumption that V1 ~= V. For example, you might think that after splitting, each world has only approximately half the value that the world before the split had, so that V1 + V2 ~= V. Then no value would be gained by splitting. A tempting thought, perhaps. But why would splitting make a world half as valuable? It's not clear. One consequence of this view is that our world, constantly splitting, would be constantly halving in value, so that its value is plummeting by orders of magnitude every second. Hmmm, that seems no less bizarre than the idea that splitting doubles the value of cosmos. Now if you thought that the other worlds already existed before the split, then you might reject V1 ~= V, since the subset of worlds in V1 would be about half (or whatever) of the total number of worlds in V. But the whole idea of splitting worlds is not that the worlds already existed. It's that they are created. So we're talking about whether creating a new valuable world adds value to the cosmos as a whole. Pending a good argument otherwise, it seems like it should.

(3.) You might reject additivity. You might say that although V1 ~= V2 ~= V, you can't simply sum V1 and V2 to get ~= 2V. I can feel the pull of this idea. If the worlds were strictly identical, you might say "well, there's no point in duplicating everything again"! But (a.) The worlds are not strictly identical, and in fact (given chaos theory) they might diverge increasingly over time. And (b.) Duplication plausibly adds value when worlds are temporally separated (at the end of the universe, it seems like not re-running the thing again would involve omitting possible future pleasures that future inhabitants of the world would have), and side-by-side splitting wouldn't seem to be very normatively different in principle.

Okay, let's pretend you're convinced. When new worlds arise through quantum mechanical splitting, that's terrific. Whole realms of value are added! The value of the cosmos as a whole approximately doubles. Now, accepting this, is there anything you should do differently?

Consider the following possible principle:

Conservation of Splitting. Over any time interval t, the world will split N times, no matter what you do.

As far as I can see, there's no reason to accept Conservation of Splitting. It seems like you can create situations in which there would be relatively more splitting or relatively less splitting. You can run more quantum mechanical experiments or fewer. You can make more quantum mechanical measurements or fewer. And if you run more experiments and make more measurements, there will be more splitting, and thus more worlds.

For example, that pair of polarized sunglasses you have, sitting in that dark sunglass case in that dark drawer? When a photon passes or fails to pass through a polarized lens, that's a quantum mechanical event -- a chance event, which, on this interpretation, results in a splitting of worlds. In some worlds the photon goes through; in others it does not. You could take those sunglasses out of the drawer. You could go to the beach and wear them in the sun. Many more photons will pass through those lenses! Maybe about 10^18 more photons per second. If each of those photons has an independent quantum chance of passing or not passing through those lenses, splitting the world, then you're creating 2^10^18 new worlds per second just by sitting on the beach -- worlds that would not have been created had the sunglasses remained cased in the drawer. Think of all the value you're adding to the cosmos!

Effective altruism is a movement that recommends using reason and evidence to do the most good you can do. Normally, effective altruists recommend doing things like donating money to charities that effectively help to alleviate suffering due to poverty. But the value of saving one life has to be tiny compared to the value of creating 2^10^18 new universes per second! So instead of staying indoors in the shade, writing a check to the Against Malaria Foundation, maybe you'd do better to spend the day at the beach.

Even if you have only a 0.001% credence that the splitting worlds interpretation of quantum mechanics is true, the expected utility of sitting on the beach might far exceed that of donating to poverty relief.

I know that it doesn't seem intuitively very plausible that going to the beach is far morally better than donating money to effective, life-saving charities, but consequentialist philosophers are often willing to admit what's ethically best might not match our intuitions about what's ethically best.

PS: I feel so bad about making this argument that I just donated to Oxfam.

--------------------------------------------------

Related Posts:

Duplicating the Universe (Apr 29, 2015)

Goldfish Pool Immortality (May 30, 2014)

How Robots and Monsters Might Break Human Moral Systems (Feb. 3, 2015)

David Duffy said...

I think I have read too many quantum ethics papers - surely you don't have to feel guilty because some other version of you has already donated to Oxfam?

Stephen Wysong said...

For those of you who might object to the Everett Many Worlds interpretation, note that it's apparently widely accepted by the scientific community. But the more likely scenario is the Level 1 multiverse, per Max Tegmark:

“The simplest and most popular cosmological model today predicts that you have a twin in a galaxy about 10 to the 1028 meters from here. This distance is so large that it is beyond astronomical, but that does not make your doppelgänger any less real. The estimate is derived from elementary probability and does not even assume speculative modern physics, merely that space is infinite (or at least sufficiently large) in size and almost uniformly filled with matter, as observations indicate. In infinite space, even the most unlikely events must take place somewhere. There are infinitely many other inhabited planets, including not just one but infinitely many that have people with the same appearance, name and memories as you, who play out every possible permutation of your life choices.”

Variants of Tegmark’s articles (PDF’s) on the Multiverse subject can be found here:

http://space.mit.edu/home/tegmark/PDF/multiverse_sciam.pdf

http://space.mit.edu/home/tegmark/multiverse.pdf

Because there are a finite number of ways to arrange particles, the repetition of You in the Level 1 multiverse (this very universe, not some other one) is a matter of simple probability in a sufficiently large universe and is, consequently, much more likely the case than the Everett proposal. Unfortunately, in the Level 1 case, wearing polarizing sunglasses only means that it’s extremely likely that a few millions of your doppelgängers will be wearing them too, but won’t add to the number of universes as might be the case with the Many Worlds hypothesis.

Carl said...

This post seems to give 'crazy' short thrift to the Born rule. Our memories, material surroundings, biology, and technology are consistent with the Born rule statistics, but not with the 'every branch counts equally regardless of its measure' rule. That's why physicists who endorse many worlds interpretations are only interested in accounts along the lines of #2, so as to reproduce the Born rule, e.g.:

http://www.preposterousuniverse.com/blog/2014/07/24/why-probability-in-quantum-mechanics-is-given-by-the-wave-function-squared/

BTW, effective altruists have considered this argument for world-splitting and generally rejected it for the above reason.

Anonymous said...

I would be very cautious. Whatever probability you give the multi world hypothesis, I would give just as much credence to the view Lewis describes in "How many lives has Schrödinger's cat?" (http://www.andrewmbailey.com/dkl/How_Many_Lives.pdf)

But on that view every conscious being is doomed to quantum immortality in horrible agony, and if you're doubling worlds you're doubling conscious beings. Better throw those glasses away quick!

Hi Eric! I discovered you through a psychologist/philosopher friend of mine, and I think you are one of the most promising philosophers of mind. I´d like to see you take on Karl Fristons Free-Energy principle/Bayesian Brain/Predictive Coding framework. I´ve yet to see much critique from philosophers. Those philosophers who are on the bandwagon (Clark, Howhy) propose that the hard problem will dissolve within this framework.

Do you have any thoughts on this?
Best,

Ben

Eric Schwitzgebel said...

David: Is it possible to read too many quantum ethics papers?!

Stephen: I agree that Tegmark's Level 1 multiverse also reasonably likely. It has other weird implications, but not this particular weird implication! (By "reasonably likely" here, I only mean worth considering as a possibility, to consider what the philosophical consequences would be.)

Carl: I'm fine with the Born rule, but I think that's a side issue here. The Born rule doesn't even attempt to address the utility or not of creating worlds, if creating a world is what happens upon splitting. It addresses only the relative weights to give to different branches post-split. (I have an extended exchange about this with Sean Carroll and others on the Facebook link to this post. I don't know if I convinced anyone, though.)

Ben: Thanks for the kind words! Predictive coding is cool, but I tend to be a late adopter of hot trends in neuroscience, since my experience is that they often look a little silly ten years later (e.g., mirror neurons). So predictive coding is in my quarantine until, say 2030 or so. Check back with me then! :-)

Eric Schwitzgebel said...

Whoops, missed Anon: Yes, that's the inversion I was thinking of, if you think creating worlds is bad!

Carl said...

"The Born rule doesn't even attempt to address the utility or not of creating worlds, if creating a world is what happens upon splitting. It addresses only the relative weights to give to different branches post-split."

Say that I anticipate decoherence of my branch into A and B, each of which will have equal amplitude, followed by the decoherence of B into equal amplitude branches C and D.

MWI that recovers Born statistics says that I should prospectively value happiness in A and B equally, and twice as much in A as in either C or D. So the decoherence of B into C and D can't double total utility unless it also doubles the utility of A. But when B decoheres into C and D it has already decohered from A: if A is physically unaffected by the C/D split, why would A's value double?

Eric Schwitzgebel said...

Carl: The value of your *happiness* wouldn't double. However, you would have done an immensely valuable thing in B that you didn't do in A: created a new world. There would also now be more counterparts or continuers of yourself in the B -> C & D branch than in the A branch, and you might value that.

Suppose I value the creation of every new world at \$10. You can impoverish me by giving me an opportunity to pay to create worlds -- but that's decision-theoretically rational if I do value creating the worlds at \$10 each. But can you Dutch book me? I don't see how. Ordinary violations of the Born rule in decision theoretical settings can be shown to be irrational due to Dutch booking. So this isn't an ordinary violation of the Born rule.

Paul Christiano said...

Note that the existence of the whole wave function seems to be the dominant view amongst people with a good understanding of quantum mechanics, but the world-splitting interpretation is quite fringe. (Also note that on a sensible version of the world-splitting interpretation, the amount of world-splitting is probably a constant independent of our decisions, and what we can most easily influence is how much world-*merging* occurs. Though frankly, if you don't care about the amplitude of possible worlds, should you even treat worlds with 0 amplitude differently? At that point it seems like the only reasonable view is that all possible worlds exist and who cares about the flow of amplitude between them...)

The Born rule is a pretty strong argument against caring equally about all worlds. And if you grant that you don't care equally about all worlds, then it seems like you should be open to the possibility that an action can lower a world's moral weight, offsetting the creation of a new world.

Or, if you like, a splitting action destroys the old world and replaces it with two new worlds. Given that we've already granted that all three worlds can have different weights, why think that the sum of the two new weights is larger than the original weight? The Born rule says that the weight of the two new worlds sums up to precisely the weight of the original world.

Eric Schwitzgebel said...

Thanks for the comment, Paul! I'd be interested to hear more about the commonality of the view that the whole wave function exists vs the world-splitting interpretation. In my experience, Everett / many-worlds interpretations aren't super clearly articulated ontologically; and one conclusion that I'm inclined to draw from my argument is that there will be some wild consequences if one takes the world-splitting talk in a certain type of serious ontological direction.

On the Born rule: Prospectively I care more about what will happen in a high-amplitude world than a low-amplitude world. If my money can only go to one, I want it to go to the high-amplitude world. In that sense I care more about that world. But on a world-splitting view, once the world is there, with X number of human lives, and Y number of awesome paintings, and etc. etc., is the world that was low-amplitude from the earlier perspective somehow a less valuable world than the one that was high-amplitude from that earlier perspective? If I find out that some hugely improbable event just happened in the recent past, should I say "damn, this world is now so much lower amplitude and worth so much less!"? In deciding about the future, you need to follow the Born rule not to be Dutch-booked, but I don't see how it follows that this is a good way to value worlds that already exist.