Warning: Only read this if you’ve already read UNSONG, since this contains major spoilers.
Done reading UNSONG?
(That means you too, Brian. Don’t read beyond this point.)
So let’s assume The Unsong theory of theodicy: The reason the world has evil in it is that God creates every possible world where the net amount of Good exceeds the net amount of Evil. What would a typical world look like?
Well, let’s start by assuming most worlds are kind of like ours. We have a planet, civilization evolves, eventually it reaches technological maturity. Let’s assume this civilization inevitably creates AI, and in almost all cases this AI is unfriendly. Is this universe consistent with theodicy?
Answer: quite possibly. For example, imagine a paperclip AI, as described in this article. As the article points out,
Plus, if we teach the AI to enjoy making paperclips (and some say these sorts of human-analogous incentives will be necessary to create true thinking machines) then at least it’ll be having a fun time.
In fact, if that AI turns the whole observable universe to paperclips and has no upper bound on the amount of joy it can feel, we’ve created a nigh-infinite source of joy. No matter how much misery humans experience in our thousands or millions of years as a species, the AI feels enough joy to counterbalance that. So while we might find theodicy reassuring in principle – the net balance of the world must be Good! – that doesn’t neccessarily mean we won’t be incredibly miserable.
But wait! it gets worse! Take a look at this graph:
This is a bell curve. It’s what happens when we have a lot of independent factors that can move things in either direction. That part isn’t new.
If we imagine the net Good in all possible universes (without theoditic intervention), we expect it to have a bell curve distribution, with neutral worlds in the middle1. After all, there are an unimaginable number of tiny independent factors that go into deciding how much good and evil goes into a potential universe. If we imagine theoditic intervention, God chops off the left half of the bell curve – everything to the left of the middle vanishes, and everything to the right stays in existence.
Now the key thing to remember about bell curves is that they are incredibly concentrated around the middle. The textbook pictures are incredibly misleading. They make it look like it’s a fairly wide distribution in the range. But it actually decays superexponentially. That 99.7% within three standard deviations of the mean? A standard deviation is on the order , where M is the total variance of the universe and n is the number of factors. Since there are an unimaginably huge number of factors that go into creating a universe, that number is tiny. Almost all possible universes are barely breaking even. And if we think of civilization as gradually growing and improving, that means we’re almost certainly at a net loss right now – the current level of misery on the planet exceeds the amount of joy, and if it weren’t for the fact that we could improve it would be better to never have existed. And even when we do make things better, it’ll be just barely enough to break even with all the evil that happened before.
So even if God only allows the good universes to exist, from the outside view, our universe is probably kind of meh.
1. It might be argued that the mean random universe wouldn’t be neutral. It seems like a totally chaotic world might be evil. Unfortunately, this just means the distribution would be even more loaded towards the break-even point once we eliminate everything to the left of it. Conversely, if the mean random universe is to the right of the break-even point, we’re doing better. But that seems less likely.↩