randombio.com | science commentary
Sunday, January 29, 2017

Reality, multiverses, and artificial intelligence

Pretending something doesn't exist doesn't make it go away. Or does it?

P retending something doesn't exist doesn't make it go away. We all learned this when we were two: if you close your eyes, others can still see you, and the world continues to exist. But it doesn't have to be that way.

One of my earliest memories was when I was in first grade, riding on a schoolbus, a trip that took 75 minutes each way. I was thinking about how, when I was just a little kid, maybe two or three, that when you closed your eyes for a minute and then opened them, objects were in different positions, as if they suddenly jumped to their new locations. How to explain that? I remember thinking that this was physically impossible, and that that must have been why I had concluded that the world had existence independent of my thoughts.

You know the bus route is too long when first graders start thinking about Piagetian developmental psychology.

I tell this story not to explain why the other kids were always beating me up, but to explain that this is a stage of development everyone goes through. The brain makes no assumptions about how the world works; even basic things like the permanence of objects and the existence of cause and effect have to be deduced empirically.

Guider crash Andromeda galaxy as might be seen from a different universe where telescope auto­guiders never work properly

We take these things as basic truths, but the brain does not; it could adapt just as well to a world where effect precedes cause and objects come and go as we think about them. We flatter ourselves that we use reason to discern reality, but in fact reality molds reason to itself. Logic and causation are not intrinsic properties of the mind, but of the universe.

Neural networks

Years later, when I was working on AI, I realized that intelligence is limited by the environment: if the environment is too simple, it is impossible to build a machine with a high intelligence.

In AI, your network learns by being exposed to a world. Usually you start with an artificial one in order to validate your theory. But it's not easy. For example, if you train a neural network with simple tic-tac-toe universes, it will never be able to understand a chess-like world (technically, it means the weights will never converge). In order to make it intelligent, it's necessary to build a complex world, with discoverable causes and effects. We could even say there's a funda­mental principle here: a mind can never be more intelligent than the universe it lives in.

That doesn't mean, of course, that an individual human IQ can be increased much, if at all, by the environment. But if there is no structure in the environment, it is impossible for complex ideas to form, and the brain will never live up to its potential. That principle holds for any system that adapts to its environment. If the world is simple and unchanging, a genome will remain simple and complex species will not evolve. A changing, complex environment creates complex creatures.

In practice this means that when you do a simulation of the brain you don't put the intelligence and wisdom in the brain. You have to put it in the universe, and you make the brain capable of discovering how to learn it.

A lot of unappreciated effort goes into creating a universe, even a little toy one. But unless I inserted myself into that universe, that network would never have a clue what I was. Moreover, it would be impossible for it ever to grasp the magnificence (such as it is) of its creator.

The question is, what is it about our so-called real universe that causes it to become “intelligent”? Philosophers and religious people have asked questions like this for centuries. Granted that we invent the concept of number, for example, to make sense of the universe, but does this mean the concept of number is inherent in the universe? It must be so.


But could there be a world where the concept of number is not implicit in its structure? We always assume that other universes are by definition unreachable. But it is only an assumption.

Crap Nebula The Crap Nebula, not as nice as our Crab Nebula in our universe

One of the least appreciated implications of the multiverse hypothesis is this: if there are an infinite number of universes, and therefore an infinite number of each of us, how do we reconcile that with our subjective certainty that our consciousness exists as a privileged reference point? How do we know that reference point (which we can call subjectivity) cannot just wander at will through those infinite dimensions? We don't. If it happened, we would never know the difference.

If the multiverse idea were true, we couldn't rule out the possibility that simply wishing the world to be different would make it appear so. If we wished for, say, a million bucks, we might simply move to a dimension where that million bucks is already in our bank account.

The multiverse hypothesis, then, makes causation irrelevant. It also implies that the world is not ‘real,’ since we would be able to change the past, present, and future at will, simply by hopping to a different dimension.

Physicists dislike the multiverse theory because it is not parsimonious; postulating that every decision creates an entire new universe violates every principle of conservation. It is about as non-parsimonious a theory as you can get.

But that doesn't mean it's impossible to disprove it. In fact, the much-maligned anthropic principle comes to the rescue here. If there are an infinite number of universes where any history happens and any physical rules apply, then there must also be an infinite number of universes where it is impossible to move to another universe. If by chance or by design you ended up in one, you'd be stuck there.

You have three wishes

The probability of getting stuck in such a dead-end universe is equal to the fraction of such universes in the multiverse. If it is, say, 25%, we could call it a three-wish multiverse: you would get, on average, three wishes, and then you'd be stuck.

There could never be any universe where the multiverse ceases to exist, but there could be some from which it is impossible to escape. In these dead-end universes it is also impossible to imagine other universes. Upon entry the rule that allows you to escape, and thereby prove multiverses exist, ceases to exist.

Maybe we are all like that teenager recently in the news who, playing in an abandoned prison, accidentally closed the cell door behind her. In that story the fire department came in and used plasma torches to get the girl out, thereby destroying her little pocket universe. Without external assistance, escape from a dead-end world would be very slow or very painful. Kind of like what we have now.

Related Articles

Does the theory of relativity prove that the world is a simulation?
On unfalsifiable tautologies in popular science.

Forging the universe
Why is there something instead of nothing? Science, religion, and philosophy have different ideas.

Without cause and effect, nothing would ever happen. But what actually is it?

Indeterminacy in Science
Is the universe indeterminate at the smallest level? If so, what about free will?

Is the universe mathematical?
We use mathematics to deconstruct the universe. Could it also build one?

On the Internet, no one can tell whether you're a dolphin or a porpoise
Name and address
book reviews