Wednesday, September 5, 2007

It’s simulations all the way down!

My buddy RT told me about something called The Simulation Argument. It's a simple argument consisting of a few reasonable statements which – when taken together – basically guarantee something that you probably won’t like.

It isn’t a proof – it’s just something to think about:

The Simluation Argument*
A) It’s possible to create a computer simulation of individuals with artificial intelligence.
B) An A.I. individual inside the simulation wouldn’t know that it’s inside a simulation – it would just be going about its daily business in what it considers the “real world”.
C) A civilization that did develop such a simulation would run it many – say millions – of times (for fun; for research, etc.)


That's it, that's all there is to it. What do you think - is any one of these statements totally unreasonable?

Well if they sound even somewhat plausible to you, then ask yourself this: which of the following is more likely?

  • That we are the one civilization mentioned in C) which happens to develop the ability to run A.I. simulations? Or,
  • That we are one of the millions of simulations that civilization C) has run?

Remember: we wouldn't know if we were in a simulation (point B); and there have been way more A.I. civilizations simulated than there have been "real" civilizations in the "real" universe (point C).

To put it another way, suppose there are 1,000,000 intelligent, self-aware civilizations in the universe…But only 1 of them is “real” while 999,999 are A.I.’s. Every one of them feels like it is “real” and lives in a “real” universe.

Are we the one in a million? Or just one of millions? Well, are you a betting man/woman/simulation?

If so, let's play a game. The game gives you just as good a chance of winning as we have of being the one "real" civilization mentioned in point C:

You and the Simulation Argument will both pick a number between 1 and 1,000,000. If the numbers match, then you win and we are the “real” civilization. But if the numbers don’t match, then the Simulation Argument wins and we’re just one of the many A.I. civilizations being run by the “real” civilization on their fancy computers.

Ready? Remember, the fate of humanity rests on your shoulders, here. Are you sure you’re ready?
If you are, then pick your number between 1 – 1,000,000 and click to see what number the Simulation Argument chose:

I AM READY TO SAVE HUMANITY


Good try!! Too bad about the fate of humanity…
...Do you feel any different? Are we any less “real”?

Switching gears for a moment, here’s an argument which isn’t quite as credible. I stole it from Stephen Hawking:

The Old Lady’s Argument
A well-known scientist (some say it was the philosopher Bertrand Russell
) once gave a public lecture on astronomy. He described how the Earth orbits around the sun and how the sun, in turn, orbits around the centre of a vast collection of stars called our galaxy
.
At the end of the lecture, a little old lady at the back of the room got up and said: "What you have told us is rubbish. The world is really a flat plate supported on the back of a giant tortoise
."
The scientist gave a superior smile before replying, "What is the tortoise standing on?"
"You're very clever, young man, very clever," said the old lady. "But it's turtles
all the way down!"

I like that little anecdote (clearly the old lady's position is ridiculous, but I’m curious as to why she chose turtles…)

Anyway, if you're wondering why I brought up both of these arguments in the same post…there’s a subtlety to the Simulation argument that really comes across as a kick in the, er, teeth:

Suppose that you accepted the Simlulation argument (or lost the number-matching game!), and we are just one of the millions of simulations that have been run by civilization C. Then at least we’re just one step away from the “real” universe, right?

…But when you accepted point A ('it’s possible to create a computer simulation of individuals with artificial intelligence'), you probably did so because we’re pretty close to developing such A.I. for ourselves…But remember - we've already figured out that we ourselves are just a simulation...which means that any A.I. entity we someday create won’t just be a simulation – it’ll be a simulation inside a simulation!

Sucks to be them!

But…this actually reveals the truly awful truth: odds are, we’re not just one of the simulations run by “civilization C”, after all. Odds are, we’re just one of the simulations run by a simulation run by a simulation run by a simulation run by a simulation (go ahead and repeat “run by a simulation” as many times as you like here) run by a simulation run by the “real” civilization C.

Ouch!


A well-known scientist (some say it was the philosopher Bertrand Russell) once gave a public lecture on astronomy. He described how the Earth orbits around the sun and how the sun, in turn, orbits around the centre of a vast collection of stars called our galaxy.
At the end of the lecture, a little old lady at the back of the room got up and said: "What you have told us is rubbish. The universe is really an A.I. simulation running inside another civilization’s computer."
The scientist gave a superior smile before replying, "What is
that
civilization’s universe like?"
"You're very clever, young man, very clever," said the old lady. "But it's simulations all the way down!”

The End.

PS: I suppose at this point I’d better say, “Thank goodness the old lady is crazy, the Simulation argument is ridiculous, and even if it weren't then clearly we are the one lucky civilization in a million million million million million which actually lives in the real universe.”
…But then again, http://en.wikipedia.org/wiki/Digital_physics#Overview.


* The Simulation Argument was
invented by Nick Bostrom.

6 comments:

Aaron said...

Its definitely an interesting theory, though not the first time I've been exposed to it.
I've heard about it relating back to sociology or philosophy but even still I'm sure the greeks and whoever else had their own versions.

Sometimes its just aliens with our brains hooked up to these diodes instead of a real computer generated "simulation" but the idea is always the same.

I have the largest gripe with the creation of the self-aware-intelligent AI systems. While I think the creation of AI is possible in "some" regard, I'm not sure that this level of complexity is attainable. That being said even if there was some kind of possibility of this being true as a software dev you should know that the program only does what its been "programmed" to do. Removing the ability for the AI to create these alternate AI systems would solve this dilemma pretty quick.

Maybe I'm too much of a skeptic for these kinds of arguments ... or maybe I'm too vain ;) who wants to admit they're just a simulation in a simulation in a ...

-2A

Johnny GoTime said...

I have the largest gripe with the creation of the self-aware-intelligent AI systems. While I think the creation of AI is possible in "some" regard, I'm not sure that this level of complexity is attainable.
One thing I should have been more clear about is that any simulation is not a completely accurate representation of the "programmer's" universe. In other words, if you and I create a simulation, then it certainly will omit many concepts that occur in our universe but are outside the scope of the simulation...For example, suppose objects in our simulation don't have mass. Obviously that's a huge departure from our universe, but it's totally distinct from the issue of whether or not our AI guys are intelligent. And to them, whatever world we've simulated for them will seem reasonable and consistent - because their entire notion of "reasonable" and "consistent" will be built on the rules of their simulated universe...

So to come back to your gripe: certainly the simulation won't be a 100% accurate portrayal of our own universe, but it will be logical and reasonable from the perspective of our AI guys.

By the same token, if you and I are a simulation, well, our universe is logical and reasonable to us because it's all we know. The "real" world programmers who wrote the simulation may have all sorts of complexity in their universe that we simply have no way to imagine, because the basic rules of our universe don't allow for them!

And by this point, you're probably annoyed at how unprovable this all is...And you're right! The argument is not a proof, it's just a "thing that makes you go hmm."

In fact, you might say it's a matter of faith...show me a more-concrete proof for the existence of God.

EasyRider said...

For the sake of argument, if it doesn't affect the way we choose to live our lives, what does it matter to us? However, for the sake of argument...

AI, despite being 'A', is still 'I'. Therefore, a simulated universe would still have to withstand constant observation. It has to look real at all times and in all places. remember the simulation in this example assumes that we are multiple AIs, not just one. Therefore, a realistic simulation could not (I would argue) afford to lose any information to compression. So, with the information ratio of the simulated universe compared to the actual universe needing to be 1:1 to be plausible, one would have to have fit an entire universe within an infinitesimal fraction of another one (the simulated one within the real one). Unless the universe possesses Buddhist physical properties, this is not possible.

Johnny GoTime said...

Hey easyrider, this isn't a cop-out but please see my response to aaron's comment. It boils down to this:

If we are a simulation, our universe will be missing many details/concepts that are present in the universe of the "programmers". Meaning that compression isn't an issue.

-J

Johnny GoTime said...

easy, one other thing just occurred to me:

A large amount of time has passed since the Big Bang. (Yes, this just occurred to me.)

But what this means is that the "programmers" could have started our simulation out on their old crappy computer hardware, and then added processors and storage every thousand years or so as their own technology improves to keep up with our expanding simulated universe...

This (in addition to my previous comment) could help mitigate the compression issue you raised.

Yorgos said...

Interesting read. Some observations:

The simuation definition from wikipedia: "Simulation is the imitation of some real thing, state of affairs, or process. The act of simulating something generally entails representing certain key characteristics or behaviors of a selected physical or abstract system."

According to that, a complex adaptive system (that by definition evolves in unpredictable ways) that is situated in a computer and has 0 and 1 (or if you wish "go" and "no go") as fundamental building blocks, is NOT a simulation. And it stops being a simulation right after you push the "RUN" button.

Exploring the notion of "simulation" we can also see tht it is an "intellectual concept". When you have a simulation of ants looking for food an taking it bakc to the nest, in reality you having nothing close to that. You just have a program that produces a pattern that your mind associates with another pattern (the one of ants). In one sentence, simulations do not exist outside our heads (the argument here could be "what does exist outside our heads?").

In the same line of reasoning, Artificial Intelligence is a very bad term. According to CAS theory, nothing is artificial in the whole universe. Human is a complex adaptive system that resulted from the evolution of other complex adaptive systems and so on, so any "artifact" made by human is just another reorganization of matter, energy and information that occurred due to the evolution of multiple CAS. Saying "I created Artificial intelligence" translates to "another CAS emerged". Nothing artificial about it (nothing artificial about anything).