Darwinbots Forum
General => Off Topic => Topic started by: Numsgil on April 30, 2006, 05:21:00 PM

Assuming that the universe is run with a fixed time step (23*10^42 units per second) and discounting any relativity effects, and assuming that there are 2.7 * 10^83 particles in the universe to calculate, and assuming these particles can be simulated with an n log n run time, I compute the power of a computer capable of simulating the universe at 10^125 FLOPS.
Given that a supercomputer now days is 280.6 * 10^12 FLOPS, and using Moore's law that computing power roughly doubles every 24 months (and assuming this growth rate is constant into the foreseeable future, which I doubt it is) we will have a supercomputer 741.16 years from now capable of simulating the entire universe in real time.

You really worry me sometimes bud...

You're like the 5th person to tell me that today
Hehe

Assuming that the universe is run with a fixed time step (23*10^42 units per second) and discounting any relativity effects, and assuming that there are 2.7 * 10^83 particles in the universe to calculate, and assuming these particles can be simulated with an n log n run time, I compute the power of a computer capable of simulating the universe at 10^125 FLOPS.
Given that a supercomputer now days is 280.6 * 10^12 FLOPS, and using Moore's law that computing power roughly doubles every 24 months (and assuming this growth rate is constant into the foreseeable future, which I doubt it is) we will have a supercomputer 741.16 years from now capable of simulating the entire universe in real time.
Umm...You were saying, Nums? I don't follow the math here, but... I've heard there is a limit how fast you can make a computer. I've heard they are close to the limit and they need to invent new types of circuits to make the computers more powerful. Some even fear the computer market will come to a halt because of it. Hmm... In 700 years from now I expect this world to be dead anyway, seeing as how we treat it.

I make no claims that Moore's law is likely to hold. It's mostly just some interesting math.

Moores law recently became more jumpy meaning have we have periods of now grow and then sudden inventions(like wit evolution ). So lets first wait until quantum computers run , which will be a big jump until we assume were the border is (Which is currently limited trough the fact you can not make chip structures smaller than molecules, or better the limit is before that as you have to deal with quantum effects etc if you go on molecule size)

Most industry analysis still sees Moore's law as holding through the next several chip generations, but you're right, there does seem to be some fundamental design hurdles involved with miniaturization.
However, a simulation of the universe would be quite parallelizable, which I think is where the new frontier in CPUs is going to take us.

Hmm ... wouldn't it be interesting to run a really massive DB evosim on a supercomputer?
I don't suppose anyone has access to one

Anyway a nice introduction to the The Quantum Computer (http://www.cs.caltech.edu/~westside/quantumintro.html). And I think he is danish...

it would rather be 550 years from now

Why 550?

cause it double by 1,5 years this days, not 2

I've heard people say that, but all the hard evidence seems to support 24 month doubling. Moore's initial article also said 24 months.

okey, but you dont have to worry who of us will win
cause this world wont exist then

The question is kind of moot anyway since any computer which is doing the modelling must also be a subset of the universe in which it exists.
If that is the case then it would be necessary for the computer to have as many memory locations as the universe has particles. Even in a computer which only requires one particle of computer memory to hold a value (somewhat impossible since how many states can a particle be in?) and not counting the parts of the circuits which are not directly involved in memory storage, the computer would have to contain at least as many particles as the universe which it is modelling.
Effectively the computer would actually be the universe. Oh wait a minute! Isn't that the way it might well work right now?
And they say computers are getting smaller

It is kind of odd. There would presumably be an upper computational limit of the amount of material you're using for your computer. As you say, any computer simulating the universe is also simulating itself, which creates a kind of paradox.

It is kind of odd. There would presumably be an upper computational limit of the amount of material you're using for your computer. As you say, any computer simulating the universe is also simulating itself, which creates a kind of paradox.
That's why we just need to catch Laplace's demon and keep him in a basement hooked up to a computer display.

I found something similar to the simulating the universe idea
Omniscience through counting
All knowledge in the universe can easily be displayed on your computer screen. How? I hear you cry (as well you might). It is quite simple, and requires nothing more than a computer than can count (which is what they are best at). Imagine, if you will, and 8x8 grid of pixels. This grid (or bitmap) contains 64 pixels, each of which may be 1 or 0 (on or off, black or white). The total number of combinations of on/off pixels is 2^64 (which is an extremely large number). This simple 8x8 grid can easily hold such images as all the letters of the alphabet, the digits 09, pictures of space invaders and PacMan, different shades of grey and so on. To display all the possible combinations simply requires the computer to treat the 8x8 grid as a 64 bit binary number (1x64 instead of 8x8). If the computer starts at zero and counts up to 2^64, displaying the binary number in the format of an 8x8 bitmap, it will show all the possible combinations of pixels that there are (including the abovementioned letters and Pacmen).
If we expand this concept a little, it is easy to see that an entire computer screen may be treated as a single binary number. If your display is 1280 x 1024 pixels (quite a common size) then it could be treated as a single 1310720 bit binary number. It can store all the possible numbers from 0 to 2^13107201 (which is more than the number of particles in the universe).
If you set your computer (admittedly, it would have to be quite a fast one) counting, from 0 up to 2^13107201, it would display all the knowledge in the universe. Every frame from every movie, every page of every book, the DNA of every creature that has ever lived, the face of every person who has ever lived, the complete source of every computer program ever written, the entire life history of every single living organism, a picture of every square inch of every planet in the universe. Everything, ever would appear on your screen. There would be an awful lot of meaningless junk, but in amongst it will be all the knowledge in the universe.
And that's just with a single bitplane. If you treat a 24bit display as a single binary number, you'll get the same information, but in colour!

Hehe, now if only my computer wasn't dead.
Maybe I can just do it by hand...

If I understand Elite correctly, that whould just be a bunch of numbers all over the screen, right? I was thinking Nums meant a program that whould paint up the whole universe where you could access every solar system there is and look at the planets and suns. Very graphical intensive!
Hmm... Maybe I will live to buy a quantum computer someday. Otherwise the one I'm thinking of buying in september will probably be the last one.

that whould just be a bunch of numbers all over the screen, right?
No not a bunch of numbers.
You would get a picture for each binary number. ie. Pixels would be turned on or off according to a numerical code. The point is that every possible picture would at some time be displayed on your computer screen.