In three and more dimensions, if you start a random walk outside some sphere there is a non zero probability that the event "the brownian motion enters this sphere" never happens.
Funnily, in dimension one and two this is not true.
(In fact I suspect he can never visit more than a vanishing subset of them, but I could be wrong, and I'm sure somebody has figured this out already...)
As you say, in one dimension you will visit all the points.
In two dimension the random walk is dense, so you pass arbitrarily close to every point in the plane.
As the dimension increases beyond two, it gets less and less likely that you hit the target sphere. In some way, space gets bigger and bigger as the dimension increases.
Suppose the d dimensional "target" sphere is very small and very close to the starting point of the path. Let some time pass, and suppose that the Brownian motion does not hit the sphere and has increased by 1 in every direction. Then the distance to the sphere is sqrt(d), which grows unbounded as d increases: if you don't hit the target immediately, on average you will be further away from it in higher dimension. So the probability of touching the target after that diminishes as the dimension increases.
I don't have an intuitive explanation of why 3 is the critical dimension though.
Are you sure of that? How would you prove it?
It seems intuitively right, but intuition has a habit of being wrong, when it comes to infinities.
Well, take the set of aleph-null length sequences (all aleph-null of them [1]) and put them in an order. Now construct the following sequence: invert bit 1 from sequence 1, invert bit 2 from sequence 2, ... etc.
Note that this sequence was never produced by the machine, since we enumerated all the sequences it produced, but they are all different to this one. So, by a standard diagonalisation, the set of events is higher cardinality.
[1] How many aleph-null length sequences come out of the machine? Well, they have to be continuous, so the only obvious aleph-null length sequence is the total sequence the machine produces. However, you can drop an entry from the beginning of the sequence to get another aleph-null length sequence. And so on, making a total of aleph-null.
Mostly, I'm unconvinced that an "event" which takes an infinitely long time to complete actually counts as an event. Usually when we talk about events they're localised in space and time.
This proof is dependent upon the lemma that time is infinite. On the other hand, the set of events we're looking for is the one where time ends. Therefore by assuming the lemma we have no need for the proof.
OTOH, if we're looking at another event that's not so confusing (one that's not the end of time), the proof holds water. It's a pretty sweet proof that I think is similarly used to prove that the set of real numbers is uncountable.
But there's no need to argue about the definition of event. I'll say instead that I'm interested in knowing whether infinite time implies that all finite-length events must eventually happen (and indeed, happen an infinite number of times as the paper claims).
However, the assumptions here seem tricky. Uniform distribution of probability seems unlikely, and if the universe is anything like the game of life, there are garden of Eden states with no predecessor state, which an evolving system cannot reach, even in infinite time.
Cardinality is indeed tricky. I had thought that all finite sequences would be greater than aleph-null, but apparently it isn't.
Order and Chaos need not be interchangeable.
There is an interesting book called "Sync" which discusses the presence of spontaneous order from chaos. He also talks about how order can turn back into chaos, but the important part of the book is that there are cases where chaos will not turn into order, and order will not turn into chaos, meaning that the stability of the two create two distinct sets, one that contains the identity of the universe and one that allows it to change.
Some of you may have read Wolframs work as well which shows that simple rules can lead to ordered stable systems. That stability is the definition of uniqueness.
Sync, it's the one book I recommend all my thinking friends read. http://amzn.to/a8s5Dl
It changed my life, and my world view.
So, I am comfortable saying that no act of randomness by any number of chimpanzees ( as they are now ) will randomly produce a work of William Shakespeare. WS, wasn't just a man, he was a man in a time with billions of events that influenced that moment when he composed those works. There was a structure and a path to those influences, and it wasn't such a frail construct that randomness happened upon it, but rather there was a beautify symmetry that occurred and allowed that to come to pass.
Another way to look at it, is that random/ chaotic signals will not sum up to collapse a bridge. however an ordered marching army has in many cased taken down the bridge.
Now that I think about it, we're talking about all possible events, which is even bigger. For any given event, there is theoretically an infinite amount of alternative possible events. Therefore, for each of the trillions of events at any given point in time, there is an infinite set of other possible events.
As far as I'm concerned having read the thread up to this point the question is still open... though my personal prejudice is still that it's not true that every event must happen.
Not exactly. Pi has no chance of occurring in an infinite set of integers.
This, of course, is impossible.
Proof: let X[n] be the set of numbers with non-zero probability of being produced at the n'th trial. X[n] must be countable, since sum(X[n]) = 1 and the sum of any uncountably infinite set of non-zero numbers must be infinite.
Let X = union(X[n], n=0...infinity). X is countable, being the countable union of countable sets. The reals are uncountable. Thus, most real numbers will NOT eventually be produced.
(It's true, pi in particular could be in X, but the vast majority of numbers could not be in X.)
However, I don't see how the set of numbers found in an infinite number of trials would be countable.
Edit: I googled some about this and can say I definitely learned some math today..
Count the first element from set 1, first element from set 2, second element set 1, second element set 2, first element set 3, third element set 1, etc. Eventually you will count every element from every set.
You can't do that for the real numbers.
"Among other things it depends crucially on an important assumption about the laws of physics: that we ought to be able to understand why they work, not just observe that they do work"
IANAP but,
In assuming that there are discrete laws of physics, we also assume there is a subset of all 'possible' events; events that 'can happen'. (Given a certain arrangement of the universe at T1, there is a set of possible arrangements of the universe at T2).
If we assume finite time, the set of events that will happen is a subset of the set of events that can happen. This is because the set of events that can happen is infinite and the set of events that will happen is finite. If both are infinite, then they are the same set.
Infinite time makes it apparent that all of the events that can happen will happen an infinite number of times, because events that can happen are events that will happen, and an event that can happen is never eliminated from the set of events that can happen by having been an event that will happen.
False. Possibilities can be mutually exclusive, so if one occurs, the other can't. Yet, after the fact, we considered them both to be events that could have happened.
(Unless you subscribe to absolute determinism, but I don't believe that's clear to be the case. Also, IME, discussions of determinism mechanisms in, e.g. quantum mechanics, tend to talk more about whether that's where Diety lives.)
3141592...
Edit: (Several people have said the same thing so maybe I can add something else)
I think another point of confusion might be the difference between "arbitrarily long" and "infinitely long". The number pi is infinitely long, a single integer can be arbitrarily long.
What's the difference? Pi has infinite digits. It is an irrational number. We can never "see" all the digits of pi, because no matter how far out we go there are always more digits.
With an integer, this isn't the case. There are infinite integers, but this property is defined on the set, not on any single element in the set of integers. Every integer (a single element in the set of integers) is finite. All of them, no matter what. They all have an end.
That said, what "arbitrarily long" means is that for any length you choose, there are integers that long (or longer). Say we choose a length 5, then 10000, 10001, 10002, ... all have length 5 or greater, but every one of them has an exact length. It just happens to be greater than 5. You can do this for any length: 1000, 10^1000, whatever. Anything. Choose any number as big as you can even imagine. Try to imagine numbers bigger than anything you can even imagine. There are integers that long, and infinitely more integers longer than that. But every single one of them has an exact, finite length.
[1] http://en.wikipedia.org/wiki/Non-standard_model_of_arithmeti...
log(infinity) < infinity
1. There are an infinite number of integers, but
2. Every integer is finite
Start counting 1,2,3,4... Do you ever get to the end? No, this proves that there are infinitely many integers. But do you ever get to an infinite number? No, you'll stay in the set of finite integers forever.
That's just plain mathematically wrong. There are many kinds of infinity (http://en.wikipedia.org/wiki/Aleph_number), and the set of all possible events has a much, much higher cardinality than an infinite axis of time. This is similar to saying that, since the set of integers is infinite, it must contain Pi.