What would you do with an infinitely fast computer?

[squareroot]

Ok, I'm probably making a fool of myself here, but I can't for the life of me convince myself that shutting off a simulation of the universe would destroy that universe.

My line of reasoning is this: a simulation is characterised by a set of rules for computing the (n+1)th state given the (n)th state (there might be complications if things are continuous, but I don't *think* that breaks the reasoning), and a set of initial conditions. By the very fact that the simulation can be computed up to some state n=k (whether k is finite or not is irrelevant to our purposes), all states previous to k in the simulation must be defined, which intuitively seems to be sufficient for any inhabitants of that simulation to continue to experience their universe as continuing to exist.

The problem that I have with the don't-unplug-it-or-you'll-destroy-the-universe line of thinking is that each of these states is clearly defined *before* the simulation is run. Hell, presuming that the rules and initial conditions can be expressed in some finite language (I'm not sure that this presumption is actually necessary, but nor am I sure that it's not, so 'better safe than sorry' comes into play), the simulation and all of its nth states for n<k can be said to, in some way, exist, long before anyone even codes the damn thing. I'd even go so far as to suggest that its existance wouldn't require any 'previous' (sorry, that's the wrong word for it, so I'm gonna have to rely either upon you guys being either telepathic or upon what I've written making enough sense that you can work out what I mean from the context, each of which is about as likely as the other) universe for it to be simulated within.

Presumably I've missed the point somewhere along the line, but I can't spot where

If I understand correctly, this was answered some 400 years ago, with "Cogito ergo sum"... just because the simulation will happen, doesn't mean it's valid. Maybe in 100 years, someone will simulate a version of Japan being mauled by Godzilla. Does that make any less fictional right now?

[TheChewanater]

This discussion reminds me of Source Code, where the machine is shut off and Stevens dies, but the simulated world still exists for some reason, implying it is a "reality" by itself.

Which is bullshit.

[undecim]

Ok, I'm probably making a fool of myself here, but I can't for the life of me convince myself that shutting off a simulation of the universe would destroy that universe.

My line of reasoning is this: a simulation is characterised by a set of rules for computing the (n+1)th state given the (n)th state (there might be complications if things are continuous, but I don't *think* that breaks the reasoning), and a set of initial conditions. By the very fact that the simulation can be computed up to some state n=k (whether k is finite or not is irrelevant to our purposes), all states previous to k in the simulation must be defined, which intuitively seems to be sufficient for any inhabitants of that simulation to continue to experience their universe as continuing to exist.

The problem that I have with the don't-unplug-it-or-you'll-destroy-the-universe line of thinking is that each of these states is clearly defined *before* the simulation is run. Hell, presuming that the rules and initial conditions can be expressed in some finite language (I'm not sure that this presumption is actually necessary, but nor am I sure that it's not, so 'better safe than sorry' comes into play), the simulation and all of its nth states for n<k can be said to, in some way, exist, long before anyone even codes the damn thing. I'd even go so far as to suggest that its existance wouldn't require any 'previous' (sorry, that's the wrong word for it, so I'm gonna have to rely either upon you guys being either telepathic or upon what I've written making enough sense that you can work out what I mean from the context, each of which is about as likely as the other) universe for it to be simulated within.

Presumably I've missed the point somewhere along the line, but I can't spot where

I agree. Stopping the simulation would not "destroy the universe".

When or how a universe is simulated doesn't matter. With any computer program, including the simulation in question, there is a deterministic outcome. User input or a random number generator could of course change that, but those could also be part of the program.

For example, rather than exactly 10 seconds after starting a program you press "a", you can simply program it to act as if a had been pressed after exactly 10 seconds. By creating a program this way, different inputs will make different programs. The results are the same, and so the simulation is the same.

The algorithm that simulates the universe still exists, even though it's not being run. I'm not referring to the simulation and data residing on a computer hard drive, either, or the source code sitting on a server. Even if those were completely erased, the algorithm would still run the same way if it were re-created years later. This is just like a mathematical object. Know one has calculated the 10^1,000,000th digit of Pi, but it still exists.

If I understand correctly, this was answered some 400 years ago, with "Cogito ergo sum"... just because the simulation will happen, doesn't mean it's valid. Maybe in 100 years, someone will simulate a version of Japan being mauled by Godzilla. Does that make any less fictional right now?

Imagine a fictional human who states "Cogito ergo sum". That person is incorrect.

So Cogito ergo sum (I think, therefor I exist), is flawed. For all we know, our universe could be completely imagined.

[Technical Ben]

This discussion reminds me of Source Code, where the machine is shut off and Stevens dies, but the simulated world still exists for some reason, implying it is a "reality" by itself.

Which is bullshit.

For all the faults of inception, at least the handwavery "it's all just a dream anyway" actually covered it. I saw the trailer to source code and just wanted to facepalm the Cinema screen.

Simulating something is fine. When you stop it, you can restart it. If it's deterministic (or re-runnable) you could even start over. As long as those states are the same, nothing is lost.
The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

[Xanthir]

Ok, I'm probably making a fool of myself here, but I can't for the life of me convince myself that shutting off a simulation of the universe would destroy that universe.

My line of reasoning is this: a simulation is characterised by a set of rules for computing the (n+1)th state given the (n)th state (there might be complications if things are continuous, but I don't *think* that breaks the reasoning), and a set of initial conditions. By the very fact that the simulation can be computed up to some state n=k (whether k is finite or not is irrelevant to our purposes), all states previous to k in the simulation must be defined, which intuitively seems to be sufficient for any inhabitants of that simulation to continue to experience their universe as continuing to exist.

The problem that I have with the don't-unplug-it-or-you'll-destroy-the-universe line of thinking is that each of these states is clearly defined *before* the simulation is run. Hell, presuming that the rules and initial conditions can be expressed in some finite language (I'm not sure that this presumption is actually necessary, but nor am I sure that it's not, so 'better safe than sorry' comes into play), the simulation and all of its nth states for n<k can be said to, in some way, exist, long before anyone even codes the damn thing. I'd even go so far as to suggest that its existance wouldn't require any 'previous' (sorry, that's the wrong word for it, so I'm gonna have to rely either upon you guys being either telepathic or upon what I've written making enough sense that you can work out what I mean from the context, each of which is about as likely as the other) universe for it to be simulated within.

Presumably I've missed the point somewhere along the line, but I can't spot where

This is a valid line of reasoning, depending on how strictly you want to define "exists".

For example, the MWI of quantum physics implies that, at minimum, every logically consistent universe with the same initial conditions as ours really exists, it's just that nearly all of them are too far away in phase-space for us to detect.

You can take this further. With just a touch of Platonism you get the idea that mathematical objects exist separately from our use and "invention" of them. Algorithms are mathematical objects, and reality can be described algorithmically, thus all universes which can be described mathematically really exist. You don't need to run a program for the result to exist, running it simply instantiates the answer for you; the answer always already existed.

This starts touching on some weird metaphysics, though. I suggest reading "Permutation City", which is an excellent scifi book exploring the subject. It takes some odd narrative turns, though, which I'll wrap in a spoiler so you can read the book first if you want, as this really will spoil some of the book:

Spoiler:

The book posits that, while all logically consistent universes exist, universes are "more real" the longer they're instantiated in a running computer. You have to run a simulation for a while, throwing enough computation at it, for it to "continue to exist" when you turn off the computer. The end of the book has an inversion, where a simulated universe with many more sentient beings than the outer universe rationalizes away the necessity for the outer universe (basically, they come up with a Grand Unified Theory which explains the physics of their world in a simpler manner than "another universe running on different physical laws built a computer which is simulating us"), which causes the outer universe to lose coherence.

[undecim]

The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

Since we're working with an infinitely fast computer, we could just have it try every possible algorithm until it found one that created a universe where every possible measurement we can obtain (e.g. every astronomical measurement we can feed into the computer) is correct to within 1%.

[TheChewanater]

The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

Since we're working with an infinitely fast computer, we could just have it try every possible algorithm until it found one that created a universe where every possible measurement we can obtain (e.g. every astronomical measurement we can feed into the computer) is correct to within 1%.

There are infinitely many universes that fit that description but do not yield reliable predictions for tomorrow's lotto numbers

[undecim]

The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

Since we're working with an infinitely fast computer, we could just have it try every possible algorithm until it found one that created a universe where every possible measurement we can obtain (e.g. every astronomical measurement we can feed into the computer) is correct to within 1%.

There are infinitely many universes that fit that description but do not yield reliable predictions for tomorrow's lotto numbers

Yes, But likely the smallest such program is the one that simulates our universe. Occam's razor, etc.

Though we are still making the assumption that the universe is deterministic. If that's not the case, then it doesn't matter how powerful the computer is. Though we still might be able to use it to get the probability of e.g. the first number of tomorrow's lotto being 11.

[TheChewanater]

The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

Since we're working with an infinitely fast computer, we could just have it try every possible algorithm until it found one that created a universe where every possible measurement we can obtain (e.g. every astronomical measurement we can feed into the computer) is correct to within 1%.

There are infinitely many universes that fit that description but do not yield reliable predictions for tomorrow's lotto numbers

Yes, But likely the smallest such program is the one that simulates our universe. Occam's razor, etc.

There are infinitely many programs that are equally small, fit that description, and are not our the same as our universe.

[undecim]

The problem with our own universe is huge though. At least measuring from our own dimension, we could never know the variables to input into the computer. Uncertainty principle, plus faster than light expansion makes some things outside of our grasp. We could simulate the begginging states perhaps and run from there, but after a few billion years, any errors in our calculations will be massively off. With things such as QM, we would not end up with the same state (planet earth etc) in a simulated universe.

Since we're working with an infinitely fast computer, we could just have it try every possible algorithm until it found one that created a universe where every possible measurement we can obtain (e.g. every astronomical measurement we can feed into the computer) is correct to within 1%.

There are infinitely many universes that fit that description but do not yield reliable predictions for tomorrow's lotto numbers

Yes, But likely the smallest such program is the one that simulates our universe. Occam's razor, etc.

There are infinitely many programs that are equally small, fit that description, and are not our the same as our universe.

If they are equally small, then there can't be an infinite number of them... There are only x^256 x-byte programs.

And even if you did only find an infinite (or otherwise uselessly large) set of universes, you could still use that set to make predictions. e.g., you might notice that in 50% of these universes, the first number of tomorrow's lotto number is 12.

[TheChewanater]

I meant to say many, not infinitely many.

Still, say you test the universes by seeing if previous lotto numbers are correct for the past 50 years. You'll look at the universes that are, and try to find which one is ours. Assume each day the lotto number is twelve digits (or six two digit numbers) and there are 365 days in a year. This means that 1 in every 1.825*10¹⁶ universes matches every single lotto number every day in the past 50 years.

There are probably around 10⁷⁹-ish particles in the universe, and for each to have a unique state they must (at minimum) be represented by just over 32 bytes. This means that the file containing the universe is 32*10⁷⁹ bytes.

This means that there are 256^3.2e78 possible universes. (256^3.2e78) / (1.825*10¹⁶) universes in which every day in the past 50 years the lotto number is the same every day. If my calculations are correct (this number is too big for Python to handle) this is about 0.55*10^{10^(39.8)-16} universes that you have to guess from.

Note that there are a lot more factors in this, and you could use more data besides lotto numbers (such as weather or not lotto exists in that universe), but nothing you can come up with would narrow it down quite as much as you think. Also, I see no reason why one of the billions of numbers would show up 50% of the time, since they'd all be essentially random.

If they are equally small, then there can't be an infinite number of them... There are only x^256 x-byte programs.

So there is only one 1 byte program? And there's also a -1 byte program?

[Moose Hole]

Yes, But likely the smallest such program is the one that simulates our universe. Occam's razor, etc.

Though we are still making the assumption that the universe is deterministic. If that's not the case, then it doesn't matter how powerful the computer is. Though we still might be able to use it to get the probability of e.g. the first number of tomorrow's lotto being 11.

Imagine a universe which is identical to the real one, except you were never born. Would the lotto numbers be the same? Yes, in many instances of such a universe. Is the information in that universe smaller than the real one? Yes, because it doesn't have you.

Occam's razor is a guideline, but it doesn't apply in this case.

[Technical Ben]

Basically, the data compression problem in reverse. The more you compress data, the further you get from the original. The less you compress it, the closer you get to the original.
So, to get a perfect match in the infinitely fast computer, would require you to already have the correct simulation for the universe. It would take us nominal* time to get a reading of our universe. If we have the reading from this universe, well, we never needed the simulated one in the first place.
Without a perfect set of data to compare to, you only have an imperfect (and near infinite set) of close results to search through. However, the closer you get to the one you want, the closer you get to actually manually inputting the details of this universe. So even taking it step by step, we end up having to look in nominal time for the state of the universe.

(Hope I've not confused myself, and typed it backwards there, any corrections would be welcome. )
*Is that the correct usage?

[ryanelm]

I hope I'm not repeating anyone but,personally I'd calculate what was going to happen tomorrow with a really simple GA, thousand monkeys style.

[undecim]

Is the information in that universe smaller than the real one? Yes, because it doesn't have you.

That is not necessarily the case.

If f(x) is identical to sin(x), and g(x) is identical to sin(x), except that g(x)=0 for -pi < x < pi, then g(x) requires more information to generate, even though it's missing a feature.

The universe isn't just a collection of states at every point in time. Every one of these states is the result of the states before it, and it began as 1 singularity. Just like when generating a fractal, you might start with a single line, and a simple rule, but end with a complicated curve.

I meant to say many, not infinitely many.

Still, say you test the universes by seeing if previous lotto numbers are correct for the past 50 years. You'll look at the universes that are, and try to find which one is ours. Assume each day the lotto number is twelve digits (or six two digit numbers) and there are 365 days in a year. This means that 1 in every 1.825*10¹⁶ universes matches every single lotto number every day in the past 50 years.

Assuming that it is possible for each of those universes to exist. I suppose QM technically makes anything possible, but the probability of many of those universes existing would be very low.

There are probably around 10⁷⁹-ish particles in the universe, and for each to have a unique state they must (at minimum) be represented by just over 32 bytes. This means that the file containing the universe is 32*10⁷⁹ bytes.

This means that there are 256^3.2e78 possible universes. (256^3.2e78) / (1.825*10¹⁶) universes in which every day in the past 50 years the lotto number is the same every day. If my calculations are correct (this number is too big for Python to handle) this is about 0.55*10^{10^(39.8)-16} universes that you have to guess from.

I see no reason why one of the billions of numbers would show up 50% of the time, since they'd all be essentially random.

Assuming they're perfectly random.

If you get a set of universes that have closely approximated our measurements of the universe for the last 50 years, we can expect that most will continue to approximate our universe, provided that it follows similar laws of physics (which would be part of our search algorithm)

When our lotto official picks a number (let's say, for the sake of argument, she takes a bingo ball out of a basket), there are only a few numbers that are even close to her hand. Some subtle differences in our universe might change one number to the number that was right next to it in the basket, but it would take quite a bit of difference for her to pick a number in the corner of the basket instead.

[joshz]

I hope I'm not repeating anyone but,personally I'd calculate what was going to happen tomorrow with a really simple GA, thousand monkeys style.

But what's your fitness function? How do you trim down the vast amount of data to something usable?

[Xanthir]

If you get a set of universes that have closely approximated our measurements of the universe for the last 50 years, we can expect that most will continue to approximate our universe, provided that it follows similar laws of physics (which would be part of our search algorithm)

No, we can't really expect that. You can show this by an argument similar to the Boltzmann Brain problem. The number of universes that have similar laws of physics and similar history is much, much smaller than the number of universes that happen to approximate the parts of our history we check but have massively more complicated laws of physics, or are simply random.

When our lotto official picks a number (let's say, for the sake of argument, she takes a bingo ball out of a basket), there are only a few numbers that are even close to her hand. Some subtle differences in our universe might change one number to the number that was right next to it in the basket, but it would take quite a bit of difference for her to pick a number in the corner of the basket instead.

Not really. A subtle difference could cause the balls to be put into the basket differently, for example, thus altering what she chooses.

[Yakk]

If you get a set of universes that have closely approximated our measurements of the universe for the last 50 years, we can expect that most will continue to approximate our universe, provided that it follows similar laws of physics (which would be part of our search algorithm)

No, we can't really expect that. You can show this by an argument similar to the Boltzmann Brain problem. The number of universes that have similar laws of physics and similar history is much, much smaller than the number of universes that happen to approximate the parts of our history we check but have massively more complicated laws of physics, or are simply random.

The Boltzmann Brain problem could be simply wrong.

Ie, our specific universe evolved from a low-entropy starting state, while there might not be any (reasonably short) paths from such a low-entropy starting state to a Boltzmann Brain universe that appears locally identical.

Ie: start with exceedingly low entropy, have laws of physics that generate increased entropy, and have limited "time" for randomness to otherwise accumulate. Find the short "time", simple start, locally similar universe -- between these restrictions, I'd be surprise if "random and chaotic universes that happen to locally look like ours" dominate. Now, "universe that locally looks like ours, but diverges reasonably fast" I expect to be exceedingly common (and ridiculously dense, so much so that within your measurement accuracy lies a continuum of reasonable universes, with no way to actually pick out this one).

And, based off the "we aren't special" principle, very few of them will be Boltzmann Brain like. Otherwise, we would actually be living in a Boltzmann Brain situation almost certainly, and we find out in 2 seconds... Did it happen? Nope?

[Xanthir]

True, Boltzmann could be wrong. But the "very quickly diverge significantly" universes are just as bad.

I'm also not sure about your reasoning about there being a short amount of time for chaotic universes to arise is valid. Given fairly simple physics, I might agree. But there's no reason to assume that an initially-low-entropy universe will have low-entropy physical laws as well.

And no, you can't use a "just wait two seconds" argument to reason that you're not in a Boltzmann Brain universe. "Two seconds" later, you could just be a spontaneously-ordered-from-the-chaotic-void brain with a built-in memory of just having had that argument and, two seconds ago in memory-time, decided to wait for two seconds and see. That's the whole problem with Boltzmann Brains - if you accept that they're possible, you (as far as we currently know) have to accept that you're almost certainly one of them, and will dissolve into the void in the next moment. Smart people are attempting to find ways to combine "we are not special" with "we are not a Boltzmann Brain".

[hideyyo]

And.....back on topic!

I would like to overclock the CPU.

[WarDaft]

There's also the possibility that QM randomness is incompressible. In such a case, you simply can't predict your own future. It doesn't matter what magnitude of infinite computing power you throw at the problem.


Also, in a universe energetic and chaotic enough to randomly have a fully operational thinking brain just pop into existence through particle vibrations, those brains will be just as likely to be obliterated before their neurons can process anything at all, including the concept of their existence - thought isn't instant.


Furthermore, we can lean on the good old anthropic principle even more... a self observing its existence is not distributed randomly across the high entropy universe, but randomly across the time that all existing entities capable of that realization live. If the majority of that time occurs in stable pockets, then again there is no paradox.


Lastly, we're forgetting the good old inflaton. Never seen, but particularly useful if it could figuratively blast away all that nasty entropy by simply super-enlarging the distance the entropy must cross to reclaim the whole of the observable universe. Indeed, this not only explains a massive low entropy universe appearing at random, but if we define a sphere outside the influence of the inflaton nova where the space is not stretched, the interior of that sphere need not even have a reduced entropy relative to it's surface area as a result of the massive low entropy universe appearing inside it! A homogeneous glow of particles spontaneously appearing much more probable than an organic supercomputer.

Post Ultimately, we can put those last two together... universal law sets without inflaton like particles would have vastly lower observer time in them, and thus we would almost certainly not be in one.

[TheChewanater]

Also, in a universe energetic and chaotic enough to randomly have a fully operational thinking brain just pop into existence through particle vibrations, those brains will be just as likely to be obliterated before their neurons can process anything at all, including the concept of their existence - thought isn't instant.

They don't need to. They just need to be created in the state of a brain that acknowledged its existence.

[WarDaft]

They don't need to. They just need to be created in the state of a brain that acknowledged its existence.

But there's no thought if the neurons don't have time to fire. If you created an exact physical duplicate of my brain, but frozen, then it's not experiencing anything. It doesn't matter what its memories say it should believe, it never actually interacts with them, so while it is frozen it is not actually believing them. If you destroy it fast enough, as few neurons will fire as would were it frozen.

Now, it is possible for a physical structure to emerge which is static but is still a computation, but the brain is not such a structure. Pictures of Rule 134 would be an example. But these are not very likely at all, because then you have to have a vastly larger volume of information snap into being.

[TheChewanater]

They don't need to. They just need to be created in the state of a brain that acknowledged its existence.

But there's no thought if the neurons don't have time to fire. If you created an exact physical duplicate of my brain, but frozen, then it's not experiencing anything. It doesn't matter what its memories say it should believe, it never actually interacts with them, so while it is frozen it is not actually believing them. If you destroy it fast enough, as few neurons will fire as would were it frozen..

How do you know your neurons are firing, and not just arranged as if they just fired?

[WarDaft]

It doesn't matter. Computation in the brain is a 4-dimensional process, without time, you just have data storage. If there's intelligence present in only the 3 spatial dimensions of the brain, it's nothing like what we know.

[TheChewanater]

...it's nothing like what we know.

Can you explain how we 'know' this without quoting Descartes?

[Moose Hole]

...it's nothing like what we know.

Can you explain how we 'know' this without quoting Descartes?

Of course! Descartes spoke French. Any useful quotes here would be in English.

[Xanthir]

They don't need to. They just need to be created in the state of a brain that acknowledged its existence.

But there's no thought if the neurons don't have time to fire. If you created an exact physical duplicate of my brain, but frozen, then it's not experiencing anything. It doesn't matter what its memories say it should believe, it never actually interacts with them, so while it is frozen it is not actually believing them. If you destroy it fast enough, as few neurons will fire as would were it frozen.

Now, it is possible for a physical structure to emerge which is static but is still a computation, but the brain is not such a structure. Pictures of Rule 134 would be an example. But these are not very likely at all, because then you have to have a vastly larger volume of information snap into being.

I don't understand why you're assuming that the act of having your neurons firing is required for thought. It's clearly required for your beliefs to change (at least, to change via the natural medium of human thought), but the act of believing is a brain-state that I can record and store (or generate ab initio from the entropic aether).

[WarDaft]

That's hardly necessarily true. You can record a single instantaneous brain state from someone who is currently believing something, but there is no reason to believe that this static snapshot is actually conscious, or even sufficient to describe the process of believing something. In any given microsecond, you have millions of neurons firing, and we are not capable of thoughts on anything near that times scale. Memories are stored as arrangements of chemical compounds. This can be serialized into a single string representing the layout of your memories. Actually ascribing meaning to any part of this requires a specific decoding process - this is because we can create a decoding process to ascribe any meaning to the string. It is only when we have locked in a specific process to manipulate the string that the string can have meaning.

And yet, this is still not enough! I can most certainly decide on some specific process that does not describe the natural functioning of my brain, yet this does not fundamentally change the nature of my consciousness. The chain of strings describing my ongoing brain state is still there, happily ignorant of whatever it is that I have told everyone to interpret it with. The only reconciliation available to us is that it is only when the string is actually manipulated by some process that that process defines the meaning of the string to the string itself.


Now, if anyone's going to go and ask me how I can be sure of that, because I could have just popped into existence with a memory full of delusions relating to the day to day act of thinking, then I must first ask you what evidence you have that there is even a problem in the first place, why aren't you delusional? I'm not the one claiming that all knowledge is potentially false and just a freak random (and unimaginably unlikely) occurrence that we believe it, after all.


And I offered an even better out, via inflation. As I said, surely a homogeneous field of inflatons is far more likely than whatever formation is really required to generate an intelligent conscious being with terabytes of very low entropy information. Indeed, to generate terabytes of random data and get a consistent coherent meaningful result, is an act so unimaginably unlikely that just about any other possible explanation is preferred.

[TheChewanater]

Everyone seems to assume that neurons exist and are firing. It's possible that everything we know about the brain is wrong, and the randomly-arranged particles forming are brains are arranged to form memories of us reading Wikipedia articles including that information. Perhaps there is one "conscious", and it exists in Plato's cave, imagining it is a human with a brain and not a random arrangement of particles.

Of course, we'd just be throwing Occam's razor out the window then.

[WarDaft]

Everyone seems to assume that neurons exist and are firing. It's possible that everything we know about the brain is wrong, and the randomly-arranged particles forming are brains are arranged to form memories of us reading Wikipedia articles including that information. Perhaps there is one "conscious", and it exists in Plato's cave, imagining it is a human with a brain and not a random arrangement of particles.

Of course, we'd just be throwing Occam's razor out the window then.

Ahem.

Now, if anyone's going to go and ask me how I can be sure of that, because I could have just popped into existence with a memory full of delusions relating to the day to day act of thinking, then I must first ask you what evidence you have that there is even a problem in the first place, why aren't you delusional? I'm not the one claiming that all knowledge is potentially false and just a freak random (and unimaginably unlikely) occurrence that we believe it, after all.

[TheChewanater]

Everyone seems to assume that neurons exist and are firing. It's possible that everything we know about the brain is wrong, and the randomly-arranged particles forming are brains are arranged to form memories of us reading Wikipedia articles including that information. Perhaps there is one "conscious", and it exists in Plato's cave, imagining it is a human with a brain and not a random arrangement of particles.

Of course, we'd just be throwing Occam's razor out the window then.

Ahem.

Now, if anyone's going to go and ask me how I can be sure of that, because I could have just popped into existence with a memory full of delusions relating to the day to day act of thinking, then I must first ask you what evidence you have that there is even a problem in the first place, why aren't you delusional? I'm not the one claiming that all knowledge is potentially false and just a freak random (and unimaginably unlikely) occurrence that we believe it, after all.

Yeah... I probably should have read that before posting.

[Xanthir]

That's hardly necessarily true. You can record a single instantaneous brain state from someone who is currently believing something, but there is no reason to believe that this static snapshot is actually conscious, or even sufficient to describe the process of believing something. In any given microsecond, you have millions of neurons firing, and we are not capable of thoughts on anything near that times scale. Memories are stored as arrangements of chemical compounds. This can be serialized into a single string representing the layout of your memories. Actually ascribing meaning to any part of this requires a specific decoding process - this is because we can create a decoding process to ascribe any meaning to the string. It is only when we have locked in a specific process to manipulate the string that the string can have meaning.

In other words, you're defining the word "consciousness" as a process, not a state, and then claiming that a brain that exists for only a moment can't run a process.

Perhaps you should instead apply your cleverness to imagining what notion of consciousness *can* be applied to a state. Namely, at any given moment, your brain is in a state of computation, with memories representing past computations. Similarly, a hard drive is filled with data, even though the hard drive itself does no computation.

And yet, this is still not enough! I can most certainly decide on some specific process that does not describe the natural functioning of my brain, yet this does not fundamentally change the nature of my consciousness. The chain of strings describing my ongoing brain state is still there, happily ignorant of whatever it is that I have told everyone to interpret it with. The only reconciliation available to us is that it is only when the string is actually manipulated by some process that that process defines the meaning of the string to the string itself.

Yes, codes have no meaning outside of their interpretation. That's irrelevant to the point, because we are specifically imagining a human brain coming from the ether, which we know how to interpret.

Now, if anyone's going to go and ask me how I can be sure of that, because I could have just popped into existence with a memory full of delusions relating to the day to day act of thinking, then I must first ask you what evidence you have that there is even a problem in the first place, why aren't you delusional? I'm not the one claiming that all knowledge is potentially false and just a freak random (and unimaginably unlikely) occurrence that we believe it, after all.

The whole point of the Boltzmann Brain argument is that, given a set of apparently-reasonable cosmological assumptions, the "freak random occurrence" is actually *incredibly* more likely than a consistent 14-billion-year-old universe (infinitely so, actually).

And I offered an even better out, via inflation. As I said, surely a homogeneous field of inflatons is far more likely than whatever formation is really required to generate an intelligent conscious being with terabytes of very low entropy information. Indeed, to generate terabytes of random data and get a consistent coherent meaningful result, is an act so unimaginably unlikely that just about any other possible explanation is preferred.

Yes, it is quite unlikely. "Unlikely" means very little in the face of infinite time and/or infinite numbers of universes.

[Prime Intellect]

This.

I think it's unlikely that we could know the original conditions of the universe, even with the right theories. Still, pretty mind bugging story.

I love his stories (I think his site is qntm.org?) but I'd definitely try to make an AI first. Imagine what an AI could do if it could learn everything in an instant. And what it would do in the following instants with such knowledge.

fake edit: wow apparently I can't even quote posts that use [url] derp. See his original post for the link.

[lebull]

Here's my question...
What would happen if you tried to make said computer sleep(n)?

[WarDaft]

I had a much longer post ready. Then the power went out in a storm. Then I sort-of forgot I was replying.

In other words, you're defining the word "consciousness" as a process, not a state, and then claiming that a brain that exists for only a moment can't run a process.

Perhaps you should instead apply your cleverness to imagining what notion of consciousness *can* be applied to a state. Namely, at any given moment, your brain is in a state of computation, with memories representing past computations. Similarly, a hard drive is filled with data, even though the hard drive itself does no computation.

Actually there's an even easier out - consider a computation of your entire life as expressed in rule 134. Such a thing could definitely form out of random chaos, and even last long enough to act so as to interpret itself as a computation of life. But the closer it comes to approximating what we consider life, the less probable it becomes at an exponential rate.

We just need to know that it is improbable enough that a homogeneous field of inflatons forming is more likely, thus resulting in a little pocket of the high entropy universe mimicking the big bang.



Actually, what we should really be asking ourselves is "what set of properties is most conductive to the existence of intelligence in areas of equal maximum (and finite) information content in an infinite universe?" Intelligence will find itself to be in the infinitely large universe where intelligent life occurs once in every 10^{10^{10^10}} bits far less often than it finds itself in the infinitely large universe where it occurs once in every 10^{25} bits. Amongst universes whose rules have equal kolmorgorov complexity, a power tower of tens that high should outweigh the greater number of less life conductive universes.

[Yakk]

My favorite illustration of Kolmogorov complexity and how it is not as intuitive as you might think....

Take a look at procedural content generation and compression. The demo scene is an example of this -- building "small" executables that generate ridiculously complex scenes. The trick is twofold -- first, generated content, and second, building the content in a way that the generated content is compact. Ie, instead of specifying where something is, you could have it positioned randomly, then search the random seeds that put it close enough to where you want.

Similarly, what we see as a really complex universe might actually be really simple -- it might be the result of relatively simple rules, a relatively simple initial state, and a whole bunch of computation on the initial state. In a sense, the "floating brain" ends up being more complex, because to describe the floating brain you need to both describe the seed of the universe that would create it and the fact that the rest of the universe is empty, together with rules that allow both the standard unfolding of the universe and allow for some way to describe "oh, and a bunch of what it described didn't actually happen" -- or explicitly describe its entire current state (which could be very complex compared to the "seed" state that would generate it).

Which is basically what WarDraft is talking about.

[WarDaft]

They're not presuming the brain's universe to be empty. They're actually presuming it to be extremely high entropy which would naturally obliterate the thinking engine a fraction of a second after it forms. This actually doesn't get us much, because to randomly generate just the brain, you just need to specify the brains location and shape as being all that's there, a relatively small statement. It also only produces one single brain, rather than infinitely many brains, which is what the infinite random chaotic universe does. My argument is that amongst equally "complicated" rule sets, say one generating an infinite super high entropy random universe and one generating an ours (which may very well also be infinite) then even a single occurrence of a thinking brain across the entirety of our observable universe constitutes a vastly higher frequency of intelligent life than the equal volume frequencies of intelligence summed up amongst all the super high entropy universes with equal rule complexity.


This does raise a question however... if something in a universe is perfectly random (not apparently perfectly random - at least as random as being not predicted to be part of a sequence by any Turing Machine, halting or otherwise), should we consider this value to be part of the 'seed' of that universe?

[brainfuck]

I wouldn't start by simulating the universe in all its complexity.
Creating a paradox can be as simple as typing:

Code: Select all

while(true){ }

Processing an infinite loop with infinite processing power takes an undefined amount of time.
There is no answer to how long it would take, however my feeling tells me I wouldn't live long enough to see the loop finish and the same applies to the computer.

Either this or something like

Code: Select all

SELECT TOP 1 FROM questions WHERE answer = 42 ORDER BY relevance DESC

[WarDaft]

while(true){ } has no termination condition. No computer, no matter what the extent of it's processing capabilities, (finite, countable, uncountable, inaccessible, anything) can run the algorithm correctly and halt.