A requested thread for Revolutionary:
This link give an introduction to the topic.

Katisara brought up this idea a while back, I think, but I couldn't remember which thread it was in.  My thoughts on it are that it makes some rather large assumptions about the computational power of 'post human' civilization, which basically amount to "we'd be able to do anything at all!" which I think requires a bit more of a leap of faith than I'm willing to give.

Simulating the entire history of humanity (which is what the main article in the link suggests), with computational consciousness for all members that ever lived is mind-bogglingly large as a computational task.  The theory assumes it will be so trivial to do that a post-human civilization would do it 'many' times.  I'm not convinced that's a reasonable assumption.

Now, simulating ONE human mind, for, say, a lifetime, might someday be feasible (though still prohibitively costly) I could imagine, but that breaks the argument made in the paper.  Even if they did this a few hundred times, there'd still be far more 'real' people experiencing life than minds simulated this way, so we wouldn't be more likely to be simulated than real.  But even one simulated might seems somewhat unlikely to me.  It seems more likely that we'd create an simulated mind, but not simulated world for it to interact with.  We've got a real world already that we don't have to simulate, why not just use it?  So an artificial intelligence seems like a more likely scenario than tons of simulated entire-history-of-humanity to me.

Another aspect of it that tends to set off alarms for me is that it's not based on anything in reality, and thus isn't falsifiable.  There's nothing we can test to see if it's likely to be true or not, as it depends so heavily on what we imagine 'post-human' technology to be like.

I agree with most of what you've said.  I just honestly think we're far more likely to asteroid/superflu/globalwarm/nuclearholocaust ourselves out of existence before we ever reach that sort of level of advancement in society.

In reply to Doulos (msg # 3):

The first point I would point out is we cannot start with the assumption the universe is anything like what we think it is in this case. Afterall, if we are no more complex to that which allowed us to think we are in this simulation is no more intricate than a person in The Sims is to us... Well, everything changes.

Likewise, a universe of mostly deadly radiation and really only of star dust... Need not be real any more than magic is when I run theHogwarts edition of the sims.

When there is a little downtime, I will lay out a number of purely phenominological reasons I actually would be not at all surprised tofind this true. I even will say i would be a little sad if it is not.

That's sort of the problem in my eyes.  We're essentially saying we can't assuming anything about reality, which means anything could be true.  You're sort of taking the "if it were really X, we'd never even realize it!" approach, which means it's not falsifiable, not testable, and sort of just making stuff up.  Sure, we can't disprove it, but that means it's pure speculation.  Reality could all be a dream of a sleeping dragon, too, but that doesn't mean it is, nor that we should assume that it is.

By our standards, it does. But remember, by the standards of someone living in 1,000 AD, we can basically do 'anything at all'. We can go across the world within a day. We seem to create food and prevent it from spoiling. We make water appear, apparently from nowhere. We can make cows pregnant without bulls. We reverse disease and even aging.



Having a computer operate all of the processes of a human mind for 80 years costs in the range of $100,000 (depending on circumstances). That computer is called 'a human mind'.

Right now, our computers are extremely simple by the standards of a biological brain. There's a lot of reasons for that. But our technology is catching up. In fact, we are currently learning how to operate mathematical transactions at the quantum level. It seems almost self-evident that technology will, at minimum, reach the computational capabilities of a human brain, and reduce them to same energy and size requirements (which means it takes less electricity than a light bulb, and less space than a desktop -- i.e., your normal laptop). After all, we know that processing capability is physically possible; it's happening right now! We accept that the minimum cost for our future society to run a single mind is about $1,000 in today's money (i.e., the cost of a high-performance laptop).

So that establishes our lower bound. But establishing the upper bound is much more difficult. We can all agree that transistors cannot be shorter than the planck length and still hold useful data, but most people will reasonably expect that upper bound to still be pie in the sky. But the problem is, we really have no idea what the upper bound is. So far, Moore's Law has continued to proven itself correct. Every time we worry about transistors getting too small, we exploit a new technology to keep shrinking them.

We also have to remember that the cost for projects like this will not be significantly higher than buying a billion laptops, since if it's cheaper to do it that way, we will. But it may be lower. So if we wanted to simulate a single country, say the US, with 200 Million people, the absolute maximum is 200 billion. But if Moore's Law holds, in ten years that'll be $20 billion. Even if we don't stick to that pace, but futher enhancements are possible (which they almost certainly are), we keep chopping that cost down.

And do recollect, we don't need to simulate the entire world, any more than Blizzard needs to simulate the bits of WOW no players are accessing. We only need to create feedback appropropriate to where the viewer is, and the entire processing is already being done by the brain! So on the one hand, the additional processing power necessary to 'run' the world is negligible. But a lot of processing our brains do (keeping our heart running, keeping our digestive track working, etc.) are no longer necessary, which gives us a SURPLUS of computing power.



Is Tycho the scientist actually asking this? Tycho the guy who plays RPGs? Tycho the philosopher? The explorer? You can't think of any reason not to simulate another world, when we have this real one just sitting around?



It is indeed currently philosophy (similar to the current theory that black holes are the seeds of new universes). The math fits, but our technology isn't far enough along yet to test it.

This is a great conversation and in some ways a relief from some of the religious ones. I think it even has within it some of the issues that pop up about faith and science.

Now, let me change the question.

Do you think there's something "Special" (in almost all but the trivial sense I mean that word) about the "stuff" of brains that "contains" and in fact exclusively implies consciousness?

Or do you think that what we understand as a computer with sufficient computational power will evolve into what we reasonably would consider to be "conscious?"

It's not as simple as you dismissed either.

This is not just some "Cartesian Deamon" we're playing with.

The point is a rather simple one (which I will oversimplify)

* If a world can be simulated, it will be simulated.
* If a world is simulated, several will be.
* If several worlds are simulated, if you seem to live in a world--most likely it's a simulate one.

Which can be reduced to...

If you seem to live in a world AND a world can be simulated, it's mostly likely that you live in a simulated world.

Now I get why you might question that a world CAN be simulated.

And, UNDERSTAND, I (as more or less a "believer") am making NO epistemological conclusions...

I am not saying... THEREFORE:

* Nothing we do matters...
* Everything is pointless...
* Eat, drink, be merry...

Furthermore, that we are in a simulation and as of yet don't know it has no moral or ethical implications what so ever.

If we come to know it there are a number of implications and none of them change, only expand.  Do we have an obligation to our "outside system setters" ...etc.

I was thinking over this today, but katisara beat me to the big punch.

We have every reason to expect that technology will eventually be able to produce a computer smaller and more powerful than the human brain. This brain is then hooked up to an interface that can connect to the historical database and can also read everything stored in the mind. Then it can create whatever the mind thinks of as plausible and fool the mind into thinking its experiencing reality.

One you have a few of these, you can start networking them together. If you have galactic scale resources, you can afford to make a computer a couple of miles across (all the human beings currently alive, if stuffed into coffins and stacked up, would be about 1.5 cu miles), which could house all of these simulated brains.

The big question, though, is are they conscious? We still use the term loosely, without any sort of testable meaning. We can already make machines that parrot semi-human responses, and machines that adapt and learn from experience, but we don't claim they are conscious. As they get more sophisticated, they will have more and more the appearance of consciousness, but how can we tell if they are conscious?

At the root of it, we each know that we are thinking. We experience other people that seem like ourselves, and assume we are not special, so they must be thinking, too. The concept of what in the world is conscious grows outwards from there and gets less certain the farther away from human we go. Do we have any basis for thinking that a machine made to simulate a person is actually thinking?

And yet, if someone in 1,000 AD had said "In 1000 years, humans will be able to snap their fingers and teleport to different planets, will never have to eat again, will live forever, and simulate all of human history billions of times on their ipads in just a few seconds!" they would have been wrong.  The fact that we can do LOTS more than we could 1000 years ago doesn't mean we can do EVERYTHING.  Likewise, though we can be confident humans will be able to do a lot more than we can now in 1000 years (provided we last that long), we shouldn't conclude that they'll be able to do EVERYTHING.


And, interestingly, unlike most of the bits of Moore's law, that number tends to go up with time, rather than down.  Not that that impacts the argument at all, but I think it's interesting to note that the cost of running one of these experiments is already quite expensive, time consuming, and increasingly so.


That doesn't seem self-evident to me.  Possible, perhaps, but nowhere near certainty.  We're no where near the complexity of the human brain right now.  People like to talk about the "number of computations per second" or "transistors per cubic meter" and compare what we can do with that of the human brain because by some measures we're already ahead of the human brain.  But when we look at any computer that's actually designed to do what the human brain does, we realize we're not even in the parking lot yet, let alone the same ball park.


True, we may someday be able to match the human brain, by copying its functionality, but I don't think it then follows that we'll make more copies of it than have ever been biological versions (a key assumption to reach the "therefore each of us is likely to be a simulation" conclusion).  The "if we could do it once, we'll do it a gazillion times" idea doesn't follow in all cases.  We went to the moon once, but we didn't keep going over and over again.  In fact, we stopped going after just a handful of times.  Far from being as common as going to the grocery store, it's returned to being essentially impossible for someone to get there 40 years later.


And even that would prohibitively expensive for simulating the whole of human history.  Just to simulate the number of people who are alive right now would take trillions of dollars.  And that's just a tiny fraction of all the people who have ever been alive.

As I said before, the idea of simulating some people's experience doesn't seem too impossible.  The idea of simulating more people's experiences than have ever been alive doesn't seem nearly as likely.


But the "it's always kept doing this in the past, so surely it will do it forever" argument is proven wrong every time someone dies.  Moore's law has held for what seems to us like a long time, but it's really no time at all in the grand scheme of things.  It's a very local observation, not a global (in the abstract sense, not the literal sense) trend.  It's a description of our current (ie, right now plus or minus 50 years or so) rate of technological advancement, not a rule that universe is bound by.  Will it go on for the foreseeable future, sure, probably.  Will that get us to simulating all of humanity's experience many, many times, not nearly as obvious or certain.  There's a big difference between saying "this car's run great the last 5 years, so I don't see any reason to expect it to stop running next year" and "this car has run great the last 5 years, so I think it's reasonable to conclude that it will never stop running ever."


Not true.  There's been WAY more than a billion people who've lived.  And we need to do it more than just once for each person, we have to do it many times for each for the conclusion to hold.  Even spending $1000 per person currently alive is more than humanity has ever been willing to spend on any project.  If we were willing to spend that much on something, I have a long list of higher-value projects than simulating reality for a billion minds.


Wait, wait, wait!  No.  You're assuming that we could right now simulate one human's existence with a $1000 laptop, which is not the case at all.  Once you say "some day we'll be able to do that" we've already made a huge leap.  If you tack on "and at that point, Moore's law will still be running just like it does today," you've made an even larger leap.  I'm not willing to accept those as "obvious" or "granted" assumptions.  The logical conclusion of this line of reasoning is that eventually we'll have infinite computational power, at zero cost, at zero energy consumption, and taking up zero space.  I don't think that's reasonable.


But for the conclusion to hold, we do need to simulate the experienes of many times the number of humans that have ever existed.  Like I've said, simulating one person's experience seems fairly possible.  Repeating that many times for every person who's ever existed seems a much, much bigger ask.




Simulating an entire world for the purpose of looking at simulated experiences of minds seems unnecessary.  For other purposes, sure, but that won't lead to the conclusion that we're all likely to just be simulations.  If we made a simulated world so we could go back and see what it was like in the ice-age, we wouldn't need to recreated the consciousness of every creature in the simulation.  Much of the point of simulating thing is to avoid having to do that.  That's sort of the difference I'm getting at.


I'm not even sure it's that far along.  It's not that the math fits, it's that "if we assume anything is possible, then this is one of the infinite number of things that follow."  There's not really math to fit, just a handful of large assumptions.  We could just as easily argue "evil people exist now, so there will probably be evil people in the future.  In the future we'll have near-infinite computing power, and evil people will probably have access to it.  If evil people could do anything evil by simulating reality, they would.  If they'd do it once, they'd do it many, many times.  Therefore, it follows that we're part of some evil person's simulation, and should be doing our best to break the simulation."  Arguing from the "in the future we'll be able to do anything at all" is a bit like starting an argument from a contradiction.  You can derive absolutely anything you want from it.




Depends on what you mean by 'special,' I guess. From what I think you mean, I would say no.  I think consciousness could be obtained by material different from our brains.  I do think our brains are special in the sense that they're the product of billions of years of evolution, and are very, very good at what they do.  It would take more than just a really fast processor, or a very big hard drive to compete with them.  I don't think any piece of technology, with the possible exception of the internet in its whole, comes anywhere near the complexity required to have consciousness.


Not a 'computer', no.  I think we'll probably develop conscious technology in the future, but not just making computers better.  It's not that you get past a certain number of flops and suddenly a computer wakes up.  You don't just plug a few thousand laptops together and get a conscious entity.  Consciousness is not just computational power.  Admittedly, we don't have a great grasp of what consciousness is, or how it comes about.  My own personal view is that it cam about from the fact that our brains needed to be able to predict how other organisms would behave under given situations.  It needed to 'model' other organisms' behaviors, which meant sort of 'simulating' those other organisms within the brain.  Once we had such a capability, we could also turn it on ourselves, and model our own behavior the same way.  This allowed us to 'tell a story' about what our selves were doing, that was much more simplified than a detailed list of all the biological goings-on.  Consciousness, under this idea, is sort of like a running narrative that explains and tries to make sense out of what we're doing.  From there we could get a sort of feedback loop, where the consciousness not only monitors and tries to explain what the self is doing, but also can influence what it does.  Obviously this is only a very rough sketch of the idea, and leaves out all the important details (because I don't know them!), and could very well be completely wrong.  The important bit, though, is that a computing-device doesn't need to be conscious to do stuff.  There needs to be a reason for it to become conscious, and just adding more processing power isn't enough to make that happen.



But it's not just a simulated world, it's the experiences of all the entities within that world being simulated as well.  Simulating a world, but not having any (or at least not all) of the inhabitants being conscious is very different from simulating a world with billions of conscious entities.


Not just 'several' by 'many, many' more than 'real' worlds exist, or have ever existed throughout time.


This part I'm okay with, if we accept the first two premises.


Does it have any implications at all?

It has several implications, including ones that seem to fit a fair number of phenomena better than other options do.  Such as "The big bang"

BTW, there's no reason a simulated universe cannot be fully played out in all time and space. Much like the potentially holographic universe we have.

I hope you realize how ridiculous the premise you've put forward is. I was going to comment that someone in 1,000 AD would not have the concept of the planets necessary to put forward the idea of teleporting to them, but of course, once you bring up ipads, it's clear you're not asking a question of a citizen from 1,000 AD, so I think that's a bit unfair.

You're also running on the assumption that in a simulation, anything is possible, and of course, that's also untrue. The simulation will have to operate under constraints. Things like maximum resolution (plank length), size (width of the universe, perhaps smaller), data loss/corruption, processing speed, etc. Remember that WoW is a life simulation, and while they could implement things like teleportation and instant food without problems, there are a lot of things which are obviously constrained.

So indeed, we will not be able to do everything in a simulation.



Computer prices have most certainly not gone up! The cost of computers within a particular market range have kept about constant, or dropped. The exception is the creation of new market ranges (server farms, for instance), which simply didn't exist before.



That doesn't seem self-evident to me.  Possible, perhaps, but nowhere near certainty.  We're no where near the complexity of the human brain right now.  People like to talk about the "number of computations per second" or "transistors per cubic meter" and compare what we can do with that of the human brain because by some measures we're already ahead of the human brain.  But when we look at any computer that's actually designed to do what the human brain does, we realize we're not even in the parking lot yet, let alone the same ball park.



Can you give a reason why? After all, the primary problem with creating human minds is their long production time and inefficient program-loading methodology. If you can make them in a factory, and load the base software on each as part of the production, that is enough to drop the price precipitously.

Ultimately remember that the human mind was developed in response to an evolutionary need. Its role is to enable this animal to find food, evade predators, and make lots of babies. It isn't built for quick production, low cost, or standardized deployment.




Fair enough, but your counter-argument seems to be 'we will one day learn everything there is to know and no further computational advancement will be possible', which seems far more unjustified a position.




But that's an O&M cost. I was capturing a snapshot in time of the costs right now. If you're already running ten billion, swapping a few thousand in or out every day isn't such a big deal. This is also assuming that one computer = one person, which isn't a fair assumption. Do we replace WoW servers when a character dies?




It's not a huge leap. I've justified each step. A human brain is laptop sized, and has lower than laptop operational requirements. Currently our technology can't let us make them, but there's absolutely no reason to assume there's something magical about a mammal's womb that lets it make super-advanced computers.

I have assumed Moore's Law holds, but that's just for guessing the future cost changes. We don't need to assume Moore's Law holds, just that technological advancement holds. If it takes 50 years to increase computing power 10x, that's still 10x.



That's also not the logical conclusion. How many times do you have to multiply 10 by 10 to get infinity?



1) You really don't. If I set up the simulation so you remember your grandparents, and everyone else aroudn you does too, that's sufficient.
2) I don't need to run them simultaneously. I'm not going to have 100 billion simultaneous consciousnesses.
3) There are easy algorithm tricks to save on processing power, by reusing data for similar experiences, by using sleep cycles to support other processing, by stripping down data storage (i.e., permitting imperfect memory), etc. So 100 billion isn't 100 billion.



Economic or sociological observations and experimmentation, historical research, alternate histories, models of rising intelligence, an efficient way for individuals to live luxuriously, the drive for creation or reproduction, as a precursor to creating a utopia, as a 'time capsule' to send a massive colony to a distant location. That's just in ten seconds, with my limited, one person, 20th century mindset.




I think this is really the fascinating question, although I'm worried going into it will totally muddle the conversation of 'is this simulation possible?' just because of the size and complexity of the two different issues.

In reply to Revolutionary (msg # 10):

That kind of "there's no reason why it can't be like..." thinking is the part that bothers.  You could be ANYTHING after the elipses, and it'd be pretty much equally true when we're speculating so wildly.  There's no reason that the whole universe isn't just a tiny microbe on a pile of magical unicorn dung, that is, itself just a simulation of magical unicorn dung created by sentient rain drops of some tropical island in the ant-world of bladittyblah.  But we also have not way of testing that theory, so its a non-issue.  We wouldn't notice any difference if its true or if its false.

As for fitting things like the "big bang" better than other theories, I'm not sure I follow.  Why would that be part of a simulation involving consciousness?  Why would they pick on particular start-of-the-universe model over any other (with the notable exception of "it's how their 'real' universe started")?

It sort of sounds like you're maybe switching ideas here slightly.  An "all of time and space" simulation, and one where the big bang 'fits better' maybe sounds like you're talking about them actually creating an actual, physical universe somehow.  That actually seems more feasible to me than some software simulation.  But to me, the former isn't a "simulation," its an actual universe that just happens to have been intentionally created by some intelligence 'outside' our universe.  I'd be much happier to accept the argument: "IF its possible for beings to intentionally create new universes, and its not prohibitively costly for them to do so, then there is a good chance we exist in such a universe."  But the "IF" clause is very big, and again, pure speculation at this point.

Sort of related to that, it seems more believable to me that we'll someday create functional 'brains' that are conscious in the real world than it is that we'll create simulated brains in a purely simulated world.  The trouble of creating the artificial external stimuli for "many, many" simulated brains seems not worth it, when we could just as easily give them stimuli from the real world for free. Simulating the stimuli for a few 'brains' seems possible, but I just don't see any reason to do it so many times that there's more artificial realities than real ones.

I think that's percisely the problem.

We're applying constraints we presently have or think we have...to governor the imagination of what *Might* (and clearly only might, this is not about proof of the claim at least for me) be or in fact what would *NEED* to be in order for this to be true.

Likewise, I think the other poster hit one of the more important questions: Are the implications or how does the "truth" of the claim inform us.

Given that a simulation would "set" the variables of constraint, it seems so odd that someone is bringing up "plank length".  Now I get it may have been a temptation since someone brought up an ipad of the future.

But do any of us think even ...without getting into out there ideas like simulation... that "PayCheck" the movie had it right, we're only a breakthrough away from having no "screens" to speak of.

When you dream there seems to be very little limitations of which you speak. And the "out there" which would be running us as a simulation wouldn't need to be ANYTHING like us at all.

Anymore than when I run "sim ants" I'm anything like an ant.