Hacker News Reader

>But it is impossible to completely simulate a system from inside that system--that's basic thermodynamics.

When you're talking about universe-scale simulations, I wouldn't be so confident in asserting what is or isn't possible.

Using our present level of scientific knowledge as if it somehow informs us as to the feasibility of universe simulation would be like paleolithic humans discussing the finer points of modern microprocessor architecture.

It's likely that time and space are both infinite, which would equate to infinite computational power. It's also possible that other universes may exist with physical laws completely alien to our own. I fail to see why our reality being a computer simulation is somehow far-fetched.

It seems far more likely that the universe is a computer game rather than a simulation. A simulation implies that it tries to simulate everything accurately - if it's a computer game, we can expect that the creators cheated in multiple ways to keep computations way down.

Obvious ways to cheat:

- Only the things you are directly experiencing and seeing is simulated to a high degree.

- If you ever came close to discovering the true state of the universe, the game might simply erase those thoughts or twists things to make them fit again.

- People who really study physics and have access to detecting whether the universe is a simulation are all NPCs :)

"It's likely that time and space are both infinite, which would equate to infinite computational power. " - someone should tell Mojang that Minecraft brakes the universe lol.

Its theoretically infinitely big, as it starts with a seed number and biomes are generated when you interact with them, in a similar way to how this universe acts on quantum levels, kind of.

The thing I always found perplexing about the idea of an ancestral simulation is - how do you initialize that? E.g. suppose we are in the year 3,000 and we have super advanced computers and I'd like to simulate the year 2017. How do I lay out all the humans/their brains/animals/trees/rocks/etc, without access to the snapshot of the state? And some attempts at patching this, e.g. by simulating physics backwards from a current state because the laws of physics are reversible is not likely due to 1) entropy/chaos when going backward, 2) the difficulty of taking a simultaneous high-fidelity particle-level-accurate snapshot in the year 3,000, and 3) missing boundary conditions.

1) you start from Big Bang and go forward.

2) you can at any state check the difference between the simulated state and historical record

3) if your simulation converges, that is the closer to present day, the less the uncertainty, then you are on a right track.

4) if you simulation diverges you go to step 1 and tweak the initial conditions

5) you don't run a single simulation - you run many in parallel and select the best match at each step.

It doesn't have to be historically accurate. Just believable by the simulated.

If a simulation contains minds is it not equivalent Dennet's "Brain in a vat" [1] model of consciousness ?

In "Consciousness Explained" Dennet postulates that a conscious mind perceiving a thing-in-the-world need not have perfect rich input data but merely a neuron that responds yes to the question: Am I seeing that thing ?

[1] Consciousness Explained, Daniel Dennet, page 8, [ Prelude, Hard Hallucination Problem ]:

"...intervene somewhere along the chain (anywhere on the optic nerve, ...between the eyeball and consciousness), tugging just right on the nerves would produce exactly the chain of events that would be caused by a normal veridical perception of something, ...producing, at the receiving end in the mind, exactly the effect of such a conscious perception."

If you tweak Bostrom's argument, it can sort of make sense. For example, if you believe a simulated human mind is "just as good" as a real mind, then it might make sense to colonize uninhabitable planets by building simulations there. It might even be easiest to colonize all planets this way. And if you suppose a future with lots of successful planet colonization, then there could easily be a lot more simulated minds in the history of the universe than real ones.

However, I think the main problem with Bostrom's argument is an epistemological one: It supposes an entire outer world, about which we have no evidence and know nothing. It's like opening Sagan's garage[1], but instead of asserting an invisible dragon, you're asserting an entire universe. It is the least parsimonious hypothesis you could make, because it doubles the complexity of the universe!

[1] http://rationalwiki.org/wiki/The_Dragon_in_My_Garage

An answer to that might be: he is supposing only the existence of an outermost world, and we know that such a world must exist because either it is identical to our own universe or our universe exists as some level of inner world within it.

Why does it need to be a complete simulation? At the extreme, it could be a simulation in which you alone are the only conscious entity. Nothing outside your immediate proximity is simulated.

I just would like to know how this game is scored and how many lifes I have left...

and what happens when the game is over :)

That's also means they may have to alter our minds sometimes to avoid expensive computations:

For example a bit contrived example: if we try to find the next Mersenne prime (or something similar, computationally hard but easy to check thing (e.g. anything NP complete)), and we use millions of computers to do that, and the simulators do not have it precomputed already - then the simulators either have to compute it too to provide the proper sensory input (the mersenne prime found) or alter our mind not to do this...

And it only needs to be simulated convincingly enough and in enough detail to fool your human mind.

> And it only needs to be simulated convincingly enough and in enough detail to fool your human mind.

If we are inside a simulation, there are too many variables that are beyond our control. What are 'we'? are 'we' 'humans'? Or are 'we' 'programmed' to 'think' so? what is a 'thought'? What is 'mind'? What does it 'mean' by 'fool' 'your' 'human' 'mind'?

And your mind can be programmed to ignore and forget anything that might indicate it is a simulation.

The way I usually see it is that the descendants will try to replicate humanity in a smaller universe, who will then recursively and eventually instantiate their own replication

Not that we're currently being replicated by our actual descendants, but by our higher level's descendents.

Eventually ofc hitting the base case where the simulation cannot support further simulation (which might be us, if we never manage to successfully simulate our own universe, with whatever hacks and approximation we can find).

And ofc, if we are a simulation of a higher level's history, and the simulation is correct (to all necessary degrees), then we must be heading towards the sane point where we would render a simulation (unless the approximation fails to support it).

So if level 1 managed to simulate, then all further levels (until base) should also simulate.

And if level 1 can simulate, then it probably will.

> And ofc, if we are a simulation of a higher level's history, and the simulation is correct (to all necessary degrees), then we must be heading towards the sane point where we would render a simulation (unless the approximation fails to support it).

Even if a universe-scale simulation is physically possible, why should we expect it to faithfully repeat the same sequence of events that happened in the "outer" universe?

Leaving aside the question of why a civilization would want to try to simulate its own history, it seems very unlikely that it would be possible to reconstruct the initial conditions to get such a simulation started. The current physical state of the universe is the result of billions of years of thermodynamically irreversible processes. Furthermore, there are vast regions of the universe that interacted in the past, but are now permanently beyond each others' cosmological horizons due to the expansion of the universe. A faithful historical simulation would require precisely reconstructing staggering amounts of information, and even with the physicist's concept of "infinitely advanced technology", it's hard to imagine how such a thing could be done.

I can answer the "why" question. Provided, our descendants will be able to empathize with their ancestors (rather big "if"), they can use the simulation to reverse the death and resurrect every person who ever lived.

As for verifying the accuracy of the simulation, you only need information about the initial state of the universe (think CMB radiation) and partial information about past states of the universe to reproduce a selected region of universe with certain level of precision.

You're using words like "only", "partial" and "a selected region" but you're still talking about staggering amounts of information, on any humanly achievable scale.

A single human body contains on the order of 10^27 atoms. If you just wanted to record the state of someone's mind, with sufficient fidelity to be worthy of the term "resurrection", you could probably get away with a coarser granularity. But you're talking about accurately simulating a chaotic system, which means you need extreme precision to prevent the simulation from diverging.

More to the point, it's unlikely that the information you're seeking still exists, even in principle. Think about Conway's Game of Life; the rules are injective, meaning many different initial states can lead to the same final state. An being who lived in the Game of Life would never be able to precisely reconstruct its history with any level of technology, because there would be an astronomical number of equally possible predecessor states. (As far as we know, our universe is continuous, not discrete, but the laws of thermodynamics have a similar effect.)

None of the above needs to happen overnight. Who knows what humanity will be capable of in 1000 years?

Furthermore, in a strict sense my question would be 'Is it possible to extract additional information with any degree of certainty about the past state of some region of the universe, using simulation algorithms and partial information about past states of this region?'

This assumption doesn't need to go to the atomic level on the first try. If you can replicate distribution of mass on the scale of galaxy clusters as compared to the present day using the above approach, it would certainly prompt further research in this area.

Regarding your last point, it is my understanding that in fact the laws of physics are entirely time-reversible in a mathematical sense, and that what thermodynamics does on top of that is say "in reality, time runs only in this direction".

> then we must be heading towards the sane point where we would render a simulation

Unless the purpose of the simulation is to observe a civilization which never achieves simulator capability or chooses not to use such for political reasons but still attains other technologies comparable to theirs in power.

"AR more potential future states in which we don't exist" - This is covered in the simulation argument as a possible solution, it doesn't break the argument if its part of it. Edit; Thinking some more about his, it doesn't matter how many potential future states in which we (or people like us) don't exist, there must be NO states in which we DO exist otherwise it would still be billions to one that we would be the 'real' one.

"But it is impossible to completely simulate a system from inside that system" - This is true, so if we WERE in a simulation you would expect some kind of 'fuzziness' to exist in most of the universe, perhaps collapsing into more exact points when important, a kind of 'uncertainty principal' :O.

Regarding you second point - you don't need to simulate the Universe at 1:1 scale to reliably replicate the past states of a certain region. For example, Earth itself can be simulated at full resolution only within a thin shell centered on it's surface, say +/-10 km from geoid. There are no observers at the core, so a simplified model would do, as long as plate tectonics is reproduced accurately enough.

A short story by Stephen Baxter, Touching Centauri, touches on this issue of model resulution and observers. [Spoilers ahead]

The protagonists use a powerful laser to bounce a pulse of energy off an exoplanet. This forces the entities who (it emerges) are simulating the universe (or maybe just our local portion of it) to massively increase the accuracy and extent of the simulation, so that the returned pulse behaves as physically expected. Previously they had been simulating the remote star system in a low resolution way due to the lack of observers there. The sudden need to increase the simulation's extent and resolution overwhelms their computational resources and the simulation crashes. A resource exhaustion attack on "reality", in other words.

That doesn't work though. Even a game of billiards is so complex and so dependent on external influences that a change in state in an electron at the edge of the observable universe can have an observable effect on the outcome.

http://www.anecdote.com/2007/10/the-billiard-ball-example/

It doesn't prevent simulating A universe. You can then compute a difference between the simulated past state and real past state based on (partial) information about the past state.

Then you go back and iterate again, eventually improving the accuracy of simulation until there is no distinguishable difference between simulated state and real past state.

The simulation might end up being not a perfect copy, but it would reach the point where it will be impossible to find any discrepancies.

If you have a way to find discrepancies between a simulation and reality, why do you need the simulation anymore?

I don't buy it. Even simulating just the earth and its environs would probably require a computer as big as a galaxy worth of matter. There's just no conceivable payback that could ever be worth the costs.

You have to actually compute the simulated state and compare it to known bits of the past state to get the results.

Yes, simulating Earth to sufficient precision might require resources of the Type III civilization. However, if something can be done, why shouldn't it be done?

The payback is twofold, actually:

First, you can get provide the meaningful afterlife to your ancestors.

Secondly, once you have hardware and software to simulate a virtual universe, you can create a multitude of them, with different properties and qualities. Think of the same logic behind splitting the physical computer into virtual machines.

I think a Type III civilization will have better things to do that precisely simulating me taking a crap every day. To me this argument smells the same as simplistic suggested solutions to real world problems. 'Why don't we just do X', such as why not just take all the excess food in the world and feed it to the starving people in Africa? Wow, how simple. Why did nobody think of that. Why wouldn't people in the future just simulate all of reality thousands of times? But easy to say or imagine != easy to do. I have severe doubts if even a Type II civilization can ever be viable.

> But it is impossible to completely simulate a system from inside that system--that's basic thermodynamics

I haven't heard quite that argument before. Couldn't it just be simulated slower than real-time? Or is memory the issue?

Regarding "Basic Thermo", everything we 'know', we subjectively acquired over a tiny period of time, through a game of telephone that refers to the current universe. We're really in no position to comment, in absolutes, on the nature of the parent universes.

In order to simulate anything you have to expend energy and create entropy to power the simulation.

But to be completely accurate, a simulation must include all the energy expenditures and entropy of the system being simulated. So the math can never add up.

"I haven't heard quite that argument before. Couldn't it just be simulated slower than real-time? Or is memory the issue?"

The "real time" in simulation is not affected by how long a single time step takes to evaluate in the real world.

https://xkcd.com/505/

Bostrom himself isn't committed to simulation. His actual argument is that either we're almost certainly living in a simulation or the stated assumptions are wrong.

Moreover, there's a difference between "the universe is a simulation" in the sense used in the article and "my conscious experiences are a simulation." The latter is unconstrained by "number of qubits in the universe." The apparent complexity of this universe you think you're inhabiting isn't real, and neither is Scott Aaronson.

I think point 2 is often handled by assuming everything that happens outside of our solar system (or wherever you choose to draw the line) is 'faked' in such a way that we couldn't tell.

> there are FAR more potential future states in which we don't exist, than states in which we do. The chances we are living in a simulation now are equally as small

It doesn't follow that all of these possibilities are equally likely, so I do not buy this line of reasoning

Your #2 implied that Bostrom himself mixed up the causality, but he did not. In a sense, it's your comment that's mixing up the causality. Just because it's confusing, the way Primer is confusing, doesn't mean it's illogical.