Living Inside A Simulation

[Deepwater6]

http://news.discovery.com/space/are-we-living-in-a-computer-simulation-2-121216.html

 

Thought's Hypogs?

 

Check out the slide show and captions in the middle of the story.

[CraigD]

Beane, Davoudi and Savage’s (the team of physicist mentioned in the Discovery article) “ Constraints on the Universe as a Numerical Simulation” paper is interesting and to be applauded as a serious effort to experimentally test the simulation hypothesis, but a well-known argument concludes that a such an experiment could only detect that our universe is a simulation if the makers of the simulation permit it: because a simulation program can be sensitive to the meaning and intentions of the the simulated physicists in it, it can cause an experiment like B, D & S’s to have a null result, even if the effect it is designed to detect actually exists. In other words, the simulation can cheat.

 

IMHO, everybody interested in questions like these should have a strong acquaintance and grasp of “Bostrom’s trillema”. Taken from the above linked wikipedia article, this is:

"A technologically mature "posthuman" civilization would have enormous computing power. Based on this empirical fact, the simulation argument shows that at least one of the following propositions is true:

1. The fraction of human-level civilizations that reach a posthuman stage is very close to zero;
   
   
2. The fraction of posthuman civilizations that are interested in running ancestor-simulations is very close to zero;
   
   
3. The fraction of all people with our kind of experiences that are living in a simulation is very close to one.
   
   

If (1) is true, then we will almost certainly go extinct before reaching posthumanity. If (2) is true, then there must be a strong convergence among the courses of advanced civilizations so that virtually none contains any relatively wealthy individuals who desire to run ancestor-simulations and are free to do so. If (3) is true, then we almost certainly live in a simulation. In the dark forest of our current ignorance, it seems sensible to apportion one’s credence roughly evenly between (1), (2), and (3).

 

Unless we are now living in a simulation, our descendants will almost certainly never run an ancestor-simulation."

While I think Bostrom's argument makes a critical unjustified assumption, I also think it's an excellent starting point for discussing its subject.

[tetrahedron]

Hoo, boy... I need a simulated drink...

[Erasmus00]

I've never been impressed with Bostrom's work, although that might be colored by my hugely negative impression of groups like the singularity institute.

 

As far as the trilemma, I think there are other possibilities. It seems probably that the most resource efficient way to do an ancestor simulation is to seed a planet with some simple lifeforms and wait- i.e. experiments might be easier than simulations. After all whatever a "posthuman" civilization might be operating over very long time scales.

 

Perhaps there is an upper bound on calculation density- Moore's law flattens out before simulating an entire civilization becomes feasible (there are physical limits to information density, there may well be physical limits to calculation density),etc

[CraigD]

Perhaps there is an upper bound on calculation density- Moore's law flattens out before simulating an entire civilization becomes feasible (there are physical limits to information density, there may well be physical limits to calculation density),etc

That’s essentially the “unjustified assumption” to what I was referring when I wrote

… I think Bostrom's argument makes a critical unjustified assumption ...

I have, I think, a strong unconscious bias typical a long-time IT pros like me toward skepticism of the prediction of practically unlimited computing power assumed common among posthuman/extropian futurist imaginings – an “I’ll believe it when I see it” attitude – but even allowing for that, I can imagine many showstoppers in creating a computer simulation of the a realistic universe.

 

One that comes to mind is that any successful “simulator” will have to actually be a universe – something like various ideas of “pocket” and “bubble” universe that a sufficiently advance engineer could make – and that, once made, it wouldn’t in a practically meaningful way be a simulation, because “simulation” implies a program that can be rewound, replayed, and inspected in unlimited detail and resolution, and the maker of such a universe wouldn’t be able to do these things. In most of the descriptions I’ve read, inspect their handiwork at all, because the created universe is causally disconnected from the one in which it was created.

 

If a sufficiently large and powerful but otherwise ordinary computer can realistically simulate a portion of something like our universe, I’ve a poorly defined suspicion that, on the scale at which quantum effects are detectable, it couldn’t work, because the rewind/replay/inspect nature of a simulation would violate quantum mechanical principles such as uncertainty. If (and this is a big if, in which I don’t personally believe) consciousness theorists like the New Mysterians are correct in their belief that brains can’t function without exploiting key quantum effects – in other words, that the human brain can’t be simulated with an ordinary computer, ancestor simulations just aren’t possible.

 

I've never been impressed with Bostrom's work, although that might be colored by my hugely negative impression of groups like the singularity institute.

SIAI, with its quoted goal of “ensuring powerful AIs are not dangerous when they are created”, does seem like a solution waiting for a problem that doesn’t yet, and may not soon or ever, exist. What AI programmer doesn’t dream of having to worry about ensuring that the powerful AI they are creating isn’t dangerous?! :)

 

While it’s waiting for potentially dangerous AIs to defang, SIAI strikes me as more of a conversation salon than a serious AI research institute. It brings together some thought-provoking ideas and people, though, so I like it. :thumbs_up

 

The concept of a Vingean singularity, from which the SI gets the first of its name, is viewed by most, I think, as a science fiction cautionary tale scenario – pretty much the killer-robot dystopia of the Terminator movies. From other perspectives – including mine – the singularity is an optimistic dream that expresses profound humility – that we can make human-like minds better than any natural human mind. So I like it, too. :thumbs_up

 

Now, a group of which I have a (I initially wrote “hugely”, but after reading some recent history, scratched that – there’s an interesting and somewhat sad/sweet story to be had around the rise, fall, and continuing life of Erik Drexler – see this 2004 Wired article) negative impression is the Foresight Institute. As a group, they seem to me pathologically unwilling to consider the most critical problems with their dream of a general purpose molecular assembler, contributing to a profound schism between by twin loves of science fiction and scientific reality.

 

As far as the trilemma, I think there are other possibilities. It seems probably that the most resource efficient way to do an ancestor simulation is to seed a planet with some simple lifeforms and wait- i.e. experiments might be easier than simulations. After all whatever a "posthuman" civilization might be operating over very long time scales.

Assuming that such simulations (that is, running a universe on a computer, rather than creating a real universe and watching it run) are possible, its users likely wouldn’t need to wait longer than they wanted for results, because just as with present-day physics simulations (eg: simulation of the formation of the solar system), there’s no requirement that simulation time progress at the same rate as real time. So our far-future simulation programmers might run a 4-billion year-long simulation of the appearance and evolution of life on Earth on a super-fast computer that would complete the run in seconds or minutes of real time – handy, if you’re trying to tweak initial conditions to produce some desired end result.

 

If these sorts of simulations aren’t possible for our imagined future folk, but engineering on planetary scales is, terraforming a real planet and running essentially a giant petri dish experiment might be the only way to go – but, coupled as experiments are to real time, the folk running them would have to be awfully patient. I’m reminded of an 1979 SF novel, Hot Sleep, where biologically ordinary human beings use medical suspended animation to live lives spanning a thousand of years by waking for only one or a few days every five years. Since our planet-size biological experimenters would presumably be good at such things, they might do something similar, perhaps using velocity or gravitational time dilation rather than life-suspending drugs.

 

A problem with this realtime-bound approach, however, is that the lifetime of many stars is not much longer than the required length of a biological evolution experiment. Our Sun, for instance, is about 5 billion years old, and has an expected lifetime as a main-sequence star of about 10 billion years. So, were we running a realtime biology experiment in a solar system like ours, there’d be time only for a few billion-year long runs before the sun went red giant the white dwarf.

 

Our super-engineering-enable biologists might need not just to build planets to order, but stars as well. It’s only my gut-instinct, by programming computers to simulate biospheres in much-faster-than-realtime seems an easier, less mind-boggling approach to me. ;)

[Erasmus00]

If a sufficiently large and powerful but otherwise ordinary computer can realistically simulate a portion of something like our universe, I’ve a poorly defined suspicion that, on the scale at which quantum effects are detectable, it couldn’t work, because the rewind/replay/inspect nature of a simulation would violate quantum mechanical principles such as uncertainty. If (and this is a big if, in which I don’t personally believe) consciousness theorists like the New Mysterians are correct in their belief that brains can’t function without exploiting key quantum effects – in other words, that the human brain can’t be simulated with an ordinary computer, ancestor simulations just aren’t possible.

 

It depends on what you mean by simulation. I'd suggest something like an emulated person is a type of simulation, even if the rewind,replay, inspect properties can't exist.

 

While it’s waiting for potentially dangerous AIs to defang, SIAI strikes me as more of a conversation salon than a serious AI research institute. It brings together some thought-provoking ideas and people, though, so I like it. :thumbs_up

 

My largest problem is that (before I'd even heard of the institute), I had a student who had learned some rather strange notions of both statistics and physics from some singularity institute literature. Its very hard to get someone pre-committed to wrong ideas to unlearn them!

 

When I dove into looking at the singularity institute, they struck me as being a lot like the cryonicists, Drexler, etc. They strike me as dilettantes- dabbling in hard problems but unwilling to do the hardwork of coming to terms with the scientific state-of-the-art so they can make real contributions. Instead, they bandy about wild-ideas and argue about meta-issues. I guess everyone needs a hobby, but these guys fund-raise and set-up as non-for-profits (and in the case of the cryonics guys offer a paid service that clearly doesn't work). I'm willing to bet many of their small donors lack the relevant expertise to separate them from real scientists. The only difference between them and con-artists is their fervent belief.

 

When last I checked, after a decade and hundreds of thousands of dollars, the singularity institute had produced only a few self-published non peer-reviewed papers, a lot of blog entries, and ~1/3 of a rather long Harry Potter fan story.

 

As a group, they seem to me pathologically unwilling to consider the most critical problems with their dream of a general purpose molecular assembler, contributing to a profound schism between by twin loves of science fiction and scientific reality.

 

I'd agree and argue a similar problem plagues the singularity institute and probably most self-described "trans-humanist" groups.

 

Assuming that such simulations (that is, running a universe on a computer, rather than creating a real universe and watching it run) are possible, its users likely wouldn’t need to wait longer than they wanted for results, because just as with present-day physics simulations (eg: simulation of the formation of the solar system), there’s no requirement that simulation time progress at the same rate as real time. So our far-future simulation programmers might run a 4-billion year-long simulation of the appearance and evolution of life on Earth on a super-fast computer that would complete the run in seconds or minutes of real time – handy, if you’re trying to tweak initial conditions to produce some desired end result.

 

If these sorts of simulations aren’t possible for our imagined future folk, but engineering on planetary scales is, terraforming a real planet and running essentially a giant petri dish experiment might be the only way to go – but, coupled as experiments are to real time, the folk running them would have to be awfully patient....

 

Our super-engineering-enable biologists might need not just to build planets to order, but stars as well. It’s only my gut-instinct, by programming computers to simulate biospheres in much-faster-than-realtime seems an easier, less mind-boggling approach to me. ;)

 

I guess I'm sort of operating on the premise that the universe is really big. Building insanely powerful, insanely energy hungry computers to run simulations might just be orders of magnitude more costly than finding N planets that fit what you want, dropping some life on them, and returning to visit every few hundred years. Assuming arbitrarily long-lived super-people, I can see that they might take the much cheaper option (this assumes interstellar travel is cheaper than mega-computers that can simulate planets). You might not be able to do a ton of experiments end-to-end, but tons of parallel runs are possible.

 

I think the only life-form we've successfully simulated (and not in full detail!) is the tobacco mosaic virus, which took weeks of super-computer time and was simulated for only about 50ns. Even assuming orders of magnitude more power, simulating whole civilizations is likely to be prohibitively expensive.

 

That said, my officemate in graduate school used to say that things like an arbitrary max. speed and a limited distance/momentum resolution were just the sort of shortcuts he'd put in a simulation.

Edited December 21, 2012 by Erasmus00