In 2010 I explained why I guess I’m not in a sim. In 2011 I explained why sims should be small, and focus on “interesting” folks. In 2001 I explained why it matters if you live in a sim. Here is Tyler
So the question becomes this:If the mediocrity principle is true then would the universe create more computationally limited sims or more computationally unlimited sims?
I think the evidence is in favor of more computationally limited sims but we don't have sims running on quantum computers yet so the answer is not in...
Presumably by not having aliens one could run a simulation of life on one world with considerably less computational costs by simply using vague approximations to handle all the physics outside of a narrow region around the planet of interest.
Fourth option: the simulators know that the suffering is somehow worthwhile or outweighed by other benefits it makes possible. Religious people often make similar arguments about gods.
Fifth option: we are just over-weighting suffering in our analysis. Maybe it's not really that bad and we shouldn't take it so seriously. It's just a dashboard indicator of your prospects for longevity and reproduction.
I find the simulation argument implausible because either (1) it's about the whole universe (which the Copernican in me thinks is most plausible), in which case that's a lot of resources to expend, or (2) it's somehow focused on me (the minimal alternative), and frankly while I'm an overconfident guy, I just don't think I'm worthy of that kind of attention.
I think there's a third option: the simulators might not know about your existence. The simulation could exist for some other purpose, with humanity as an obscure and unexpected side effect.
Yes - Bostrom-style, we can say that one of the following is true: (a) the simulators are evil, or (b) all suffering people are NPCs. Since I know I'm not an NPC, I know the simulators are evil, but cheery folks are in a more difficult epistemic position.
I think we'll probably have a nematode simulation in about ten years. People have been working on this for at least 15 years [1], so that would be 25 years for a nematode simulation. The amount of discovery and innovation needed to simulate a nematode seems maybe 1/100th as much as for a person. [4] Naively this would say 100 * (15+10) or 2500 years for human whole brain emulation. More people would probably work on this if we had initial successes and it looked practical, though, giving us maybe a 10x boost? Which still is (100/10) * (15+10) or 250 years.He pulled these factors of 100 and 10 out of the air as far as I can see.
May I presume that we have some Civilization players in the reading audience? I'm usually a more peaceful than warlike player. There are three settings for barbarians: normal, off, or "raging." Raging barbarian hordes can certainly make things more interesting, but they also tend to muck up my neat plans for civilization development. Sometimes you want to control or eliminate those potentially interesting variables.
I would just add that IMO Robin's simplifying assumption that 'story value' is the only consideration affecting inclusion of aliens in a sim is a bit rash. It seems to me that there are a number of other considerations which might affect that decision (some of which are mentioned in other comments). Given this, basing one's estimate of p(silence|sim) on story value alone may introduce a bias. I find it difficult to say more because we're getting pretty far out on a speculative twig, if not leaf:-)
1. By Occam's razor, we are not in a sim. 2. If we were in a sim, there would exist (or have existed) at least one set of "aliens" - the beings that created and/or are running the sim. And in that case, those beings would seem to be the only "aliens" that should actually matter much to us.
I'm surprised that nobody's mentioned that it could be much less computationally intensive to run a sim of human experiences on earth than to run a sim of the entire apparent observable universe.
The information content that gets through to earth is presumably many orders of magnitude lower than the information content of the information content that describes the observed universe if we're not in a sim.
Indeed, Robin has touched on the theme of the implications of bounded computational resources on sims in one of the blog posts that he links to:
Today, small-scale coarse simulations are far cheaper than large-scale detailed simulations, and so we run far more of the first type than the second. I expect the same to hold for posthuman simulations of humans – most simulation resources will be allocated to simulations far smaller than an entire human history, and so most simulated humans would be found in such smaller simulations.
To put zmil's and arch1's criticism on a more rigorous footing:"And if you can’t tell if aliens help or hinder a sim story, then not seeing aliens gives no info about if you are in a sim."This claims that p(sim|silence) = p(sim) if p(silence|sim) is unknown. But actually, since p(sim|silence) = p(silence|sim)*p(sim)/p(silence), the claim is that "if you can’t tell if aliens help or hinder a sim story," then p(silence|sim) = p(silence). But this doesn't follow; if you have no information about p(silence|sim), you might set it equal to 1/2; if p(silence|not sim) is lower than that, silence is evidence for being in a sim.
I agree that aliens probably add to a story, so I would set p(silence|sim) below 1/2, but I'm nowhere near certain enough to put it at the astronomically low values that constitute the low range of estimates for p(silence|not sim), and I'd be shocked if anyone claimed to be.
@Peter Gerdes:
Agreed!
So the question becomes this:If the mediocrity principle is true then would the universe create more computationally limited sims or more computationally unlimited sims?
I think the evidence is in favor of more computationally limited sims but we don't have sims running on quantum computers yet so the answer is not in...
Presumably by not having aliens one could run a simulation of life on one world with considerably less computational costs by simply using vague approximations to handle all the physics outside of a narrow region around the planet of interest.
Fourth option: the simulators know that the suffering is somehow worthwhile or outweighed by other benefits it makes possible. Religious people often make similar arguments about gods.
Fifth option: we are just over-weighting suffering in our analysis. Maybe it's not really that bad and we shouldn't take it so seriously. It's just a dashboard indicator of your prospects for longevity and reproduction.
I find the simulation argument implausible because either (1) it's about the whole universe (which the Copernican in me thinks is most plausible), in which case that's a lot of resources to expend, or (2) it's somehow focused on me (the minimal alternative), and frankly while I'm an overconfident guy, I just don't think I'm worthy of that kind of attention.
Ah yes, good point - the third prong of the Problem of Evil, omniscience.
I think there's a third option: the simulators might not know about your existence. The simulation could exist for some other purpose, with humanity as an obscure and unexpected side effect.
Yes - Bostrom-style, we can say that one of the following is true: (a) the simulators are evil, or (b) all suffering people are NPCs. Since I know I'm not an NPC, I know the simulators are evil, but cheery folks are in a more difficult epistemic position.
This is his entire analysis:
I think we'll probably have a nematode simulation in about ten years. People have been working on this for at least 15 years [1], so that would be 25 years for a nematode simulation. The amount of discovery and innovation needed to simulate a nematode seems maybe 1/100th as much as for a person. [4] Naively this would say 100 * (15+10) or 2500 years for human whole brain emulation. More people would probably work on this if we had initial successes and it looked practical, though, giving us maybe a 10x boost? Which still is (100/10) * (15+10) or 250 years.He pulled these factors of 100 and 10 out of the air as far as I can see.
Robin, what do you think of Jeff Kauffman's take on brain emulation? It was linked in the comment's at MR.
May I presume that we have some Civilization players in the reading audience? I'm usually a more peaceful than warlike player. There are three settings for barbarians: normal, off, or "raging." Raging barbarian hordes can certainly make things more interesting, but they also tend to muck up my neat plans for civilization development. Sometimes you want to control or eliminate those potentially interesting variables.
I'm also with Brian Nachbar and multifoliaterose.
Brian, really well put - thanks.
I would just add that IMO Robin's simplifying assumption that 'story value' is the only consideration affecting inclusion of aliens in a sim is a bit rash. It seems to me that there are a number of other considerations which might affect that decision (some of which are mentioned in other comments). Given this, basing one's estimate of p(silence|sim) on story value alone may introduce a bias. I find it difficult to say more because we're getting pretty far out on a speculative twig, if not leaf:-)
1. By Occam's razor, we are not in a sim. 2. If we were in a sim, there would exist (or have existed) at least one set of "aliens" - the beings that created and/or are running the sim. And in that case, those beings would seem to be the only "aliens" that should actually matter much to us.
Exactly. Thanks.
I'm surprised that nobody's mentioned that it could be much less computationally intensive to run a sim of human experiences on earth than to run a sim of the entire apparent observable universe.
The information content that gets through to earth is presumably many orders of magnitude lower than the information content of the information content that describes the observed universe if we're not in a sim.
Indeed, Robin has touched on the theme of the implications of bounded computational resources on sims in one of the blog posts that he links to:
Today, small-scale coarse simulations are far cheaper than large-scale detailed simulations, and so we run far more of the first type than the second. I expect the same to hold for posthuman simulations of humans – most simulation resources will be allocated to simulations far smaller than an entire human history, and so most simulated humans would be found in such smaller simulations.
not necessarily, the examples you see of extreme suffering might have no internal experience simulated.
Maybe they are running the simulation to determine what would have happened to Earth if humans had never met aliens.
To put zmil's and arch1's criticism on a more rigorous footing:"And if you can’t tell if aliens help or hinder a sim story, then not seeing aliens gives no info about if you are in a sim."This claims that p(sim|silence) = p(sim) if p(silence|sim) is unknown. But actually, since p(sim|silence) = p(silence|sim)*p(sim)/p(silence), the claim is that "if you can’t tell if aliens help or hinder a sim story," then p(silence|sim) = p(silence). But this doesn't follow; if you have no information about p(silence|sim), you might set it equal to 1/2; if p(silence|not sim) is lower than that, silence is evidence for being in a sim.
I agree that aliens probably add to a story, so I would set p(silence|sim) below 1/2, but I'm nowhere near certain enough to put it at the astronomically low values that constitute the low range of estimates for p(silence|not sim), and I'd be shocked if anyone claimed to be.