The Singularity and Simulation

There is a nice video of a recent talk by David Chalmers on the singularity available here. Dave also summarizes the argument in a recent post at his blog Fragments of Consciousness (here). He also gave this talk at the Graduate Center, which is where I saw it last Wednesday. It is an excellent talk and I hope it starts people talking about these interesting issues. Assuming you believe that AI is a possibility I find the general line he is pushing very persuasive and would be interested to hear what others thought about it. 

One thought that I had was that if the second premise of the argument is right then we might have some kind of evidence that we are not living in a simulated world. If we were we would be the AI and the second premise says that once you have AI it will be a matter of years before you have AI+, but we haven’t had AI+ yet (i.e. strong A.I.) so we are not AI. When I asked about this Dave responded that ‘a matter of years’ should be interpreted as in the time scale of the next world up. If we are indeed in a simulated world then the simulators of our world could presumably manipulate the time scale in the simulated world. So what may seem like a long time to us could be a few seconds for them. Ah well, I guess we still can’t be sure that we aren’t in the Matrix. 
This allows me to clarify the point of my previous post. In discussion with Dave about it he pointed out that what I describe is just one kind of dualism and that it is not the kind that the zombie argument deals with. This is a fair point. Looking back at the post I see that I was sloppy in presenting the argument. I should not have been saying that the zombie argument by itself is an argument that we are in a simulated world. What I should have said is that this account of what a nonphysical property is is the only one that is one the table. But when we adopt this as a theoretical account of what non-physical properties are even zombies can have them and so they do not seem to threaten physicalism. If there is some other account of what a nonphysical property is then we can examine it and one cannot say that an obvious example of a nonphysical property is seeing green or feeling pain. What is needed is an account of what it would mean to say that feeling pain is nonphysical. I, for one, can’t even conceive what that would mean except in the way Dave does in his matrix paper.  

14 Comments

  1. gualtiero

    Richard, thanks for the post. I’d like to understand why we are supposed to take this singularity business seriously. From what I’ve heard (without having read the relevant literature), these singularity people fallaciously shift from (1) computers are becoming more powerful (true) to (2) computers are becoming more intelligent (true if ‘intelligence’ means what comes with computation power; questionable if ‘intelligence’ means what it usually means, including insight, intuition, etc.) to (3) computers are about to become conscious (a non sequitur). What am I missing?

  2. gualtiero

    One more comment about conceiving of nonphysical properties. As per my comment to your previous post, I have no idea what it means to say that a property is nonphysical, period. It doesn’t surprise me that you haven’t found any account of that (except for the one in the matrix paper, which I still find unintelligible).

  3. Richard Brown

    Yeah, I really think this is the biggest challenge to the dualist. If they can’t supply even some general account of what kind of thing a nonphysical property is supposed to be then dualism isn’t even a coherent position.  Usually they appeal to the qualitative properties at this point but that is just question begging in the extreme. So as it stands dualism is either radically question begging (since it simply defines qualitative properties as nonphysical) or simply incoherent. 

  4. Richard Brown

    Dave was cautious on this point and suggested that we might have AI within 3,000 years and that people who think it will happen within the next 50 years or so are a bit overly optimistic. Nor does Dave think that computer consciousness is necessary. All that he needs is that we can can make an agent that reasons better than we do in the second sense. The argument that we will be able to do this is pretty simple. Dave thinks that evolution did it naturally and so our best bet will be to try to artificially evolve an intelligent agent in a simulated world. If nature could do it using only chance then it seems reasonable to think that adding a little intelligent design to the process would do it even better. This is not a knock down argument that we will produce AI but it at least shows that it seems plausible. Now, as to how long it will take that is another issue. And it is true that Dave thinks, for independent reasons, that functional duplicates of human beings here in the actual world will be conscious (since consciousness is a ‘functional invariant’ for dave) the argument for intelligence explosion doesn’t depend on it. 

  5. Well, here’s one suggestion for a nonphysical property: being even, as in an even number. Really just about any mathematical property will do, or so it seems to me.

    This is something I’ve wondered about (but never taken the time to really research): unless I’m really wrong in my conception of mathematical properties being nonphysical, this seems like a potential problem for those physicalists who think that math is important and true (and I imagine many do).

  6. Richard Brown

    Hi Corey thanks for the comment.

    I think it is question begging to assume that mathematical properties are nonphysical. What we have is a datum, say that the number four is even, this is something that everyone can agree one. The empiricist/physicalist does not dispute the fact that four is even they give an empiricist account of this fact,  which usually takes a broadly Humean form (though there are other candidates like fictionalism and Mill’s view). What we need here is an argument that the property of being even isn’t a physical property, or alternatively we need to compare accounts of what it means to be even and see whose account is simpler, more elegant, etc. This requires an independent account of nonphysical properties which no one can give, or at least no one has given it yet. On the other hand we have generally well worked out accounts of what physical properties are and the kinds of things they can do so until we have some general account of nonphysical properties theories that only appeal to physical properties should be preferred. 
    Usually, the only evidence offered that mathematical properties can’t be physical properties are experiences we might call ‘rational seemings’. Rational seemings are experiences as of being rationally compelled to believe that something is the case, as when one contemplates the validity of modus ponens (this is what the ‘clear and distinct’ metaphor amounts to in Descartes for instance). However rational seemings are fallible (as evidenced by things which have historically seemed to be a priori true which weren’t) and there is a plausible empiricist alternative for why we have these experiences; viz some things seem to be rationally compelling because they are the kinds of things which were selected for via evolution and so ultimately are justified empirically. This completely explains why some things seem to be justified independently of experience but aren’t. 
    As for the second point physicalists in the Quinian tradition usually take it the other way. they argue that since math is important because it is important to physics and since physics is an empirically justified science that means that math is empirically justified as well. But as you say, math is only a problem for the physicalist if one assumes that mathematical properties aren’t physical (I have discussed these issues before over at Philosophy Sucks!  herehere and here if you are interested)
  7. Richard,

    Thanks for the comments. Here are some further thoughts:

    I suppose much of our disagreement is going to depend on what counts as a physical property, and where we each think the burden of proof lies. I would think that a physical property is either one that is in the domain of physics, or that supervenes on (or reduces to, or has some strong relationship to) properties you find in physics. And I’m certainly assuming that numbers are nonphysical objects, so their properties are as well. Whereas you characterize this as question-begging, I would ask: what could possibly make it true that numbers (and their properties) are physical? Maybe we cannot make sense of physics (and all that supervenes upon it) without numbers, or maybe we can (I think this is one of Hartry Field’s projects). But that just seems to be a separate issue.

    As to the fallibility of what you call “rational seemings,” while that may be true for many experiences, I do not see how I could be wrong about the mathematical facts that I know. You give the example in one of the links you provided of counting a set of 6 things, a set of 7 things, and putting them together to get a set that everyone says has 14 things. This is supposed to be an example of a possible empirical refutation of six plus seven equalling thirteen. Why not say, instead, that this is an example of a very surprising property of whatever it is that’s being counted? When things get weird in quantum physics, the math and logic doesn’t go out the window: the (initially) straightforward metaphysics of the objects in question do.

    Since that’s obviously a very messy issue, here’s another candidate. Perhaps the property of being a genuine dollar bill is a nonphysical property, because a physical duplicate of a genuine dollar bill is, quite plausibly, a counterfeit, and thus not a genuine, dollar bill. So the “genuine dollar” facts do not supervene on the physical facts (insofar as we might think of the physical facts as being the occurrent, three-dimensional, empirically-measurable physical properties). Or, quite similarly, being holy water is probably a nonphysical property that similarly depends on having the right kind of history, and not on physical facts.

  8. Richard Brown

    Thanks Corey for these thoughts, some in return;

    I certainly agree with your charectarization of physical properties but I see no reason to think that numbers must be, or even actually are, nonphysical. One way it would be true that numbers were physical is if numbers and mathematical truths in general were simply formal systems of inter-related and inter-defined concepts (this is roughly Hume’s relations of ideas), another way would be for Harty’s project to be true. If there are no numbers and yet we can do science just fine then our world could be completely physical and we would still have the mathematical truths (well, we would have enough to  get the science done just by their being true in some possible world), still another would be Mill’s account in terms of empirical generalizations. On that view mathematics is just an extremely well-confirmed empirical theory. As far as I can see these are all live possibilities, though I tend to oscillate between Hume and Mill…
    I also agree that it is extremely hard to see how mathematical facts could be anything other than necessary but of course the argument I gave predicts that we would feel this way. Suppose that mathematical truths are actually just very general statements of regularities in the physical world that could be otherwise. In such a world it is reasonable to think that creatures that evolved there might evolve representations designed to track these regularities. From the point of view of one of these evolved creatures the mathematical truths will seem necessary even when they aren’t. To put this another way, because of the innateness of the representations these creatures might think that their beliefs are justified independently of their experience when really these things are the result of a lot of empirical ‘tuning’ by the environment. Once we have this alternative explanation for why some things rationally seem necessary but aren’t the question becomes one of which story is more likely. I find the evolutionary story much more plausible. 
    I also agree that we might be tempted to throw out the metaphysics of sheep or whatever, but we can design the thought experiment to rule this out. The point, I think, is just that it seems conceivable that if we had the right kind of evidence we would be forced to alter our conception of mathematics and if that is the case then math is in fact an empirically justified science. So, we could tell the story that this is a wide spread phenomenon that happens every time you have groups of 6 things and seven things combined, and further analysis reveals that these are ordinary objects that do not metaphysically misbehave. 
    Finally, I think that the kind of historical facts you are talking about can be deduced from a completed micro-physics and so are themselves physical facts. Your intuition that we could have physical duplicates of genuine dollars that were counterfeits is evidence that being a genuine dollar bill is a contingent property of dollar bills not evidence that it is a nonphysical property. Of course some philosophers deny this (Block is a notable example I think), and say that no high level properties are deducible from the micro-physics but I think there are convincing arguments against this position. 
  9. Thanks Richard,

    I think I may hear that characteristically dull thud of a basic intuition clash here! But this is fun.

    First, I’m confused about why being part of a formal system of concepts makes something a physical property, except insofar as concepts might have to be had by physical systems of some sort or another.

    Also, I don’t know what it would mean to suppose that mathematical facts are just generalizations of experience: that’s just not taking math very seriously, or it’s taking “generalization” way too seriously.

    And I’m really uncertain of how your thought experiment could go. I have no idea what it would mean to have an environment in which a set of six sheep wander over to a set of seven sheep and the local counters then find that there are 14 sheep (without saying something about the weird metaphysics of sheep popping into and out of existence, or that their names for numbers are just different than mine), just as I have no idea what it would mean to have an environment in which the local seers can both see a tree in front of them and not see a tree in front of them, or see a circular square. And I really have no idea how evolution would work in such an environment.

    What I can imagine is an environment in which the physical facts–which I take to be generalizations of experience–differ. It is physically necessary that stuff falls around the surface of the planet, and this fact seems to come from some combination of experience and the innate setup of our brain. So I can imagine creatures evolving in gravity-free environments generalizing their experiences with objects such that they would not think that objects must necessarily fall. But this is just to make a distinction between physical and logical necessity; I would include math on the logical side, whereas it seems that you want to put it on the physical side. I know what it would mean for physical necessities to be different, even though I have no experience with different experiences, whereas I do not know what it mean for the logical necessities to be different.

    So I think this is the dull thud: I do not think an environmental regularity in which six and seven objects result in a 14 objects, where the numbers mean what they normally do, and addition means what it normally does (i.e. no metaphysical weirdness), is conceivable, just as I do not think an environment full of circular squares and non-copper samples of copper is conceivable. I can say the words, but that’s it.

    Finally, you may be right about a completed physics, but that seems awfully optimistic. It seems unlikely that we could ever build a holy water detector, given that what makes a sample of water holy or not is a fact about that bit of water’s history (i.e. its interaction with particular people). What sets of values of the physical properties of water could determine whether a sample of water had the right history or not?

  10. Richard Brown

    Hi again Cory,

    I agree; this is fun.

    “First, I’m confused about why being part of a formal system of concepts makes something a physical property,”

    Sorry I meant to be talking about a physicalist who has this view. In a Humean world where there were only physical properties we could have conceptual relations. Therefore, we could have the conceptual relations that give us the mathematical truths. Therefore the mathematical truths could be physical truths. This way of doing it has the advantage of saving something like the necessity of mathematical truths (because they are analytic)…although I tend to think that even analytic truths are, at least in part, empirically justified.

    “Also, I don’t know what it would mean to suppose that mathematical facts are just generalizations of experience:”

    The idea is just that there has been this very long regularity in nature involving the kinds of experiences that confirm basic mathematical truths. Or alternatively arithmetic is just an extremely well confirmed physical theory, no different than quantum mechanics in that respect. So far as we can tell every prediction made by basic arithmetic has been born out empirically. The difference in the case of mathematics is that it is so well confirmed that it was ‘built into’ our early ancestors (much like basic logic and Newtonian physics seems to be). But if our experience had been different the things which seem inconceivable would be different. So I agree that it is very hard to make sense of the idea that 2+2 could be other than 4 but that could be simply because I am so well trained by my experience. The problem here is that both your account and my account predict all of the same experiences, or rational seemings. If I am right then it will seem to me to be inconceivable that 2+2 be other than 4 even though it could be other than 4 whereas if you are right it will seem that way to me and I will be right. So, how do you decide between them? At this point it seems to me that the only way to do so is via considerations about simplicity and basic theoretical coherence and at that level the nonphysicalist fares very poorly.

    “It seems unlikely that we could ever build a holy water detector, given that what makes a sample of water holy or not is a fact about that bit of water’s history (i.e. its interaction with particular people).”

    You would have to be able to “rewind the tape” as it were and retrodict the water’s history. This would require knowing the positions and forces governing all other fundamental entities but it seems in principle doable. It doesn’t seem any different, say, than the way that we would determine the velocity of some object. To do so we need to know were it was earlier and where it will be later but once we have that information we can deduce its velocity.

  11. Richard,

    Seems we’ve gotten on a bit of a tangent here!

    First, I think that there’s a simple terminological dispute. Like I said before, I’m taking “physical property” to be either things like charge, spin, and whatever else the physicists talk about, or properties that have a pretty straightforward relation to sets of these properties, like density, being a noble gas, and so on. That’s vague, but it’s meant to refer to the kinds of things that people who do what we call physical science talk about. It seems to me that you’re taking “physical property” to be properties that can be had in some way or another by physical things. I don’t think there’s any interesting relationship between physical properties (in my sense) and being a copy of Moby Dick, although it’s probably true that, to be a copy of Moby Dick, a thing has to be a physical object of some sort or another (which, I take it, would suggest to you that being a copy of Moby Dick is, in fact, a physical property). Obviously I like my way of talking better, but I don’t think we’re actually disagreeing.

    So back to really disagreeing!

    Any story about innateness explaining our inability to think of mathematical/logical facts being different seems like a non-starter to me. Some kind of folk physics might be innate, and there are a bunch of ways of seeing that are innate (hence optical illusions). There might even be a good evolutionary story about how this all came to be. But that doesn’t preclude our knowing what it would be like for things to have been different. I can’t fail to see craters lit from a certain angle as hills, but I can imagine an organism not having this perceptual feature, because I can imagine an organism evolving in a world in which, say, all the light comes from the surface, rather than the sky. Our species, evolutionarily-speaking, has had virtually no experience with physics not near the surface of the earth, which is probably part of the explanation for our innate physical beliefs. But that doesn’t at all prevent anyone from thinking about how the world could be different with respect to whatever made us have those innate physical beliefs. So being innate doesn’t preclude being able to conceive of differences.

    So it seems that any explanation from innateness would have to give us a story about why imagining variation in one kind of fact is so much easier than another, given the parity of experience we’ve had with the two kinds of facts. From what you might want to call an “evolutionary perspective” it would seem that folk physics is much more confirmed than math as a whole (as a species, we’ve had way more experience–presumably without any variation–with things falling near the surface of the earth in a certain way than we have with quaternions or tensors); nevertheless, it’s easy to imagine the physical facts being different, but not the mathematical facts.

  12. Richard Brown

    Hi Corey, sorry I haven’t got back to you sooner! 

    I am not sure if we are disagreeing about ‘physical property’ or not. For me a property is a physical property if it is one of the kind that you would call a physical property or is one that can be deduced from those properties. So, I take it that the existence of tables can be deduced from talk about ‘charge, spin, and whatever else physicists talk about’. Is it the case that you think being a table is a nonphysical property that some arrangement of properties like spin, charge, etc have? If so then the dispute may be merely terminological…
    I like your objection to the innateness hypothesis I put forward. Firstly, though, I would want to restrict our attention to discussion of simple arithmetic truths like 2+2=4, that 1+1=2, or that 1/2(30)=15. These kinds of basic arithmetical truths have been confirmed just as much as the hypothesis that things near the surface fall. This goes as well for the axioms of first-order logic (or the rules of inference in natural deduction). But I agree with your general claim that being innate doesn’t preclude us from imagining differences; I merely claim that it makes it harder to do so. I think this is the case even with the laws of physics. Aristotle took the laws of physics to be necessary truths, whereas Humeans disagree. Descartes thought that God could have made any laws of logic or mathematics that He wanted to and so that they were contingent on His choices. Of course it was hard to conceive of how this could be the case until things like the liars paradox emerged. So it seems to me that people have imagined that the mathematical laws be other than they actually are. Is it the case that, in some sense, it is harder to imagine 2+2 not being 4 than it is to imagine objects that don’t fall when dropped? I think yes but this is actually because simple arithmetic truths are better confirmed. There are not any seeming counter examples to it (whereas there are for gravity). 

Comments are closed.

Back to Top