On the "Simulation argument"
-
- Posts: 4696
- Joined: February 1st, 2017, 1:06 am
On the "Simulation argument"
Bostrum's original paper is here:
https://www.simulation-argument.com/simulation
After reading Bostrum's paper several years ago I wrote a rebuttal. I intended to submit it for publication, but never got around to it. So here it is. (It is a full essay, so those with short attention spans may wish to move on).
-------------------------------
Simulation Argument? Not!
G. E. Morton
I was pondering Nick Bostrom's "simulation argument" when my cat, Katy, just awakened from a catnap, yawned and hopped into my lap for a snuggle. As she lay there purring and enjoying her quality time, I began to wonder if Katy might be a simulated cat. Of course she looked and felt and sounded like a real cat, but according to Bostrom, that is no guarantee that she is a real cat. For, conceivably, a sufficiently advanced "posthuman" civilization might develop computers powerful enough to simulate, not only Katy, but me as well. This simulated "me" would observe and interact with a world indistinguishable from our present, presumably "real" world, complete with purring cats. Furthermore, Bostrom goes on to suggest, if the simulation were allowed to run long enough, the inhabitants of this simulated world might themselves create powerful (virtual) computers which could generate compelling simulations of their own. Thus Katy might be, not just a simulated cat, but a simulation of a simulated cat, or even a simulation of a simulation of a simulation of . . . .
At that point my thoughts were interrupted by a slight trembling of the floor beneath my (simulated?) feet. A simulated mild earthquake? But then I identified the cause of the tremor -- it was a shock wave generated by the bones of the venerable Occam spinning in his grave. Bostrom, I realized, was offering an explanation of the cat in my lap which involved multiple layers of ontology, all perhaps constructed somewhat differently (for there is no requirement that the virtual hardware and software running each successive simulation be identical), but each perfectly invisible and transparent to its own virtual humans.
Well, that realization dampened my enthusiasm for the simulation argument somewhat. Occam's Principle, which holds that explanations for our experience ought not multiply entities needlessly, had always seemed a fairly prudent maxim to me. Instead of simply supposing Katy to be a real cat, and trying to account for her presence in the world via the usual biological and evolutionary investigations, Bostrom was suggesting we may be obliged to reconstruct a number of layers of virtual software lying between our virtual selves and the real Katy, the prototype Katy, wherever she might be.
Now, note that being a *simulation* presupposes a prototype. A simulation, I take it, is not a sui generis invention, but a copy or imitation of something else, rendered more or less faithfully. Thus a series of simulations must terminate eventually in a real world, with real cats. So even if we are living in a simulated world, there must be a real one somewhere. Moreover, our efforts to fully explain our experience must eventually explain this real world, for without that explanation all of the simulated layers are left hanging, as it were. They become "nomological danglers." Indeed, we cannot begin to explain any of them unless we have already explained, in some detail, the real world they are presumed to simulate.
Thus it appears that accepting Bostrom's hypothesis gains us nothing in terms of explaining the world in which we find ourselves. We still must explain a real world, with all of the familiar difficulties of that enterprise. But in addition, we must now also explain an uncertain number of intervening simulated worlds, and the mechanisms that generate and maintain them, lying between us and that real world. And by hypothesis, these layers neither add nor substract anything to our experience --- they are perfectly transparent, perceptually inseparable from the real world they are simulating. What explanatory advantage, then, do we gain by postulating them?
But perhaps we can imagine that our world is not actually a simulation, strictly speaking, but is indeed a sui generis construct, a work of original art created ex nihilo in the mind of the programmer, which exists nowhere except in that mind and in the computer running the software. Perhaps the finished reality --- the one we experience --- does not even exist fully realized in the mind of the programmer; perhaps the software itself has some capacity for creativity, and once launched, generates a world not even the programmer fully anticipated. After all, we ourselves can write computer software that behaves in unexpected ways. Our world, in that case, is not a simulation, but is wholly artificial. And following Bostrom, we can even imagine that this software might generate virtual beings themselves sophisticated enough to write similar software. Et cetera. In this case our regress need not terminate in a real world, since there is no original prototype world; each virtual layer is sui generis and perhaps unique, each the product of the creativity of its virtual programmers.
Whew. But does this simplify things? It's true that we no longer have a "real world" to account for before we can adequately account for the successive simulations of it. On the other hand, this succession of virtual worlds need not terminate anywhere, and indeed cannot do so in any principled way. It is an infinite regress, for the inhabitants of none of the virtual worlds has any better reason for supposing their world is the "real" one than we do. None of them, even as we, no matter how successfully they believe they have explained their own world, can be certain they are not the products of a virtual creator. We might therefore imagine them asking, just as we do, "What is the point of this hypothesis, anyway?"
Having quieted the stirrings of Occam's soul, other oddities of Bostrom's hypothesis occurred to me. Brother William's bones are not the only ones becoming restless. Those of Baron von Leibniz are also thumping at the lid of their coffin. Leibniz proposed a principle, known as the principle of the Identity of Indiscernibles), which holds that if you are presented with two allegedly distinct entities or states of affairs, and there is no means conceivable for distinguishing between them, then they are the same thing. II is actually closely related to Occam's principle: why postulate two things when one will do? And if two alleged things cannot be distinguished in any possible way, then one will clearly do. Adding the second makes no difference, because you can't distinguish a situation where it is present from one in which it is absent, provided the first is present.
But Bostrom's scenario presents two presumably distinct situations, two possibilities --- that we live in a real world, and that we live in a virtual world --- and then goes on to assert that these two situations cannot be distinguished by any conceivable means. We cannot tell which of them is the case. But if we cannot, then why concoct the two scenarios in the first place? Why should we even think we have two scenarios, rather than just one, described in different ways?
Bertrand Russell once offered the following as an examples of scientifically meaningless propositions:
1. The universe and everything in it is doubling in size every second.
2. The universe was created 10 minutes ago, complete with fossils and human beings with totally fictitious memories of their pasts.
These are commonly cited in logic and philosophy of science texts as examples of empty empirical hypotheses. They purport to make assertions about the world, or reality, but because there is no possible means by which we can establish their truth or falsity --- no possible observations which would allow us to distinguish worlds in which they are true from worlds in which they are false --- they actually tell us nothing whatever about the world we live in (or indeed about any world). Any observable world looks exactly the same in either case. Thus they are empty in the information-theoretic sense; they convey no information by which two possible states of affairs can be distinguished.
We can now hear the rattling of Karl Popper's bones also. Popper, of course, held that for a hypothesis or theory to be considered scientific, it must be falsifiable (in the case of theories, at least one of its key claims or predictions must be falsifiable). That is, there must be some conceivable observation which would show the hypothesis false. A hypothesis that is consistent with any observation whatsoever is useless, empty. It expresses no knowledge.
Yet, that is the precisely the situation in which Bostrom places his virtual observers --- they and their hypothetical world are crafted in such a way as to make it impossible for them to distinguish between a real world and a virtual world. They can't rule out either alternative, any more than they can rule out that their universe is doubling in size every second.
Now some may find Bostrom's scenario to be more plausible than Russell's doubling-universe conjecture. The reason it may seem more plausible is that we have a third-person perspective on the situation of the virtual-worlders. They cannot distinguish between a real world and a simulation, but, at least in their case, we can: we know their world is virtual because we have hypothesized that it is. We know something they don't: it is, after all, our computer (or that of our descendents) in which their world is being simulated.
But this move only provokes more minor earth disturbances. Now we have Kant and Schopenhauer, among others, thrashing in their graves. For we are trying to make this simulation scenario plausible via "trasnscendental knowledge": knowledge obtainable only through some supernatural means. It is knowledge which cannot be obtained through ordinary empirical methods, but only via divine revelation, cosmic insight, or something similar. As Schopenhauer put it,
Transcendental knowledge is that which, going beyond the boundary of
possible experience, endeavours to determine the nature of things as
they are in themselves; while immanent knowledge keeps itself within
the boundary of possible experience, therefore it can only apply to
phenomena.
Bostrom's scenario seems plausible to us because we have transcendental knowledge --- knowledge not available, in principle, to the inhabitants of the world about which we are hypothesizing. From their point of view, we are God, and our knowledge is divine knowledge. But hypotheses about the world --- and the the "simluation argument" clearly is such an hypothesis --- which depend upon transcendental knowledge for their plausibility are illegitimate hypotheses, regardless of who is claimed to possess such knowledge. They are barren hypotheses which cannot be tested by the inhabitants of the world to which they purportedly apply.
Someone may wish to argue that even if the simulation argument involves transcendental knowledge, that knowledge is not supernatural. The possibility that we might someday build computers powerful enough to simulate our world, or portions of it, or even to synthesize coherent novel worlds from scratch, is a reasonable projection of present technological trends, and is entirely consistent with the known, or at least accepted, laws of physics. It invokes no magical or other ad hoc mechanisms.
But it is not magic that renders transcendental knowledge objectionable. It is the inaccessibility of such knowledge to the agents to whom it purportedly applies. If transcendental knowledge is allowed, then any hypothesis whatsoever can be put forward, and cannot be ruled out by empirical means. For example, suppose an advanced race has mastered the art of terraforming, has vast energy resources at its disposal, and amuses itself by creating entire worlds with elaborate, coherent ecologies. Not wishing its own existence to be suspected by the intelligent creatures placed on some of these worlds, it seeds such new worlds with various red herrings, such as aged bones of "extinct" creatures, certain ratios of carbon isotopes and lead-argon, and so on, intended to make the world appear much older than it is. There is no need for it to create its intelligent creatures with false memories: they wake up on their first morning with no memories, but with the usual range of capacities and instincts. At first they are all very confused and frightened, but soon they get to work satisfying basic needs --- to eat, find shelter, and so on. After a generation or two, questions about their origins crystallize into myths.
After a few thousand years, their technological skills have improved dramatically --- they set about investigating their origins in a systematic way. But their creators, of course, were far more skilled than they, and have made sure no evidence of their true history is there to be found. They've covered their tracks perfectly.
How can we rule out such a history for own own world?
Obviously this scenario is a variant of one of Russell's, above, with some technological window dressing added --- plausible projections of current technologies. Does the window dressing render the hypothesis any more fruitful?
We've developed over the millenia a number of rules and precepts that have proved useful in constraining and focussing our ongoing efforts to explain and understand our experience. We've devised the rules of logic, for instance. They impose specific constraints on theorizing; they say, "Be sure your hypotheses, your descriptive propositions about the world, are not contradictory --- that is, that they do not assert a state of affairs and then deny it in the next breath. A theory that does that will tell you nothing about the world; it can give you no guidance in dealing with it."
But inconsistency is not the only way empirical hypotheses can go wrong. In fact, the only thing wrong with logical inconsistency is that inconsistent propositions convey no information. They cannot uniquely select one state of affairs from among a number of possibilities. But propositions can fail to do that for other reasons --- for example, if they are not falsifiable. A proposition which is not falsifiable, as Popper recognized, is one that is compatible with any observable state of affairs. It cannot distinguish among them. Thus it contains no information. Propositions which purport to assert two states of affairs which cannot be distinguished, per Leibniz, likewise convey no information, or rather, neither conveys more information than the other. We can't decide between them by examining the world; we can only decide between them (if at all) pragmatically: by judging which of them is the simplest, which of them conveys whatever actual information they contain in the fewest bits (per Occam).
I asked earlier, of Bostrom's two alleged scenarios, i.e., that either we live in a real world or a virtual world, why should we even think we have two scenarios, rather than just one, described in different ways. And that is the real question here. We have only one reality; the only thing in question is, what is the simplest, most fruitful, way to describe it? Both scenarios Bostrom presents describe it after a fashion, and observation of reality alone cannot distinguish between them. A hypothesis that there exists a realm of demons, each of whom is in charge of some aspect or reality, and who manipulate it pursuant to some purposes known only to them, might describe it also. Why not?
Bostrom's "simulation argument" presents itself as a hypothesis about the world. But it actually is not. It conveys no information by which, of various conceivable states possibly true of the world, one or more might be selected or ruled out. In other words, we know no more about the world after reading Bostrom's essay than before. What we do know is that virtual worlds, or worlds manipulated by demons, will require far more explanatory machinery than the real world as we ordinarily understand it, even though yielding no more information.
So we can set it aside as empty, just another of those sterile descriptive exercises in which philosophers are wont to indulge. And I can rest assured that Katy is a real cat.
-
- Posts: 199
- Joined: April 28th, 2008, 11:27 pm
Re: On the "Simulation argument"
Not entirely true. Though the word "simulation" implies a copy of something genuine, nothing prevents the computers in question from generating synthetic displays of things that never existed in their real world.
-
- Posts: 3119
- Joined: November 26th, 2011, 8:10 pm
- Favorite Philosopher: Terry Pratchett
Re: On the "Simulation argument"
-
- Posts: 2181
- Joined: January 7th, 2015, 7:09 am
Re: On the "Simulation argument"
It's true we can't know, and it makes no practical difference to us, but Bostrum also says this which is the spirit GE responded in -
The paper itself makes several acknowledged and unacknowledged assumptions and then tries to quantify probabilities, which undermines it imo, where-as other popular What Ifs which catch the zeitgeist avoid, like brains in vats, the Matrix, Last Thursdayism. There's just no way of assessing their probability in principle. You might see religions as more of the Simulation type of What If in that you can try to assess their likeliness using reason, as classical theism does.Apart form the interest this thesis may hold for those who are engaged in futuristic speculation, there are also more purely theoretical rewards. The argument provides a stimulus for formulating some methodological and metaphysical questions, and it suggests naturalistic analogies to certain traditional religious conceptions, which some may find amusing or thought-provoking.
- Pattern-chaser
- Premium Member
- Posts: 8268
- Joined: September 22nd, 2019, 5:17 am
- Favorite Philosopher: Cratylus
- Location: England
Re: On the "Simulation argument"
I know you wrote a great deal more than the above. But your reasoning matters (to your argument), and this is a misinterpretation or misuse of Occam's Razor. The Razor is appropriately deployed when we need to decide between two or more options, and have no more rigorous way of reaching a decision. This is not such a case.GE Morton wrote: ↑February 6th, 2023, 1:30 pm At that point my thoughts were interrupted by a slight trembling of the floor beneath my (simulated?) feet. A simulated mild earthquake? But then I identified the cause of the tremor -- it was a shock wave generated by the bones of the venerable Occam spinning in his grave.
This is a speculative topic. We don't need to choose between simulation and ... not-simulation. We are surely here to consider if the simulation theory is possible, and what the consequences might be if it was true.
I think we all already know that it is possible, and that we have no way to determine if it is the case or not. So the only point to this topic, that I can see, is those consequences, and the ideas that might come out of discussing them.
"Who cares, wins"
-
- Posts: 4696
- Joined: February 1st, 2017, 1:06 am
Re: On the "Simulation argument"
What consequences might those be? Or, specifically, what consequences would follow from the simulation scenario that would not follow from the non-simulation scenario?Pattern-chaser wrote: ↑February 8th, 2023, 9:44 am
I think we all already know that it is possible, and that we have no way to determine if it is the case or not. So the only point to this topic, that I can see, is those consequences, and the ideas that might come out of discussing them.
Would you defend Russell's "The universe and and everything in it is doubling in size every second" speculation on the same grounds?
- Pattern-chaser
- Premium Member
- Posts: 8268
- Joined: September 22nd, 2019, 5:17 am
- Favorite Philosopher: Cratylus
- Location: England
Re: On the "Simulation argument"
Pattern-chaser wrote: ↑February 8th, 2023, 9:44 am I think we all already know that it is possible, and that we have no way to determine if it is the case or not. So the only point to this topic, that I can see, is those consequences, and the ideas that might come out of discussing them.
Well, that's the joy of discussing speculations like this one — to see what the consequences might be, and to wonder about them. There are, after all, no definite conclusions to be reached here.
None. That's rather the point of these things. The actual nature of Objective Reality makes for interesting speculation. But the starting point is always that, whatever the starting question/theory/idea, it results in the apparent reality that we see and live with. All such ideas are identical, indistinguishable, in that one respect, if no other.
If such a theory did not explain apparent reality, then it wouldn't be a candidate for consideration, would it?
I'm not defending anything. I do acknowledge that anything possible could be true, but that's as far as I go toward a "defence".
"Who cares, wins"
-
- Premium Member
- Posts: 750
- Joined: December 11th, 2019, 9:18 am
Re: On the "Simulation argument"
Interesting topic and a nice written story!
As mentioned in the other topic Simulation Theory is an increasingly 'hot topic' in philosophy and has been worked on by David Chalmers since he began his career.
David Chalmers: From Dualism to Deism (book Reality+ about the cutting edge of VR, AI and philosophy)
A philsopher comes full circle.
viewtopic.php?f=2&t=17829
I personally would disagree with the theory but David Chalmers has written over 1,000 pages to convince readers of the plausibility of Simulation Theory.
After some re-considering I did consider the potential moral plausibility of the development of simulations because it would enable life to fulfil more within the scope of its potential of existence. This idea is not based on a perspective on its potential possibility but solely on a moral plausibility perspective.
What is the purpose of life? Enhancing potential can be considered an answer. One would need to add 'rightly so' (morality) but despite that, a simulation might provide life with an opportunity to escape the confines of time and space.
A brain in a vat type of simulation is nonsensical in my opinion because a brain is a posteriori in the face of the senses and the senses are a posteriori in the face of the potential required for sensing, which is moral valuing.
A simulation that performs morally 'as life' (if that were to be possible) might enhance life in a way that might be morally desirable. (this is just a consideration).
2023/2024 Philosophy Books of the Month
Mark Victor Hansen, Relentless: Wisdom Behind the Incomparable Chicken Soup for the Soul
by Mitzi Perdue
February 2023
Rediscovering the Wisdom of Human Nature: How Civilization Destroys Happiness
by Chet Shupe
March 2023