## Preface HANS: (*Blasts open the door, looking disheveled*) Good afternoon, Nate! I have just come to an incredibly unsettling revelation that I must tell you about at once! NATE: (*Looks at his watch*). Hans, it's 9:30 in the evening, that is hardly the afternoon. You know my mind is no good this late—my brain is in a heavy fog. Can this wait until the morning? HANS: I wish it could but it simply cannot! I must share this delicious line of reasoning I've just come to. NATE: Alright, alright. Let's hear it. HANS: Excellent, here goes. \*begins to pull an unnecessarily bulky rock out of his pocket\* NATE: (*Looks at rock with bemusement and exhaustion*). That thing must weigh 20 pounds, how long have you been lugging that around for? It just cannot be necessary to make your point. HANS: (*Ignoring the comment*)—Let me get straight to it: I believe I've constructed a water tight argument that concludes that *this rock* can be interpreted to contain a simulation of a conscious mind! Put another way, this rock *is conscious*! NATE: (*Opens his mouth to start responding*) HANS: But that is just the start of it! My argument actually shows that *anything at all can theoretically be viewed as a simulation of any possible world*. NATE: As I'm sure you expect, I'm very skeptical of this. Let's hear this argument. HANS: Alright, here it goes. First, let me make sure you are on board with a few definitions that will be required to get us off the ground. Let us define a simulation to be a set of abstract rules and entities, instantiated and executed in some physical substrate. A classic example would be that of a weather simulation—in a very real sense it is an abstract system governed by it's own internal rules, but of course it must be physically instantiated as a physical process in order to interact with it. However, the substrate itself does not matter so long as it supports computation—the simulation could be run on either the silicon chips in our laptops, a super computer, a [Human Computer](Human%20Computer.md), or [a set of spring loaded dominos](The%20Domino%20That%20Didn't%20Fall.md). Would you agree that this is a fair definition of simulation and that I am clearly not up to any funny business so far? NATE: I would. HANS: Very good. Now, this abstract set of rules that define the simulation, as we just agreed they must be instantiated in some physical substrate. Let us call that process *encoding*. An example should help here. Consider the weather: it is governed by a set of abstract rules—we may call them it's "laws of physics". To create a simulation of the weather, we encode these rules on some computer and then execute the simulation. This involved taking the set of rules and programming them in some language—say Python. From there, the Python code descends a ladder of execution: it is compiled to bytecode, interpreted by the Python runtime (itself compiled to machine code), and ultimately executed by the CPU according to its instruction set architecture. At the lowest level, this reduces to electrons moving through the semiconductor materials of the CPU’s transistors. Each of these levels represents the same abstract process—it implements the abstract rules of our weather system. In order to interact with anyone of these representations you must decode it into a form that is interpretable. Some of these representations are easier to interpret than others. For example, you may be able to read (a form of decoding) the python code and have a reasonable idea of what is going on. Or you may write a specific interpretation program that takes the state of the weather system and renders it via some geo visualization software. But if you were to look through an electron microscope at the silicon chip itself during the simulation, you would most certainly not have any clue what is going on. It could be simulating anything at all as far as you can tell—or it could be stuck in an infinite while loop generating random numbers and creating heat. Thus any abstract system that we wish to simulate always brings along with it two constraints: we must be able to encode it into a physical system, and we must be able to decode it into a form we can understand. Do you have an quibbles with this? NATE: No I am with you so far, please continue. HANS: Excellent. The final definition is that a conscious mind is comprised of a set of rules and relationships—a program—that is run in a physical substrate. Currently the only substrate where we have instantiated consciousness inside of brains, but this need not be the case—any physical substrate that supports computation could in principle do the trick. Thus, consciousness could be simulated. NATE: (*Eyes discretely dart to the clock while nodding in agreement*) HANS: Excellent. Now that we agree about just what a simulation is, let me explain the [Problem](Problem.md) that this created for me. You see, earlier today as I was walking home from the lab I recalled an idea proposed by Eric Drexler in one of his books[^1]. He had a clever design for a computing system the size of a sugar cube that could perform roughly $10^{21}$ operations per second. Given the small physical size of this machine, I started to wonder: could certain every day objects be performing computations—even full blown simulations—all while we go about our lives none the wiser? Could this rock in my hand contain a simulation of a conscious mind? NATE: I suppose it *could*. And by that I mean that it is *possible*— HANS: (*has stopped listening at this point due to his excitement*) NATE: But what really matters is whether or not we have a good *explanation* as to what the rock contains. HANS: Great, so we both agree that it is possible—that we can't logically rule it out. The question we must answer is how can we identify if a simulation has been *encoded* in a physical substrate. For instance, we are looking at this rock and want to determine if a physical simulation of a mind has been encoded within it. And here is my clever maneuver—it is actually more fruitful to turn the question around. Rather than asking how or why a simulation has been encoded in the rock, we can say that if we could *decode* a simulation from the rock, we know that one had been encoded. NATE: And how will you go about this decoding process? Generally when we encode a simulation in a physical substrate we do so via a specific set of rules to ensure that we accurately capture the abstract system of interest, and we then decode it via a specific set of rules that are tied to the simulation. These rules are not arbitrary—they are quite hard to vary and serve a very specific purpose. HANS: Well that is the thing. If we reason from first principles, surely you can agree that the safest choice here is to accept all mathematically possible decodings. NATE: Let's slow down for a moment, why would we want— HANS: Before you criticize that idea, will you grant me the following: that you could be looking some physical system and not know that it is encoding a simulation? For instance, imagine looking at the physical transistors of a silicon chip as it runs a simulation of a chess match. You *see* the physical system, but to you it is entirely opaque whether or not it is performing any meaningful computation. It could be running a simulation of anything for all you know, or just performing a bunch of random calculations and generating heat. You—from your vantage point outside the system—cannot tell. It is only if you have the correct *decoding mechanism*—a sequence of programs that take this cpu state, and decode it back to some external viewing program where it is presented as pixels on a screen—that you would know it running a chess match. NATE: I will grant you that. HANS: Well in that case you have effectively just conceded to my argument! You may look at a physical system today and—given your knowledge, current decoding programs and so on—interpret it to be doing nothing of interest. But tomorrow, perhaps armed with a new decoding scheme, interpret that it is actually simulating the effects of different heat levels on the wing of an airplane. Thus, as Karl Popper—yes I know how much you admire him—would say: all observations are theory laden. As our knowledge grows, our decoding schemes and interpretations will as well. If our criterion for reasonable decoding schemes is based on our present knowledge, we will certainly miss out on some. This will lead us to incorrectly classify some physical systems as *not* containing a simulation when they in fact do. It is a way to guarantee we end up with false negatives. NATE: Why would— HANS: I know exactly what you are going to retort next: "Why would having some false negatives be such a bad thing?" Because we would be doing so all the while *knowing* that we were misclassifying certain systems. That is irrational. Besides, I am simply reasoning logically, from first principles. This is done all the time—consider when a physicist deconstructs some problem into it's absolute core elements and relationships, discarding all unnecessary and parochial cruft. That is all I am doing. I am getting to the core of the issue and removing any unnecessary constraints—such as our knowledge of decodings schemes *today*—and only keeping the fundamentally critical constraints, such as a decoding scheme must be mathematically possible. Would you really argue against reasoning from first principles? NATE: No, most certainly not. But I don't think that is what you are doing. It is rather slippery and I can't quite pin down why, but you are doing something different. It looks like first principles reasoning in *form*, but there is something about it that is distinctly different. HANS: Well I know we can both agree that that is not an effective criticism. Shall I move on and discuss the consequences of this argument? NATE: (*Unsettled and very much awake*)—well I know you won't leave until you shared it, so carry on. HANS: Very well. Once we have taken aboard the previous components of my argument, the consequences are quite dramatic. Given that we have granted that all mathematically possible decoding schemes are accepted, one possible decoding scheme is that of using a *lookup table*—a mathematical object—to map internal rock states to states of a conscious mind! This is a decoding scheme by our criterion— NATE: Let's not go ahead calling it "our" criterion... HANS: —and thus this is a legal move. Although the word "move" implies that I am *making* this choice. But as you can certainly see, I am not choosing this move, it is simply an implication of the argument as it has been constructed. Thus based on this decoding scheme it quickly starts to become clear that *any physical process* can be viewed as running *any simulation*. HANS: And if you will let me put a final bow on this argument, we can also see that this argument, growing out of the premises and techniques of physical science, has the unexpected consequence of demoting physical existence to a derivative role. A possible world is as real, and only as real, as conscious observers, especially inside the world, think it is! HANS: (*stares at Nate with anticipation*) And that is the full argument! I must know, what are your thoughts? It is quite compelling wouldn't you say? NATE: I'll grant you that as you presented it, it is awkward to argue against. But I feel like since you have arrived my brain has descended into an ever thicker fog. I keep reminding myself that if you grant an absurdity you can derive nearly any consequence you like. But I can't put my finger on the absurdity—I feel that it is hidden rather well. HANS: I must say I feel slightly hurt that you think I would purposefully *hide* an absurdity. As I have already said, I am just reasoning from first principles—I have laid the entire argument out in plain view for you to see and criticize. It seems unfair that you would accuse me of arguing in bad faith. NATE: I should be more clear. I don't think that you are doing this intentionally. Rather I think you are a victim of poor reasoning. But I can hardly blame you—at this hour I can't even spot where you went wrong. And I know, I know, that is not a criticism. But I am confident there is an misstep *somewhere*. Your argument has an attractive form—it moves from premise to conclusion, mimicking first principles thinking the entire time. Thus I think there is something deeper going on here, something that I must reason through with a fresh mind. How about this, give me a few days to think this through—a good night sleep, fresh coffee and several walks should clear my mind. After this why don't you come back over and I can give a more effective criticism to your argument. HANS: Very well, I look forward to it! [^1]: The book is *Nanosystems: Molecular Machinery, Manufacturing, and Computation*