Reality+ : Virtual Worlds and the Problems of Philosophy (9780393635812), page 40
Much more realistic are cases in which people move knowingly between the nonvirtual and virtual worlds. This is already happening every day to users of video games and other virtual environments. We enter the world of Grand Theft Auto and we talk of stealing cars. However, the virtual world of Grand Theft Auto doesn’t contain any real cars. It contains only virtual cars. When we say, “There’s a car over there” while we’re in the game, are we saying something false?
My view is that language is a malleable instrument that bends to our purposes. If we want to extend use of the word “car” so it applies to virtual cars, we can do that. Philosophers and linguists have long recognized that what a word refers to can be context-dependent. “Tall” means one thing when we’re talking about basketball players (a six-footer isn’t tall) and another thing when talking about academics (a six-foot philosopher is tall). Words adapt to the role that’s useful in their context.
With the onset of virtual worlds, much of our ordinary language has become context-dependent in this way. When I say, “There’s a car” in an ordinary, nonvirtual context, I mean it’s an ordinary, nonvirtual car. But when I say, “There’s a car” in a virtual context, I use “car” in a way that includes virtual cars.
It’s also possible that the meaning of “car” could shift so that it covers both nonvirtual cars and virtual cars by default. In chapter 10, I said that a category X is virtual-inclusive when a virtual X is a real X and virtual-exclusive otherwise. On current usage, car and hurricane are virtual-exclusive (virtual cars don’t count as real cars), while computer and communicate are virtual-inclusive (virtual computers are real computers). But as we saw then, words can evolve to become more inclusive. It’s entirely possible that “car” could evolve to become virtual-inclusive by default. In that case, we may mean the same thing by “car” in both virtual and nonvirtual contexts.
The idea of virtual-inclusiveness can help us to analyze externalist cases. It’s arguable that virtual-exclusive words such as “hurricane” will refer to one thing (hurricanes) on Earth and to another thing (virtual hurricanes) on Sim Earth. That is, they’ll be anchored to their environment, in classical externalist style. On the other hand, a virtual-inclusive word such as “computer” won’t be anchored to its environment. It will refer to computers on both Earth and Sim Earth: virtual computers are computers too.
Over time, as virtual reality becomes more and more central in our lives, it’s natural to expect that many words will gradually shift from being virtual-exclusive to being virtual-inclusive. As this happens, our usage of language may come to put less emphasis on what things are made of and on how they are anchored in our environment. Instead, a virtual-inclusive language may put more emphasis on factors that are common between virtual and nonvirtual reality: the structured patterns of interaction among things, and how they connect to minds like ours.
Putnam on externalism and Cartesian skepticism
In his 1981 book Reason, Truth and History, Hilary Putnam used externalism and the causal theory of reference to analyze the brain-in-a-vat scenario. He didn’t say anything explicitly about skepticism, but it’s easy to draw conclusions about skepticism from his discussion. His conclusions are not the same as mine, but they are interestingly connected.
We’ve already encountered Putnam’s main thesis in chapter 4. He argued that the lifelong brain-in-a-vat hypothesis is incoherent, or contradictory. In effect, he thought he could use externalism to prove that we’re not brains in vats, or at least that we’re not lifelong brains in vats (I’ll take “lifelong” for granted in what follows). He did not explicitly discuss the simulation hypothesis, but he almost certainly held that this is contradictory as well.
Here’s how Putnam’s argument goes. Take a situation like the one depicted in figure 50, in which it looks like Putnam is a brain in a vat. However, Putnam insists, “I am not a brain in a vat.” He reasons as follows. Externalism tells us that if he’s in the situation depicted, then when he says “brain,” he’s referring to a virtual brain, a brain-like object made of bits, like the virtual brain he’s pointing to inside the simulation. So when he says, “I’m not a brain in a vat,” he means I’m not a virtual brain in a vat. And that’s true! In the scenario as depicted, he’s not a virtual brain in a vat. He’s not like the object next to him that he calls a “brain.” Instead, he’s a nonvirtual brain in a vat, in an entirely different world that’s not made of bits. So when Putnam, inside a simulation, says, “I’m not a brain in a vat,” he speaks the truth. Generalizing, Putnam argues that whether or not he’s in a simulation, he will always be correct in saying, “I’m not a brain in a vat.” Each of us can reason this way for ourselves, thereby proving that we are not brains in vats!
Figure 50 Hilary Putnam as a brain in a vat. Is he right when he says “I’m not a brain in a vat”?
Let me run through the reasoning for myself. If I’m a brain in a vat, externalism tells me that what I call “brains” are virtual brains—simulated entities that are part of my own simulated environment. But I can’t possibly be a virtual brain in my own simulated environment. It’s possible that I’m the sort of thing that the simulators call a brain in a vat—but that’s a different sort of thing entirely. I can know I’m not the sort of thing that I call a brain in a vat. That is to say: I can know I’m not a brain in a vat.
If Putnam is right, the very idea that I’m a brain in a vat is subtly contradictory. To be a brain in a vat, I’d have to be the sort of thing that I call a brain in a vat. But if I’m a brain in a vat, what I call a brain is something in my environment entirely different from the sort of thing I am. So to be a brain in a vat, I’d have to be something I’m not.
There’s a lot to say in response to Putnam. One loophole that Putnam touches on himself is that the argument can’t rule out the possibility that he is a simulated brain in a vat, like the second brain depicted in figure 50, experiencing a second-level simulation as depicted on the second screen. If he’s in that situation, what he calls a brain is a virtual brain (like the third brain depicted), and he is in fact a virtual brain (like the second brain depicted), albeit in the next world up. In this scenario, it’s arguable that he is what he calls a “brain in a vat” after all. If so, Putnam’s argument shows, at best, that he is not an unsimulated brain in a vat: he’s not a level-one brain in a vat, but he could still be a level-two brain in a vat. That’s not enough to defeat Cartesian skepticism; he still needs some separate way to deal with the skeptical hypothesis that he’s a brain in a vat at level two.
Furthermore, Putnam’s argument doesn’t work nearly as well for the simulation hypothesis as for the brain-in-a-vat hypothesis. The reason is that, as we saw earlier, externalism works better for some terms than for others. Words like “New York,” “water,” and “brain” arguably have their meaning tied to our environment, so they can’t refer to something in the next universe up. But there are other words whose meaning doesn’t depend on the environment in the same way. Examples include “zero,” “person,” “action,” “computer,” and “simulation.” A computer is defined in largely structural terms, independent of any particular environment. As a result, even if I’m in a simulation, there’s no problem talking about persons, actions, computers, and simulations in the next universe up.
If this is right, then Putnam’s reasoning can’t rule out the hypothesis that I’m in a computer simulation. Even if I’m in a computer simulation, I can truly say, “I’m in a computer simulation.” As I said in chapter 4, if Sim Putnam says, “I’m in a computer simulation,” then what he says is true.
Perhaps Putnam could say that the word “simulation” is like “New York” or “water” and is anchored to a specific system existing only in our ordinary environment and others just like it. But this seems wrong: When we talk about computers and simulations, we’re talking about something much more general than that. I can straightforwardly speculate that I’m in a computer simulation in the next world up, and I might be right.
For this reason, I don’t think Putnam’s main argument can work as a general reply to global skepticism. However, Putnam briefly makes a quite separate argument that is also relevant to global skepticism and much closer to my own approach. In one paragraph in Reason, Truth and History, Putnam suggests that brains in vats have mostly true beliefs. He argues for this as follows:
By what was just said, when the brain in a vat (in the world where every sentient being is and always was a brain in a vat) thinks, “There is a tree in front of me,” his thought does not refer to actual trees. On some theories that we shall discuss it might refer to trees in the image, or to the electronic impulses that cause tree experiences, or to the features of the program that are responsible for those electronic impulses. These theories are not ruled out by what was just said, for there is a close causal connection between the use of the word “tree” in vat-English and the presence of trees in the image, the presence of electronic impulses of a certain kind, and the presence of certain features in the machine’s program. On these theories the brain is right, not wrong in thinking “There is a tree in front of me.”
The basic argument is an appeal to the causal theory of reference. When a brain in a vat says, “There’s a tree,” its use of “tree” is caused by virtual trees. So “tree,” for a brain in a vat, refers to virtual trees. And when the brain says, “There is a tree,” there really is a virtual tree. So what the brain in a vat says is true! Something similar goes for other beliefs about the world.
More or less the same line of argument was explored briefly by the American philosophers Donald Davidson and Richard Rorty. It was summarized crisply by Rorty as follows:
That brain too is reacting to features of its environment. But its environment is the computer’s data bank. The only way you can translate the noises it makes is to correlate them with the bits of data that the computer is feeding in. So the noises that sounds like “It’s Tuesday the 7th of October 2003, and I am eating tofu” must mean something like “Now I am hooked up to sector 43762 of the hard drive.” For most of the envatted brain’s beliefs, like most of ours, must be true. It is not as easy to delude a brain as the evil scientist thinks.
I think that Putnam, Davidson, and Rorty are all essentially correct. A brain in a vat refers to entities in its environment, and as a result, its beliefs are mainly true.
Still, I don’t think that any of them have given a strong argument for this claim. Their argument rests on an overly strong and implausible form of externalism. As it stands, the argument seems to assume an extreme version of the causal theory of reference, in which every word refers to an item in its environment that causes it. But this extreme version is false. Many words, like, say, “witch” or “ether,” refer to nothing at all. The women to whom the word “witch” was first applied were not witches. The ubiquitous ether of 19th-century science does not exist. A Cartesian may well say that “brain” and “tree” for a brain in a vat are like “witch”: they refer to nothing. Putnam has given no real argument against this Cartesian view.
Furthermore, we have seen that while externalism may work for words like “tree” and “brain,” it doesn’t work so well for words like “three,” “computer,” and “philosopher.” The externalist analysis doesn’t really apply to sentences like “There are three philosophers over there,” or “I am using a computer.” Nevertheless, simulation realism requires that sentences like this can be true inside a simulation. So Putnam’s externalism hasn’t shown us how simulation realism can be true across the board.
I think it is possible to rebut these criticisms. I’ve already made a case, in chapter 9, that “brain” and “tree” in a simulation refer to virtual brains and virtual trees. Roughly, what matters is that brains on Earth and virtual brains on Sim Earth play a similar structural role. The argument in chapter 9 also suggests that “There are three philosophers over there,” and “I am using a computer,” uttered inside a simulation, are true. I’d diagnose this by saying that these beliefs largely concern structural matters that are shared between Earth and Sim Earth, so that they can be true on Sim Earth even with their ordinary meaning. Even if I am in a simulation, I may still see three philosophers or use a computer. However, the argument in chapter 9 and the analyses here rest not on externalism but on structuralism.
Ultimately, it’s structuralism, rather than externalism, that drives the simulation realism that drives my response to skepticism in turn. The chapters that follow will develop the case for structuralism.
Chapter 21
Do dust clouds run computer programs?
IN THE CLASSIC 1994 SCIENCE-FICTION NOVEL PERMUTATION City, by the Australian author Greg Egan, simulations are everywhere. People make simulated copies of themselves, which inhabit virtual worlds where they have conscious experiences much like the originals. The main character, Paul Durham, is a copy with low status and little legal recognition who experiments with creating simulated worlds of his own.
In the novel, it turns out that an ordinary full-scale simulation is not required to create a world. Durham modifies the simulation he inhabits so that its parts are wholly disconnected in space and time. Even so, he continues to exist. He breaks the simulation of himself down into ever-smaller disconnected pieces and he still exists. Even when the pieces are completely scattered in space and time with no connection between any of them, he and his world still exist.
As Durham scatters pieces of himself, he forms the idea that the universe itself fundamentally consists of scattered pieces of unorganized dust:
Squeak. “Trial number four. Model partitioned into fifty sections and twenty time sets; sections and states randomly allocated to one thousand clusters.” “One. Two. Three.”
Paul stopped counting, stretched his arms wide, stood up slowly. He wheeled around once, to examine the room, checking that it was still intact, still complete. Then he whispered, “This is dust. All dust. This room, this moment, is scattered across the planet, scattered across five hundred seconds or more, but it still holds itself together. Don’t you see what that means? . . .
“Imagine . . . a universe entirely without structure, without shape, without connections. A cloud of microscopic events, like fragments of space-time . . . except that there is no space or time. What characterizes one point in space, for one instant? Just the values of the fundamental particle fields, just a handful of numbers. Now, take away all notions of position, arrangement, order, and what’s left? A cloud of random numbers.
“That’s it. That’s all there is. The cosmos has no shape at all—no such thing as time or distance, no physical laws, no cause and effect.”
This theory is the dust theory. It postulates a large cloud of randomly scattered atoms of dust, outside space and time and with no cause and effect. The core idea is that a scattered dust cloud like this could execute any possible algorithm and therefore simulate every possible world, resulting in the existence of vast numbers of conscious people. An even more speculative version of the theory says that a dust cloud like this is what underlies our own reality.
Figure 51 The dust theory: A randomly scattered cloud of dust underlies computation, which underlies reality.
The dust theory is fascinating. It would have all sorts of consequences if true. Even if there’s no such dust cloud, our world contains a whole lot of matter. If dust can run every algorithm, then so can this matter. If so, then pretty much every possible computer program is running somewhere in the world and every possible simulated world exists. Every possible simulated person exists, too. That’s a vertiginous picture. If simulating a world or a person is as easy as this, the whole idea of simulation may be trivialized.
The dust theory raises any number of questions. One question, pertaining to the plot of Permutation City, is why anyone in the novel bothers to create full-scale simulations. This would seem pointless if every simulation is already running in the dust. Another question concerns Durham’s partition experiment. Can these simulations be truly scattered and unconnected, given that the program he’s using always has the ability to reunite the pieces? If they aren’t truly scattered, Durham’s inference that scattered dust can run simulations seems hasty. A third question is how science can work if every possible world exists. Most worlds will be chaotic and unpredictable. It would be surprising to find ourselves experiencing a highly ordered world as our world seems to be.
There’s a deeper problem with the dust theory. It rests on a false assumption. The assumption is that patterns of cause and effect are irrelevant to executing an algorithm, to generating reality, and to generating consciousness. In fact, complex structures of cause and effect are crucial to all of these things. These structures aren’t present in Egan’s dust cloud. As a result, dust does not support genuine algorithms, simulated worlds, or simulated people.
This point is key to my argument for simulation realism. Computer simulations are not just clouds of aimless dust. They’re fine-tuned physical systems whose elements interact in complex patterns of cause and effect. This causal structure is what makes them genuine realities, on a par with the nonsimulated world.
Clarifying this argument requires us to examine the relationship between algorithms and physical systems.
Computation in physical systems
What’s the relationship between computer programs and physical systems? There’s a vast mathematical theory of computation. The theory postulates abstract systems such as Turing machines, finite automata, cellular automata (like the Game of Life), and algorithms of all sorts. It tells us what sort of problems the various computing systems can solve and how they go about solving them.
