Coining the Faceless Wind
Long ago in a philosophy class my teacher touched upon the well-known thought experiment called “the brain in a vat,” in which an imaginary subject’s brain is placed into a tank of something approximating cerebrospinal fluid and hooked up to a supercomputer that feeds it artificial stimuli that is comparable to kind the “real world” would provide. At its most basic level, the experiment brings into question what is “real” or “true” since the mind (we assume the brain is the mind, here) in the vat, by definition, is unable to determine if it is in the “real” world, or merely a brain in a vat. These kinds of theories were popularized through cyberpunk fiction and movies like Ghost in the Shell and The Matrix, which in turn affected the way we think about computers and the Internet.
Though I am not really equipped here to discuss the real implications of the possibility of a brain in a vat, I thought that another interesting area of inquiry might be how some (evil?) demiurge might construct such a mechanism using what we currently know about real-time virtual reality– or, in other words, video games.
As I see it, there are two main approaches to achieving the brain in a vat scenario.
The first is summed up by The Matrix, wherein all of humanity are brains (plus their bodies) in vats, and interact in a kind of shared hallucination. The video game equivalent of this is probably a massively multiplayer online game, like World of Warcraft, or a more freeform online world such as Second Life, where each avatar is piloted by a human brain through the world of the game and interacts with other brains in their own vats through the various methods that the game provides.
It doesn’t take long for the technical issues of this approach to become manifest. Every brain would need its own powerful client machine to “render” and feed the brain stimuli; the client machine would also need to connect to a central server in order to reconcile each client’s version of “reality.” Coordinating the actions of many brains in vats together– people jostling each other in a crowd, for example, or playing in a band, or multiple cooks in a kitchen– would tax a system of any known design enormously. Complex physical simulations would need to be carried out either by the central server and propagated instantaneously to all clients or calculated completely deterministically on each client simultaneously. Such coordination is extremely difficult to achieve even with the relatively simple information that networked games must share today. (The only way I can think of that this might be possible is if the system does not actually run in real-time, and that each “frame” of reality we experience is actually the result of days or weeks or centuries of calculation on the part of the demiurge’s computer system. But one imagines him getting as impatient as we do with poor frame rates.)
There are other problems with a Matrix-like scenario, too: how would such a system handle human births and deaths? Is our pet dog or cat a smaller brain in a smaller vat somewhere? The line between what is simulated versus what is “real” blurs in many ways in this arrangement, making a consistent illusion of reality difficult to define, let alone manage.
The second major brain in a vat scenario is the solipsist one: you are the only brain in a vat, and the vast world you perceive is generated solely for your benefit. Serious philosophers usually recoil from solipsism, and for good reasons. One is that there’s no place good to take it (the world is false– okay, so now what?). Another is that it doesn’t feel right morally; we must accept the existence of other beings in order to behave well. Believing that others do not really exist leads one down the road to psychopathy.
The games that best encapsulate this form of the brain in a vat are in my opinion Bethesda Softworks’ recent role-playing games, The Elder Scrolls IV: Oblivion and Fallout 3. In these games the entire world is created solely for the entertainment of the singular player, who is the only “character” possessed of agency in their vast worlds.
As a game developer, the solipsist model seems much easier to manage. After all, one of the most important things a game does is decide what not to do: the game’s renderer decides what not to draw, the sound system decides what doesn’t need to play, the artificial intelligence decides what enemies do not need to think. The complex graphics, soundscapes and interactions we enjoy are due to these judgements– if our consoles actually tried to process in full what was going on around us they would chug to a halt.
We have all heard the koan about the tree falling in the forest and its sound, or lack thereof, in the absence of observers. In terms of a video game, the answer is obvious that if there is no observer present there is no reason to calculate the observable property (armchair quantum mechanics enthusiasts may bring up the Heisenberg uncertainly principle here). In terms of technology, a singular brain in a vat could be much more easily convinced it is in a “real” place, especially if it has never known any other world.
Between these two extremes there are some compromises available, such as the idea that perhaps only two or four or six brains exist in vats– though situations like that seem more the province of science-fiction plotting than a serious possibility.
None of this idle speculation is to suggest I believe the world is indeed simulated, so I’ll close with a paraphrase from Borges: the world, unfortunately, is real; I, unfortunately, am Matthew.
Commentary | January 20, 2010
Reader Comments (5)
I mean why not just posit instead that we could all be A.I. characters in a ridiculously advanced Sims game? That actually closes up even more loopholes. For example the whole "glitch in the Matrix" factor could be eliminated since our senses, thought processes and memories could be edited. Not to mention the fact that being an inherent part of the simulation we would have no clue how low-fi it would be compared to the external world.
Why yes, I do still consider the original Matrix film to be worthy of examination and discussion. Thanks for asking.
But really, I'm with Seth: the idea of a mind in a "real" brain receiving stimuli that doesn't correspond to a "real" referent is the product of Descartes' writings, and he had manifestly no understanding of the mechanism of consciousness. All those enlightenment guys read their Plato, then their Aristotle, then tried to pull themselves up by their physics-and-biology-haven't-fully-been-invented-yet bootstraps. I'm reminded of Leibniz, who argued that the soul must exist, because if there was a machine that could see (much less think), you could not identify any particular part of it that did the seeing (so much more so the thinking). Folks back then just believed that thinking was a single-threaded process that could be described in precise narratives in the first person masculine singular, "Man." We moderns know that there's no rule that says a computer can't play the game, so to speak.
Your question reminds me of Eliezer Yudkowsky's passionate argument for taking steps to ensure that the Singularity be a moral one, for otherwise people might spawn immense simulations just to model realistic people that they can then torture for any amount of time. We need to outlaw that, he says, because those simulations would also be people, despite not being able to vote or directly influence our physical reality (which conditions I think mean that those people are quite screwed).
You know, many Japanese videogames (and other media) over the years have featured the creators of a simulated universe as their Final Bosses. These frequently beg the question of why these beings would bother making the world in the first place, or why they would make it as it is. But from a game design perspective, shouldn't those goals come first? A perfectly designed game of this sort would change its "players" into exactly what the designer wanted. To what end? What does the demiurge desire? In reality, I'd expect a purpose like "to express power through cruelty," but "I Have No Mouth But I Must Scream" is decades old, so I hope we can come up with some new ideas.
And keeping time with a band composed of remote clients in a game is similar to keeping time with a band playing at the other end of a football field. So I'm not sure that the appearance of local synchronization in our reality proves that it couldn't be manufactured by some algorithm. You could even argue that whoever designed our universe put those speed limits in precisely because they helped to mitigate the overall system load; and that uncertainty in quantum physics is a hack to prevent endless recursion (wave collapse as a kind of try/catch statement). One thing is certain; in a simulated 3D universe, for every voxel faster per tick your light moves in the sim, you'd have to pass a spherically increasing volume of data out of the client (relative observer) and into the universe. Come to think of it, you might be able to make a case for general relativity being proof of a bandwidth limitation on the network running this whole thing.
Setting all the solipsistic and brain-tank implications aside though, what about other game paradigms? What about Spore, where the AIs generated for one player's world are created by another brain in a tank and then sent on their merry way? They may be complicated clocks by the time you see them, but they're still a unique product of another player's actions, not just scenery from the original world design. Also, when you consider how much real-world interaction, cause and effect is essentially "turn-based" as opposed to synchronous, I think any designer worth his salt would try to leverage those interactions, say, by writing two separate engines, or at least an event structure that prioritized certain actions over others.
Anyway, great post, and excellent food for thought.
The brain-in-a-vat idea has always been a fun one to think of. Unfortunately, the explicit references to brains and vats tend to confuse thinking on the subject.
Given a computer system sophisticated enough to convincingly feed a human brain enough logical stimuli to prevent it from ever realizing it wasn't actually living in a real world, the flesh-brain itself is likely to be the weakest part of the whole system. The world, as it has proven to me many times, is more complex than my mind.
I think a much better way to think of it is as another commenter suggested above. What if we're all just on silicon? Alistair Reynolds' and Charles Stross' books explore the idea of mind uploading into purely simulated worlds in their books. It's a Singularity thing, yeah.
So here's my point. Whether the computing power of the whole system (mind+world sim) is shared between a physical brain in a vat and a silicon computer or whether it all runs on silicon (or whatever advanced computer substrate) is irrelevant. All that matters is the nature of the computations going on, not the chemical or physical pathways that are supporting them.
One brain+computer/Multiple brains+computer/All silicon are all ultimately just different physical forms of computer running the same program.
There's lots more fun stuff to argue about this, but I'll stop here.