Reality+: Virtual Worlds and the Problems of Philosophy

by David J. Chalmers

Third: the simulator will usually know a great deal about the simulation. Again, some video games and the like may conceal the full state of a world from her. But good universe-simulation software will include devices that she can use to k...

And again, simulators with access to the data structures could use these to monitor any entities in the world.

Fourth: will the simulator be all-good? There’s not much reason to think so.

All sorts of beings may have access to simulation software, and sterling character isn’t typically a requirement. Some simulators may well be like my nephew, whose attitude toward his subjects was far from benign. Some may be like Rick, exploiting the subjects for selfish ends. Some simulators may want to see their subjects thrive, but simulators like that may be in a minority. Many may simply be looking for entertainment or for information. To a first approximation, our simulator comes close to satisfying three of the four criteria. Sh...

The godliness of simulators will also be limited in another way. The simulator might be the creator of our universe—but not of the entire cosmos.

The simulator is perhaps closest to what Plato called a demiurge. In ancient Greece, a demiurge (or dēmiourgos) was an artisan or craftsman. In Plato’s dialogue the Timaeus, he describes the demiurge as a divine being who “fashioned and shaped” the material world. The demiurge is often treated in the Platonic tradition as a sort of second god, with the one true cosmic god above it. Plato’s demiurge was benevolent, but later, in the Gnostic tradition, demiurges were regarded as evil. The simulator can likewise be treated as a second deity, perhaps benevolent and perhaps not, who is responsible for the creation of our world.

Do we really want to found a religion around our simulator?

The classic ontological argument for the existence of God was put forward by Saint Anselm of Canterbury, a Benedictine monk, in the 11th century. It goes something like this: God is by definition an absolutely perfect being. We can’t conceive of any greater being. God has all the perfections—knowledge, goodness, power, and so on. But existence is a kind of perfection, too! An existing God is clearly greater than a nonexistent God. So if God doesn’t exist, he is imperfect. Since God is perfect by definition, God must exist!

In any case, the god of the ontological argument doesn’t resemble the god of the simulation. As we’ve already seen, the god of a simulation may be imperfect in many ways.

Another classic is the cosmological argument for the existence of God. Versions of this argument are found in many philosophical traditions, but it is associated especially with medieval Islamic philosophers such as al-Ghazali. It goes something like this: Everything has a cause. Therefore, the universe has a cause. That cause must be God.

What causes God? If nothing causes God, then the premise that everything has a cause is false. But if something causes God, then God isn’t the ultimate cause after all. To get around this, al-Ghazali restricted the premise to be “Everything that begins to exist has a cause,”

But now we can ask, what if the universe has existed forever? If an eternal God doesn’t need a cause, nei...

If we’re in a simulation, the simulator does some of the work that the cosmological argument asks for. She serves as a cause for our universe.

The simulator certainly began to exist, for example. This raises the obvious objection: Who or what causes the simulator to exist? We’ll get a chain of causes that eventually goes back to the cosmos as a whole and its creation. Some will stop the chain at the cosmos. Some will bring in a cosmic god. Some will ask for a cause for the cosmic god. It’s not obvious why bringing in God and then stopping makes more sense than any other stopping point.

Another influential argument for the existence of God is the argument from design. Our universe exhibits impressive design. Humans and other animals are amazingly well-functioning mechanisms. Nature is extraordinary in its complexity. Noth...

A more sophisticated version of the argument from design still exists, though: the argument from fine-tuning. This argument concerns the physical laws of our universe—laws concerning gravity, quantum mechanics, and so on. If those laws had been just a little different, then our universe would have been much less interesting and life would never have evolved. By some reasonable measures, most ways of setting up the physical laws of a universe wouldn’t lead to life, but the laws in our universe did. So this universe seems to be fine-tuned for life.

This fine-tuning requires an explanation. The obvious explanation is a fine-tuner—namely, God.

The god of the fine-tuning argument is quite consonant with the g...

It seems plausible that people who run simulated universes will often be more interested in universes that contain life than in universes that don’t. Indeed, simulators may often simulate universes in order to simulate life, as when they simulate episodes in the history of their own species. In this case, simulators will quickly abandon settings that don’t lead to life and will focus on those settings that lead to interesting forms of life. In this way, the preferences...

The fine-tuning argument is controversial. Some people deny that the necessary conditions for supporting life are all that special. Others think we might just have been lucky. Still others say that if the universe hadn’t supported life, there would never be anyone there to notice. Given that we notice, it’s no surprise that ou...

Anthropic reasoning works best if there’s a multiverse, a cosmos made up of many universes. Perhaps some spring from black holes in other universes, for example, or perhaps there is a sequence of universes arising after successive Big Bangs. Some cosmologists hold that the laws of the universe are partly determined by what happens just after the Big Bang, so the laws may vary from one universe to another within the multiverse. If there are enough universes with varying laws, it’s highly...

The multiverse solution to the fine-tuning problem is usually put forward as an alternative to the God solution. But the simulation argume...

Suppose we have simulators who are utterly uninterested in life or observers. They’re interested only in charting the physical dynamics of many different universes with different laws. In that case, they’ll create many universes, and if they creat...

the solution to the fine-tuning problem doesn’t lie in the god’s designing our world but in the existence of a multiverse. If design plays any role, it’s the overall design of the multiverse that counts.

This reflects a general weakness in the multiverse solution to the fine-tuning problem: What explains the fine-tuning of the multiverse? The multiverse itself is presumably the consequence of some underlying laws or principles. If those laws had been different, there might have been just one universe, not a multiverse. So we need to explain this fine-tuning too. It won’t help to postulate a higher-level multiverse in which our multiverse is contained, since this just iterates the problem. So we need some other solution here. Perhaps fine-tuning for a multiverse was unsurprising, or perhaps it was lucky. Or perhaps it involved a designer.

On the other hand, the simulation analogy also brings out a weakness in any argument from design. The designer itself must presumably be an impressive creature—one that indicates design. Surely that design itself requires explanation. Another designer just pushes the problem back: What explains the design of that designer, or of the whole system of designers? Someone might say that God is exempt from explanation, but this looks l...

It’s easy to find problems in arguments for the existence of God. The fine-tuning argument is perhaps the strongest of them, but it seems far from definitive.

Might the simulation argument then be the most powerful argument for the existence of God? Of course, it’s mainly an argument for a creator and not for a being who is all-powerful, all-knowing, and all-good. But the same is true of the cosmological and design arguments, and, unlike those, the simulation argument suggests a god with considerable power and knowledge beyond the mere power of creation.

If the simulation argument is even approximately as good as the design argument, it deserves to be in the pantheon of arguments for God’s existence.

A version of the simulation argument could have been made long before computers existed. Where the argument talks about simulation, one could talk about universe creation instead. Here’s a cousin of the initial simulation argument in chapter 5: A few top-level populations will each create many populations. If a few top-level populations each create many populations, then most intelligent beings are created. If most intelligent beings are created, we are probably created. ______________________ So: We are probably created.

Here a “top-level” population is one that wasn’t created by anyone (except perhaps by a cosmic god). “A few,” “many,” and “most” can be understood along the lines of the numerical simulation argument in chapter 5: e.g., at least one in ten, one thousand, and 99 percent. The case for premises 2 and 3 goes through roughly as before. Premise 2 follows from mathematical reasoning, and premise 3 seems to follow at least if we are typical intelligent beings. If one had made this argument a century ago, the main objection would presumably have been: Why believe the first premise...

What the simulation idea adds is a reason to believe ...

Like the simulation argument, this creation argument does not really get us all the way to the conclusion that we are probably created.

The simulation argument could be blocked by holding that humanlike sims are impossible, or that they’re possible but most humanlike nonsims won’t create them. Likewise, the creation argument could be blocked by holding that creating humanlike beings is impossible, or that this is possible but most top-level humanlike beings won’t create them. Still, we can get to the same sort of three-way conclusion: either most beings are created, or most humanlike populations won’t create humanlike populations, or creating humanlike beings is impossible.

Before computer simulation technology, an atheist could easily enough accept the second alter...

The simulation route leads to a distinctive sort of god. The simulator is a natural god, one who is part of nature.

The ontological, cosmological, and design arguments are often used to argue for a supernatural god, one who stands outside nature. The simulator is beyond our own physical universe but not beyond nature as a cosmic whole. In principle, the simulator can be explained by the natural laws of the cosmos.

As a result, the simulation hypothesis is compatible with naturalism. Naturalism is a philosophical movement that, at a minimum, rejects the supernatural. It holds that everything is part of nature and can be explained by natural laws. Many have thought that naturalism and God are hard to reconcile, so that naturalism should lead to atheism. The simula...

The most famous argument against the existence of god is the problem of evil. An all-good, all-knowing, and all-powerful god would not permit evils such as natural disasters and genocide in t...

But it’s worth noting here that the problem of evil is no obstacle to a naturalistic simulator god. As we’ve seen, a simulator need not be all-good. She may ...

In Stanislaw Lem’s 1971 short story Non Serviam, Professor Dobb, a specialist in “personetics,” creates a society of artificial “personoids.” After many generations, the personoids speculate about the nature of their creator. The personoid Edan 197 assumes that their god requires reverence and gratitude in order for them to win salvation; if the personoids don’t believe in their creator, they won’t be saved. Adan 900 says that this would be unjust: God hasn’t given them strong evidence of his existence, so he cannot justly punish them for not believing in him; a perfectly just god would save nonbelievers as well. Adan 900 goes on to suggest that since an almighty god could have provided certainty, the fact that God has not provided certainty suggests that God is not almighty.

Under the assumption that we’re living in a simulation, we can speculate about the nature of our simulator. Is the simulator probably humanlike, or some sort of artificial intelligence? Is the simulator running the simulation for entertainment? For science? For decision-making? For historical analysis?

For example, we might wonder whether our simulator is more likely to be a biological or quasi-biological entity in its own universe, or something more like an artificially intelligent system or perhaps a simulated being that inhabits its own simulation.

At least in our world, it seems probable that in the long run, AI systems will be both much faster and much more capable than biological systems. If so, we can expect that AI systems will produce many more simulations than biological systems. It seems not unreasonable that the same may apply throughout the cosmos. If so, we should expect our simulator to be an AI system and not a biological or a quasi-biological system.

Here’s another application of statistical theological reasoning: It’s plausible that simulations run for scientific purposes will be more common than those run for entertainment purposes. Scientific simulations will be run in large numbers at once, whereas entertainment will require many fewer simulations, perhaps not more than one per person at a time. If so, it’s much more likely that our simulator is a scientist rather than a fan.

Suppose we want to study just how fine-tuned our universe is for life. If we have good-enough simulation technology, we can set up a large portfolio of different simulated worlds with different laws and different initial conditions. We can run all the simulations and see how many end up evolving life. The more simulations we run, the more accurate information we’ll get. For all we know, scientists in the next universe up are running billions of simulations of this sort.

Another role for simulations is in decision-making. In the Black Mirror episode “Hang the DJ” (spoiler alert!), people using dating apps on their phones routinely run simulations in order to determine how compatible they are. It’s typical to run 1,000 simulations instantly. Each simulation places simulated versions of the prospective couple together and sees whether a successful relationship results. If 998 of 1,000 sim...

Do simulators allow the use of simulation technology by the people within the decision-making simulation? If they do, the computational costs will be enormous, and a huge chain of simulations within simulations threatens. If they don’t, then the simulated reality will be quite unlike the unsimulated reality (where use of simulation technology will be ubiquitous). Either way, when a given simulation technology becomes widespread among a group of people, it thereby becomes less useful for purposes of predicting what those people will do.

In any case, simulating people like us (who lack advanced simulation technology) won’t be useful for making decisions in a world where that technology is present. So perhaps our simulator is more apt to be a scientist than a decision-maker.

There’s reason to think that most science-based simulations in the cosmos will be part of large batches of simulations all similar except for tiny tweaks.