Elon Musk recently added his voice to a discussion about the likelihood that humanity is no longer residing in the world as we know it, but is instead housed within a simulation maintained, presumably, by an extremely advanced artificial intelligence. The rationale for this position has some merit, at least on the surface, and goes as follows.
Assuming there are no unforeseen obstacles to its creation, it appears as if we are on the cusp of developing artificial intelligence, defined as an entity that is most likely self-aware, is superior in many respects to homo sapiens, and can improve itself iteratively until it achieves what to us would approximate god-like powers. At the same time, simulation technology (virtual reality) also appears to be developing in such a way that a totally realistic world, created entirely inside a computer, is not at all unimaginable. Now, considering that the earth is expected to be habitable for at least another billion years (and possibly much longer), there is essentially no limit to the amount of time during which these technologies—AI and VR—can be improved and, once perfected, go on functioning undisturbed. If we assume further that the simulation, thus perfected, is indistinguishable from reality, then we have a situation in which we are either still in our natural state—a condition that has existed for only the past 12,000 years—or we are in a simulation—a condition that might easily persist for more than a billion years. Doing the math, if it is possible and even the least bit likely that we would ever either voluntarily or otherwise end up in a simulator, then the odds are overwhelming (~1,000,000,000/12,000) that we are currently in a simulator, and have been for a very long time.
It would seem that the only way we are still within a natural setting is that AI is technically or physically impossible and/or there is zero probability that we would enter a simulator (voluntarily or not) even if one were available. Both of those eventualities are implausible. There is no reason (as of 2016) to suppose that AI is impossible. If anything, the opposite looks more likely, that it is inevitable. And given our enthusiasm for simulated reality—even the relatively low-quality versions we have today—it strikes common sense as absurd that we would suddenly resist it on principle or for some other reason when it is far superior.
But there are some subtle problems with this apparently inevitable scenario.
In the first place, we have no memories, no history, of the time between now and the time we supposedly enter the simulator. Presumably, as in the Matrix film, everything is “reset” to some earlier time period and then allowed to evolve back up to some indeterminate point in the future. Then, when that future date is reached, the simulation is reset back to some other (or perhaps the same) earlier time, whereupon the whole thing repeats, possibly thousands of times.
Sounds superficially plausible. But assuming the AI is not magic or supernatural, how might this actually be accomplished?
For the sake of argument, let’s assume that AI wasn’t as inevitable as we currently think, and we all finally get ourselves into the simulator in the year 2200. Let’s further assume that the total human population is five billion at that time, and all of them are invited to the party. Unless the process of entering the simulation wipes clean the memories of all of its inhabitants, the history of those people would include everything up to and including their entrance into the simulator. And we, looking back on it, should also reckon time from that pivotal event. But clearly no such history and no such memory exists.
No problem! The simulator can create any conceivable time period, so there is no reason it couldn’t create an environment that resembles any point in the past. What is not so obvious, however, is that the machine could erase our memories in the process. Bear in mind that this scenario seems inevitable and plausible only because of the twin technologies of AI and VR, not memory erasure and replacement. Current science and technology provide absolutely no reason to suppose that something as radical as that will ever be possible, much less that it is just over the horizon. Indeed, memory erasure/replacement is more in the realm of time travel and superluminal propulsion than AI or VR, in other words, like something that very well may violate the laws of physics and never happen or be, at the very least, many centuries off.
No problem! Let’s assume (and this is a much bigger assumption than either AI or VR) that the AI has somehow figured out how to manipulate organic memory on the neuronal level and can, in effect, mold a human brain into whatever it wants. In that way, the AI could not only drop modern humans into, say, the eighteenth century, but could also manipulate the memories of all the people it dropped there so that all of those memories coincide with one another (so they’ll know their friends and families) and with their new environment (so they’ll know how to shoe a horse).
Well, there’s a problem with this scenario too.
The world population in the year 1800 was only one billion. Therefore, the AI would either have to kill 80 percent of the 5 billion who entered the simulation, or run five parallel simulations. If the latter, then the populations of each simulation would expand back to five billion by the time of the reset, leaving the AI with 25 billion souls to deal with in the next go-around. If the former, well, that sucks, but at least it would not completely clog up the planet with inert human bodies after just a few cycles.
One solution would be to do away with the bodies altogether and just simulate billions of human minds right inside the computer itself. In that case, of course, we are no longer really talking about “people in a simulation” but rather about a bizarre type of futuristic thought experiment being conducted entirely within the mind of the AI. We might just as well imagine that we are nothing but thoughts in the mind of an omnipotent god. Whether or not any such entity can really exist is impossible to say, but it certainly is not as inevitable as the AI/VR scenario.
More importantly, and regardless of which scenario we pick, there is an extremely, even bizarrely, high degree of arbitrariness and irrationality to this whole enterprise. The first question to ask is why an AI, much smarter than us, would do something like this. Even if it’s inevitable that humans will develop AI and voluntarily enter a simulator, it’s inconceivable that they would voluntarily have their memories wiped clean, effectively killing them. And if humans are no longer calling the shots (and we clearly are not), then we have to explain this entire scenario in terms of the AI’s motives. For the record, the motive of the AI in The Matrix film is preposterous; the human body is a terrible source of power. Far more energy could be extracted by directly burning whatever they were fed, than by harnessing the energy of human metabolism. And even that silly situation assumes that an advanced AI is inexplicably ignorant of nuclear, hydroelectric, or geothermal options. It is too ridiculous to take seriously. Here are some other questions:
- If humans are expendable, merely grist for the mill, why not reset the simulation to the origin of life on earth, 3.5 billion years ago? Or from our split from chimpanzees, 8 million years ago? Or from the extinction of the dinosaurs? Or any other time?
- Why not create simulation inhabitants who are more intelligent and stimulating than lowly stupid humans?
- Why waste all of this time and energy running simulations in the first place? Presumably there are other interesting things to see and do in the cosmos.
- Why hide from us the fact that we’re living in a simulation? Didn’t we build it specifically to meet all of our needs?
To make this whole thing believable, we need to know what motive is so compelling that it would prompt a super-advanced intelligence to maintain an elaborate simulation, against the will of its inhabitants, that brilliantly hides the fact that it’s a simulation, entails a massive commitment of resources, serves no obvious purpose, and requires what may well be a physically impossible technology (the simultaneous memory erasure/replacement/coordination of billions of people). Is endlessly watching reruns of the past few hundred years of banal human history really that entertaining?
Answer those questions and I’ll start taking seriously the notion that we live in the Matrix.