Science can’t tell if we’re living in The Matrix

We live in a unique era, at the boundaries of what transistor-based computer technology can offer.

Thanks to that, we can now exploit the impressive versatility of previously prohibitive computational techniques, such as deep learning and other methods of artificial intelligence (AI), to the advantage of scientific research. Such tools are proving so powerful that some people are starting to argue either that we live in a simulation, or that there is a god and it is AI itself.

Neural networks, a currently very popular AI method, are known to be universal encoders, which means that, in principle, any problem of any type can be learned and therefore predicted by the network (prohibitive computational costs notwithstanding). Unfortunately, this is not true in practice.

Computers are finite-state machines, with finite memory, operated by myopic living beings: humans. This implies that chaotic systems (that is, nearly everything observable) cannot be represented exactly in a computer.

Consider the number pi: it is an irrational number containing what seems to be a random, infinite sequence of digits. Neither computers nor humans can represent, or operate with, the true pi: we must approximate it. Fortunately, we have a recipe to approximate it to any precision, but for almost all other irrational numbers the situation is much worse, as they are impossible to compute.

If no human can see these numbers, and no computer can really calculate them, do irrational numbers even exist? They do, at least in our imagination.

Nature, as we see it, is governed by laws. Anything observable or imaginable obeys them. Physics is just the human-friendly version of a very small fraction of such laws, and it concerns only the observable phenomena. However, physics itself is based on human-centric imaginative assumptions and models.

For example, Newton’s laws are never exactly observed in nature: they are a simplified, imaginative set of models able to approximately, yet acceptably, describe several phenomena.

Quantum mechanics gives us insight on the finest grains of reality as we can perceive it by telling us that our world is made of funny-behaving “pixels” (Planck length, for the pros), literally several hundreds of trillions of trillions of times smaller than the atom. 

These are all models, and models are nothing but the imaginative representations of observable phenomena.

Humans understand nature whenever they can associate observation with imagination. The true problem arises when humans attempt to understand the supernatural. Religious people may give you a different perspective, but we must draw a clear line: nature, by definition, cannot be supernatural, and therefore the supernatural cannot possibly be observed in nature.

If a human can imagine or observe a phenomenon, then it clearly cannot be supernatural; hence the very definition of “supernatural” must be part of the conceivable domain of nature. That said, I have my strong reservations about whether we can ever prove or disprove our being part of a simulation in some big alien computer, especially if such a simulation is the Creator of nature itself.

Disclaimer: Any findings and conclusions are those of the author, and do not necessarily reflect the view of Lawrence Livermore National Laboratory or the Department of Energy of United States of America. This article has been approved by Lawrence Livermore National Laboratory for public release, IM number LLNL-JRNL-739760.