The actual Fanchurin published a paper “This World as a Neural Network” on the pre-printing server arXiv last August, exploring the possibility that the entire universe is a neural network. It is amazing because few scientists dare to publish such a study so boldly; after all, the universe being a neural network is usually considered the realm of Science Fiction or folklore, and scientists are not very serious when they say it.
Fanchurin’s core idea is simple: every observable object in the entire universe can be modeled by a neural network, which means that in a broad sense, the universe itself may be a neural network. And under appropriate constraints, the dynamics of the cosmic neural network can be fully approximated by the still-developing quantum mechanics and general relativity.
Two different types of dynamical degrees of freedom are identified, including “trainable” variables (e.g., bias vectors or weight matrices) and “hidden” variables (e.g., state vectors of neurons). In the stochastic evolution of trainable variables, van Churin shows that the dynamics can be well approximated by the Madelungequations when close to equilibrium and by the Hamilton-Jacobi equation when far from equilibrium. This implies that trainable variables can indeed exhibit both classical and quantum behavior. And when studying the stochastic evolution of hidden variables, it was similarly found that the learning dynamics of neural networks can indeed exhibit the approximate behavior described by quantum mechanics and general relativity.
So Fanchurin came to the extremely bold conclusion that the universe at its most fundamental level may be a neural network, and that this is the actual way the world around us works. He thought it would be easy to disprove his idea by simply finding a physical phenomenon that could not be described by the principles of neural networks, yet no such phenomenon seems to be found yet.
Fanchurin even goes further and suggests that natural selection may occur at all scales, and that cosmic neural networks follow the same principles of natural selection. Whether it is the cosmic scale larger than 10^15 meters, the biological scale from 100 meters to one millionth of a meter, or the subatomic scale smaller than 10^-15 meters, natural selection may be ubiquitous.
According to natural selection, the more stable structures will survive. Then what we now call atoms and particles may actually be the result of long evolution starting from some very low complexity structures; and what we now call macroscopic observers and biological cells may be the result of even longer evolution.
It seems that this is hard to force the universe into the way of a giant creature, so what are we if the universe is a creature? Each person is a node in a giant neural network? The purpose of our existence is well explained, so what is the meaning of the existence of this cosmic creature?
Of course, Fanchurin himself believes that the current idea that natural selection may be relevant at all scales is still speculative, but looking at the universe as a giant neural network does seem to offer a very interesting new perspective for the observer.
So what is your view? How would you prove that the universe is a giant neural network, or simply a giant creature?
Recent Comments