It's an interesting paper, but I really worry that headlines like this veer into mysticism. The TL;DR is that if we want to understand why neural networks are so effective, we need to remember that the sets they've been trained to understand follow a small number of physical laws, as opposed to just focusing on the abstract math of how NNs work.<p>That's an interesting perspective (both authors are physicists, without much experience in NNs per se, although Max Tegmark has long been philosophizing about the future of AI) but it's not as if neural networks have a particularly magical key to "the nature of the universe." Original headline said the link is "extraordinary," when the link is just that... NNs exist in reality.
My article.<p>The Link Between Quicksort and the Nature of the Universe.<p>Consider Quicksort. In an array of 100 digits, there are over 10^157 different permutations, more permutations than there are atoms in the universe. To go through every permutation, even at a billion permutations per second, would take longer than the lifetime of the universe. Yet your computer can sort through an array in milliseconds. Why? At the heart of this mystery, says Tegmark, lies the true nature of the universe.