How a Neural Network Learned to See Quantum Foam ⚡ экспресс
Scientists are teaching a neural network to predict important numbers in quantum gravity theory — much like a program recognizes faces in photos rather than recalculating pixels. One network checks if the result is zero, another gives the exact value. This speeds up computations many times over. Will a machine ever learn what the universe is woven from?
The concept of quantum foam dates back to Roger Penrose and Carlo Rovelli: space is like soap foam, where each bubble is a microscopic chunk of reality. These bubbles constantly burst and merge, and to understand how the Universe works at its deepest level, one must compute the probabilities of their interactions. Previously, such calculations took hours even on powerful computers.
Neural networks handled it in milliseconds. The first network, like a meticulous sorter, instantly discards impossible variants (where probability is zero). The second, like an experienced appraiser, gives precise values for the rest. Astonishingly, it correctly filters out zeros even in unfamiliar scenarios—much like a person who has seen a couple of cats recognizes any breed.
So far, this works on a simplified model, but it paves the way to testing hypotheses about the moment of the Big Bang and the structure of curved spacetime inside black holes. Thus, soap foam brings us closer to solving the mystery of the birth of the Universe.
🎯 The idea that space may not be continuous but consists of minuscule grains dates back to the ancient Greek atomists. In modern physics, it was revived by [scientist:John Archibald Wheeler]John Archibald Wheeler[/scientist], who coined the term 'quantum foam'.
🎬 In the novel 'Blindsight', aliens communicate instantaneously thanks to the quantum nature of space. Perhaps understanding quantum foam will turn this science fiction into reality.