A new computer algorithm, or set of rules, that accurately predicts the orbits of planets in the solar system can be adapted to better predict and control the behavior of plasma which drives the melting devices created to reap on Earth the melting energy that powers the sun and stars.
The algorithm, devised by a scientist at the US Department of Energy (DOE) Princeton Plasma Physics Laboratory (PPPL), applies machine learning, the form of artificial intelligence (AI) that learns from experience, to develop predictions. “Usually in physics, you make observations, create a theory based on those observations, and then use that theory to predict new observations,” said PPPL physicist Hong Qin, author of a paper detailing the concept. Scientific Reports. “What I’m doing is replacing this process with some kind of black box that can produce accurate predictions without using a traditional theory or law.”
Qin (pronounced Beard) created a computer program in which it fed data from past observations of Mercury orbits, Hesperus, Land, March, Jupiter, and the dwarf planet Ceres. This program, along with an additional program known as a “service algorithm,” then made accurate predictions of the orbits of other planets in the solar system without using Newton’s laws of motion and gravity. “Basically, I overlooked all the basic components of physics. “I go directly from data to data,” said Qin. “There is no law of physics in between.”
The program does not happen with accurate predictions by chance. “Hong taught the program the basic principle used by nature to determine the dynamics of any physical system,” said Joshua Burby, a physicist at the DOE Los Alamos National Laboratory who earned his Ph.D. at Princeton under the mentorship of Qin. “The solution is that the network learns the laws of planetary motion after seeing very few training examples. In other words, his code really ‘teaches’ the laws of physics. “
Machine learning is what makes software like Google Translate possible. Google Translate examines a large amount of information to determine how often a word in one language is translated into another word. This way, the program can do an accurate translation without learning any language.
The process also manifests itself in experiments of philosophical thought such as John Searle’s Chinese Chamber. In that scenario, a person who did not know Chinese could “translate” a Chinese sentence into English or any other language using a set of instructions or rules that they would substitute for understanding. The thought experiment raises the question of what, at the root, means to understand something, and whether meaning implies that something else is going on in the mind besides following the rules.
Qin was inspired in part by the philosophical experiment of Oxford philosopher Nick Bostrom that the universe is a computer simulation. If this were to be true, then the basic laws of physics must reveal that the universe is made up of individual pieces of space-time, like pixels in a video game. “If we live in a simulation, our world must be discreet,” Qin said. The black box technique, the invented Qin, does not require physicists to believe the simulation of simulation literally, though it is based on this idea to create a program that makes accurate physical predictions.
The resulting pixelated view of the world, similar to what is portrayed in the film MATRIX, is known as a discrete field theory, which views the universe as composed of individual pieces and differs from the theories that humans normally create. While scientists usually create general concepts of how the physical world behaves, computers simply collect a collection of data points.
Qin and Eric Palmerduca, a graduate student at Princeton University The program in Plasma Physics is now developing ways to use discrete field theories to predict the behavior of plasma particles in fusion experiments performed by scientists around the world. The most commonly used melting devices are donut-shaped tokamaks that limit plasma to strong magnetic fields.
Melting, the power that drives the sun and stars, combines light elements in the form of plasma – the hot, charged state of matter made up of free electrons and atomic nuclei that make up 99% of the visible universe – to generate massive amounts of energy. . Scientists are looking to copy the union on Earth for a virtually inexhaustible supply of energy to generate electricity.
“In a magnetic coupling device, the plasma dynamics are complex and multidimensional, and the effective governing laws or computational models for a particular physical process we are interested in are not always clear,” Qin said. “In these scenarios, we can apply the machine learning technique I developed to create a discrete field theory and then apply this discrete field theory to understand and anticipate new experimental observations.”
This process raises questions about the very nature of science. Do not scientists want to develop physics theories that explain the world, rather than just collecting data? Are not basic theories about physics necessary to explain and understand phenomena?
“I would argue that the ultimate goal of any scientist is prediction,” Qin said. “You may not necessarily need a law. For example, if I can perfectly predict a planetary orbit, I do not need to know Newton’s laws of gravity and motion. You could argue that by doing so you would understand less than knowing Newton’s laws. In a sense, this is correct. But from a practical point of view, making accurate predictions is doing nothing less. ”
Machine learning can also open up opportunities for more research. “It significantly expands the range of problems you can address because all you need to keep up is the data,” Palmerduca said.
The technique can also lead to the development of a traditional physical theory. “While in a sense this method excludes the need for such a theory, it can also be seen as a path to one,” Palmerduca said. “When you are trying to come up with a theory, you want to have as much data as you can. If you are given some data, you can use machine learning to fill in the gaps in that data or otherwise expand the data set. “
Reference: “Learning and machine service of discrete field theories” by Hong Qin, November 9, 2020, Scientific Reports.
DOI: 10.1038 / s41598-020-76301-0