Devoting time and human brain power towards solving complex problems is not always something that is feasible. Some of the greatest discoveries of our time emerged from prominent thinkers over many centuries. The rate of technological and scientific discovery has been massively influenced through the help of machine learning and a wide range of leaps throughout our recent past.

Complex subjects like quantum mechanics, for example, have been based on a series of mathematical ideas that rose to prominence by applying new statistical models. The same theory crafting often occurs when discovering new concepts using machine learning. Researchers at ETH Zürich in Switzerland are developing methods to discover new laws within physics. New formulations on physical laws would not commonly have a place within physics research for human researchers; machine learning, however, is leading to simulations that are delivering faster answers on subjects that we have faced in reality.

Neural networks can find new laws within physics and even perform observational cataloging. In current simulations with the pendulum experiment, miniaturizing physics problems like the positions of planets in our night sky can be regulated using various machine learning simulations. The networking continues to use variables to come up with solutions and then generate further results using relevant laws of motion from the miniaturized results.

The research here is the first demonstration in which an artificial neural network is capable of compressing data revealing laws of physics. Computationally approaches are working to derive new laws, and a number of other researchers are using machine learning to make discoveries as well. Cornell University, for example, is using genetic algorithms to ascertain new data about the process of evolution.

As we continue to discover new physical concepts using neural networks we can drive innovation forward. Rather than having to rely on the next great thinker or large-scale research budgets to determine the laws of our universe, we could run machine learning simulations and simplified data that has been a mystery to humankind.