Many classical physics, such as Newton's laws of motion, can be described by mathematics. You can also verify the theorems by doing experiments and collecting data from measurements and observations.
It is easier for us to "recognize" or understand the space and time dimensions (3 dimensional space x, y, z axis and time dimension, total 4 dimensional space). The laws of classical physics scientists describe nature are mostly mathematical formulas and equations in a 4-dimensional space. Taking experimental verification of Newton's second law F = ma as an example (the simple relationship between the acceleration and mass of an object), we will measure the weight, coordinates, time, and force of the object.
When the use of artificial intelligence and machine learning methods becomes mature, it opens up a new way of exploring and discovering new things.
Here is an example. NASA recently used artificial intelligence and machine learning methods to identify new exoplanets that were not previously identified from data records.
Astronomers around the world have historically used telescopes to capture and collect large amounts of data from the universe. For example, they may be the position coordinates of planets and stars, date and time, brightness, various radiation, temperature and other values, which are all "input features" of artificial intelligence terms. However, it is not easy for humans to check and analyze all data within an appropriate time, and accurately predict whether a star is an exoplanet (in AI terms, it is the output class of either Yes or No). As you can imagine, even if we have a large amount of observational data, it is very difficult to establish their internal relationship – a formula or equation that covers all the characteristics to describe the observation.
As long as we measure more features, the mathematical dimension will increase, which is even harder for us to imagine.
Here AI and machine learning help. AI does not try to derive an equation like scientists do. It just builds a mathematical model based on what it has learned – we provide AI training data sets, including feature data, and output labels (whether there are exoplanets). AI uses an optimized mathematical model to classify or predict the next input feature set. There is still a lot of mathematics involved behind machine learning, but it can be a non-linear and multi-dimensional space that we cannot easily imagine.
If Sir Isaac Newton is alive today, he might try to use neural networks to verify Newton's second law!
Image source : Wikimedia Common
Machine Learning Helps NASA Confirm 301 New Exoplanets
Newton's laws of motion