The scientific method has been extremely successful, leading to exponential growth in our understanding of the Universe. But in real life no exponential can go forever. We will either explain everything in one grand unified theory, or we'll hit the wall. We hope for the former, but have to be prepared for the latter.
So far "understanding" meant decomposing complex systems into simpler components. That's what math is good for, and category theory in particular. That's how we write programs. But there is a new way of dealing with problems by training neural networks. Does that lead to "understanding" though? Is AI the killer of the scientific method?