Another way to think about books is as training data for a deep learning network. When you feed a thousand photos of airplanes into a neural network your goal isn’t to get the network to remember the specific airplanes in your training dataset. If you did you’d probably have made the mistake of overtraining it. Rather, your goal is to teach the network to recognize the abstract concept of an airplane. Books could be similar. Their purpose could be to feed many variations of a single concept through a human brain so it can absorb the pattern.
Another way to think about books is as training data for a deep learning network. When you feed a thousand photos of airplanes into a neural network your goal isn’t to get the network to remember the specific airplanes in your training dataset. If you did you’d probably have made the mistake of overtraining it. Rather, your goal is to teach the network to recognize the abstract concept of an airplane. Books could be similar. Their purpose could be to feed many variations of a single concept through a human brain so it can absorb the pattern.
Super interesting and likely worth developing into a longer post if you’re so inclined. Really like this analogy.