The better your neural network solves any problem, the worse it is as a neural network as a whole. How so?

Question

Computer networks

The better your neural network solves any problem, the worse it is as a neural network as a whole. How so?

Answer 1

what does whole mean? A neural network is created to solve specific problems, and it either solves it or not. You can't force a neural network designed for checkers to recognize kittens from pictures.

Answer 2

Maybe a question about generalization and overfitting?
The better the network fits answers to the training sample, the worse it copes with tasks outside this sample, with “real” tasks.