Yes, it was programmed, but sadly not for mathematics.

Interestingly, I think the architectures we create for neural networks are or can be similar to the brain structures evolution came up with. For example, groups of biological neurons correspond to hidden layers, action potentials in dendrites are similar to activation functions, and the cortex might corresponds to convolutional layers. I’m pretty sure we will eventually invent the equivalent of neuroplasticity and find the other missing pieces, and then singularity or doomsday will follow.