Can we open the black box of AI?