*AI, A Modern Approach," was given that title to break from previously narrow approaches to duplicating desirable qualities of human thinking.
Although Bayesian networks require somewhat resource intensive computational elements, the importance of Bayesian inference and probability are still of paramount importance in that some of the highest scientific thinking require mastery of them. Furthermore, developing in silicon dies (or possibly graphene nanites) the machinery to perform elementary probability computations in massively parallel architectures may arise over the next few decades. The use of video DSP circuits to implement ANNs is a notable segue into this kind of development.
I would not dismiss the techniques you just read about. If your intention is to capitalize on the recent crazes, you may enter the river of wannabees chasing every current trend, implement many systems that other people will rewrite later, and have a meaningless yet profitable career. I would recommend following your inquisitiveness instead.