MLPs were fun, but as you must have observed while playing with MLP codes in the previous section, the time to learn increases as the complexity of input space increases; moreover, the performance of MLPs is just second to the ML algorithms. Whatever you can do with MLP, there's a high probability you can do it slightly better using ML algorithms you learned in Chapter 3, Machine Learning for IoT. Precisely for this reason, despite backpropagation algorithm being available in the 1980s, we observed the second AI winter roughly from 1987 to 1993.
This all changed, and the neural networks stopped playing the second fiddle to ML algorithms, in the 2010s with the development of deep neural networks. Today DL has achieved human level or more than human level performance in varied tasks of computer vision like recognizing traffic signals (http://people.idsia.ch/~juergen/cvpr2012.pdf), faces (https://www.cv-foundation.org/openaccess/content_cvpr_2014/papers/Taigman_DeepFace_Closing_the_2014_CVPR_paper...