Neural nets have been around for a long time. Why are they popular only now?
Here are two slides from Geoffrey Hinton's talk summarizing the problems:

!<{margin-right: 20px; auto;display:block;border: 1px solid gray;}blog/deep-learning-what-was-wrong-in-1986?wrong1986.jpg 35% !

!<{margin-right: 20px; auto;display:block;border: 1px solid gray;}blog/deep-learning-what-was-wrong-in-1986?wrong1986.jpg 15% !

!<{margin-right: 20px; auto;display:block;border: 1px solid gray;}blog/deep-learning-what-was-wrong-in-1986?viewofwhatswrong1990.jpg 35% !

!<{margin-right: 20px; auto;display:block;border: 1px solid gray;}blog/deep-learning-what-was-wrong-in-1986?viewofwhatswrong1990.jpg 15% !

I recommend the whole talk