>>4Machine learning has had far too marked and rapid a progression in recent years to dismiss off-hand anymore. New techniques such as attention [1, 2], generative-adversarial networks [3], deep q-learning [4], etc, though requiring large clusters of GPUs to train effectively, have pushed forward the state of the art in numerous problem domains substantially and offer a clear path for future gains as well. Corporate support from the likes of Baidu, Google, and Facebook attracts a steady flow of grad students ensuring continued growth and innovation. Naysayers like Sussman will have to eat their words eventually or face complete and utter irrelevancy.
[1]
https://arxiv.org/abs/1409.0473[2]
https://arxiv.org/abs/1502.03044[3]
https://arxiv.org/abs/1511.06434[4]
https://arxiv.org/abs/1312.5602