Gary Marcus

I think you've overstepped your competence here. You can certainly argue about approaches to human cognition, but this is an area where you are entirely wrong. Physics has always employed numerical methods, DL is another tool that can improve these methods.

A friendly word of advice to @GaryMarcus , do not step into this arena (i.e. using DL for Physics). You will be eviscerated. Stick with what you know best, human psychology.

But it's not just the physical sciences. There are compelling arguments to employ deep learning in neuroscience and biology. DL methods are extremely powerful. They might not get you human complete intelligence, but they will get you something else that's valuable.

Here are the physicists and neuroscientists at Standford employing deep learning to understand the functioning of the retina: https://papers.nips.cc/paper/9060-from-deep-learning-to-mechanistic-understanding-in-neuroscience-the-structure-of-retinal-prediction …

Here is @KyleCranmer discussing an entirely new way of doing simulation with deep learning:https://twitter.com/KyleCranmer/status/1192088743970254850 …

Deep Learning is 'Artificial Intuition' technology. Like 'Artificial Logic' that preceded it, it's utility will have a tremendous effect on the sciences and this is despite not achieving human intelligence. 'Artificial Logic' and 'Artificial Intuition' are both superhuman.