Neat critique of deep learning.

Bupe brings useful perspective to the increasing controversy over the depth and reach of deep learning. He acknowledges its utility. It is of course important not to throw the baby out with the bathwater. But he then carefully enumerates its shortcomings. Foremost amongst these is its inability to do one/zero shot learning. Its absence of real world context, or priors, also does not help. It is important to remember that the human brain takes different approaches to different challenges. The emotional pathways operate in a different fashion from the visual cortex. AI, or AGI, when we get there, will doubtless be the same. AI research will simply have to continue until we have figured it all out. And it will.

Link to article:

You may also like to browse other AI posts: