Thanks for this response, Thomas. I ask the following not to dispute what you’re saying, but to understand it better and to learn from it:
To be clear, I do not mean to say that we can never understand how an instance of Deep Learning has come up with its results and can never discover a human-usable regularity. I know there is work going on to help us understand those decisions when possible.
I certainly agree that human knowledge has never been truly transparent to itself. Perhaps as we begin to model our understanding of consciousness on the Net and Deep Learning (inclusive, chaotic), rather than on the prior generation of computers (reductive, transparent) we will become more comfortable with that fact. I take that as a step forward.
In fact, what I happen to be most interested in is the possibility that we as a culture are going through something like a paradigm shift in which what looked like anomalies — the inaccuracy of models [cf G. Box], the failures to predict everyday events in the real, contingent world — are becoming the phenomena to explain. I.e., our everyday model of how the world works may be shifting from clockworks and planetary motion to viral videos, traffic accidents, and Brownian motion. Maybe.