Robert Alexander
Nov 4 · 1 min read

As a medical doctor I wrote my doc thesis on “expert systems” in 1985 even sweating over Prolog to capture and compute over human extracted rules. As we all know it became quickly apparent that extracting those explicit rules was very hard and the implied unexpressed knoweldge necessary to solve problems. Winter ensued.

I know see the misunderstanding associated to the antropomorphic terms “learns”, “sees”, “understands” used routinely also with the current subsymbolic approaches is driving expectation on ML/DL being able to infer facts from data and the subsequent reaction.

I wonder how much bringing together knowledge based, connectionist, Bayiesian and other “AI” approaches together might push us towards truly useful applications. I also wonder if we still have teams with all of these skills to produce such possible systems.

Thank you for the very nice article.

    Robert Alexander

    Written by

    Medical doctor, technology expert, lifelong learner, polymath,