John wrote, “… I think there is a judgment element that humans perform in an apparently random-like fashion that machines cannot do.”
It isn’t random. It’s a cognitive process built on biological neural networks. It’s true that we don’t understand our own cognition very well, but we know it’s definitely not random.
Thing is, we are gradually improving our understanding of how neural networks work in biological organisms. And that improved understanding is bleeding over into computer design. Neural networks are big in computer science right now, and they have practical applications that were impossible just a few short years ago (like autonomous vehicles). Now we are at the stage where machine cognition, applied to driving decisions, outperforms human cognition by a considerable margin. It’s not that autonomous vehicles will never make errors; it’s that they will make far fewer errors than humans can manage at the same task.
Comparing the rate of improvement in machine cognition to the rate of improvement in human cognition is like comparing the speed of an SR-71 Blackbird to a slime mold.
Harnessing biological computation is just one avenue of research, and one which has not yet yielded practical applications. You don’t have to presume that’s the only path forward for machine cognition.