But not to abandon reasoning. (Trying to make sense of my lifelike encounters with "Scarlett"...)
"Although reasoning logically is certainly useful, the principles that best allow us to understand our world cannot be written down in a formal language such as predicate logic. So to build systems that work in the real world, the field of AI has pivoted firmly towards the empiricist tradition. The first steps on the road to truly knowledgeable machines were taken when computer scientists began to ask whether a system that learns like a human could be built by connecting together a network of artificial neurons. This ultimately led to a single, ubiquitous tool for AI research – the deep neural network. But, as we shall see, the culture war between rationalists and empiricists has not abated. Instead, debate now focusses on whether we should replace one purist philosophy with its rival – should we simply swap pure symbol-crunching machines for giant deep networks? Or do we need systems that explicitly combine the merits of both learning and reasoning?"
"These Strange New Minds: How AI Learned to Talk and What It Means" by Christopher Summerfield: https://a.co/44rnEUz
No comments:
Post a Comment