How People Think
Tyler Brooke Wilson (NYU)
Abstract:
What is thinking? If the computational theory of mind is right, it’s some kind of computation. But which? In this talk, I show how the two major paradigms for research in AI – Neural Networks and Symbolic Systems – both exhibit critical weaknesses when evaluated as models of human cognition. I use these weaknesses to motivate a third approach, based on a synthesis of the two paradigms. On this theory, thinking is realized by the use of neural methods for on-the-fly construction of symbolic mental models, tailored to task demands. I argue that architectures of this type stand to capture key aspects of human cognition.