Abstract

Recent research in artificial intelligence and machine learning has largelyemphasized general-purpose learning and ever-larger training sets and more andmore compute. In contrast, I propose a hybrid, knowledge-driven,reasoning-based approach, centered around cognitive models, that could providethe substrate for a richer, more robust AI than is currently possible.

 

Quick Read (beta)