ABSTRACT
After attempting for years to capture and implement the elusive quality that makes things intelligent, researchers settled on studying and generating “purposeful behaviour” by various means. This resulted in a class of autonomous agents that can learn and behave effectively in a variety of new situations, by exploiting statistical patterns in their environment, which eliminated the need for explicit rules of behaviour but introduced the need for large amounts of training data. This shortcut produced a new paradigm for the field that is based on machine learning and incorporates new expectations, new tools, and new success stories. Recommender systems are better representatives of artificial intelligence (AI) agents than theorem provers. The new language of AI is the theory of probability and optimisation, no longer that of logic and reasoning.
