Prof. em. Dr. Clayton Lewis 

What does the success of predictive large language models (PLLMs) like GPT3 or chatGPT mean? While many hope for widespread practical applications, and others are skeptical, this book project argues that the larger significance of these models is what they suggest about human cognition. The book will review many issues and topics in cognitive science, and identify ways in which these predictive models offer new ideas, or reinforce older ideas, about the mental mechanisms involved.

The presentation starts with a description of the Prediction Room, in which a Prediction Agent  is enclosed. The basic operations of PLLMs, and the Prediction Agent, are to create and use a predictive model of the flux of events they observe. For current PLLMs this flux of events is limited to a stream of text, but the Prediction Agent  is presumed to be able to carry out physical actions, and to perceive the effects of such actions, and other physical events in the outside world, things that current PLLMs cannot do. 

The presentation then considers a wide-ranging collection of topics in cognitive science. For each topic, phenomena for which the Prediction Room model might provide an account, and ones for which it is unable to account, are discussed. The book closes with a general discussion that takes stock of the ideas, and challenges, that have emerged.