Abstract: Simulated on a computer, PURR-PUSS is a machine that can be taught. When connected to a robot body in the real world, ‘she’ can learn on her own or with the help of a human teacher. Her experience is stored in small ‘pieces’ which are used for prediction, decision and novelty-seeking.
Unlike other products of Artificial Intelligence research, PURR-PUSS illustrates the difference between computation and thought by learning formal techniques and skills within the unstructured, informal substrate of her stored experience. A multiple context provides the necessary basis for generalization and discrimination, and suggests a reason for the success of parallel computing in the brain.
Simple in concept and implementation, this system opens up new areas for research by simulation in education, psychology and neurophysiology.
John H Andreae, “PURR-PUSS: Computer Simulation of a Teachable Machine,” SAGSET Journal. 8. 123-133.