A Wholistic Integrated Cognitive System (SEER-T1) that Interacts with its Environment Over Time
University of Wisconsin-Madison Department of Computer Sciences
MetadataShow full item record
This paper describes, presents and discusses a computer-programmed model for a wholistic system that performs the variety of functions usually thought of as "cognitive" while interacting with an environment of objects that move about and change over time. A relatively sophisticated " recognition cone" system that perceives and describes scenes of moving and changing objects (Uhr, 1975) is embedded in a larger "SEER" (for SEmantic learnER, or Sensed Environment Encoder Recognizer and Responder) system that also associates into memory to respond to queries and access relevant information, deduces solutions to simple problems, parses and "understands" language utterances, effects actions upon the external environment, and learns its memory of transforms and their interconnections. These functions are kept as simple as possible, and combined in as integrated a way as possible. Each calls upon and helps the others as needed, by implying things, processes and acts into a small number of common lists to which all have access. A unified "cognitive memory" is built up using a single general type of transform, one that is powerful enough to do pattern recognition " feature extraction," scene description "configurational characterizations", memory "associations," deductive "productions", and language "rewrite rules." Embedding this system in time allows for new interesting and natural appearing interactions among the various processes as they work in parallel and serially call upon one another over time. The program is coded in EASEy (Uhr, 1973f) an English-like variant of SNOBOL designed to enhance list-processing abilities, and to be easy to read, so that the program itself can be examined and understood.