AiS Challenge Team Interim Report

 

Team Number: 70

School Name: Silver High School

Area of Science: Artificial Intelligence

Project Title: Grosvenor: An Implementation of a New Cognitive Model

 

 


Grosvenor: An Implementation of a New Cognitive Model

Roeland Hancock
Presence Designs
rhancock@presencedesigns.net


One the greatest challenges of man has been to understand himself. A multitude of theories model the mind, but these are only theories—no operating implementation has ever been constructed. Computational technology has now progressed to the point that a partial, if not full, implementation of a mental model can be constructed. One of the most important aspects of a cognitive model is its ability to learn. A successful creation and implementation of such a model will provide an extremely revealing look at the mind and revolutionize cognitive science.

This project's goal is the development and implementation of a new cognitive model. The model will be a cross-disciplinary approach drawing from many sources with most of the foundation based on the “building-block” approach (Minsky 1988).

Grosvenor's ultimate goal is a cognitive model and creation of the first true artificial intelligence (AI), but will have many side applications of immediate practical and commercial use, primarily as a search engine, expert system and reference tool. These commercial opportunities will no doubt be explored in the course of future development as a means of financing further work. This goal will be resource and time intensive; Grosvenor is not expected to come to fruition in the available time frame.

To learn, Grosvenor must have a source of knowledge. Grosvenor will parse English webpages found with an Internet search algorithm based on google.com (Brin & Page, 1998) to autonomously add knowledge to its system. This new knowledge will be stored in a “knowledge net”, a specialized graph data structure interfaced with a PostgreSQL database for non-volatile storage. This data can then later be easily searched and linked together to create associations and comprehension. Grosvenor will be written in C, C++ and possibly a logic language such as PROLOG.

Due to the vast amount of information Grosvenor will encounter and search and the nature of the knowledge net search algorithm, Grosvenor will greatly benefit from, and may even require, a parallel implementation.

Current progress

After much research and thought, the cognitive model has been developed. This model may not be an entirely accurate model of human mental processes, but it is sufficiently valid to be educational and serve as a working definition.
In the way of programming, an internet search algorithm employing dictionary.com and google.com, data structures, and some search algorithms have been coded. Because writing a natural language parser (NLP) is difficult and there are already several available, an existing parser developed at Carnegie Mellon was used.

Although Grosvenor was originally conceived for parallel systems, there have been difficulties implementing this. Grosvenor requires live Internet access for knowledge acquisition, making LANL's theta supercomputer unusable. Mode will allow Internet access, but until very recently PostgreSQL was not installed.

Problems
There were some difficulties getting an appropriate computational environment, but these have, for the moment, been resolved. The developer's unfamiliarity with the PostgreSQL API and the poor API documentation have made developing the data handling code slower and more difficult than expected.

 

Basden, B. H., et al. (2002). Part-set cuing of order information in recall tests. Journal of Experimental Psychology, 25, 517-530.

Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual web search engine, http://www7.scu.edu.au/programme/fullpapers/1921/com1921.htm.

Chambers, T. C. et al. (2002). Circumscribing referential domains during real-time language comprehension. Journal of Memory and Language, 47, 30-49.

Descartes, R. (1948). The meditations. (J. Veitch, Trans.). La Salle: Open Court Publishing. (Original work published 1641).

Downs, R. M., & Stea, D., ed. (1973). Image and environment. Chicago: Aldine.

Geraci L., & Rajaram S. (2002) The orthographic distinctiveness on direct and indirect tests of memory: delineating the awareness and processing requirements. Journal of Memory and Language, 47, 273-291.

Guha, R. V., & Lenat., D. B. (1990). Cyc: a midterm Report. AI Magazine.

Johnson, et al. (2002) Text induction algorithm. IBM Systems Journal, 41.

Mattys, S. L., & Clark, J. H. (2002) Lexical activity in speech processing. Journal of Memory and Language, 47, 343-359.

Minsky, M. (ed). (1969). Semantic information processing. Cambridge: MIT Press.

Minsky, M. (1988). The society of mind. New York: Touchstone Books.

Minsky, M. (2002). The emotion machine. http://web.media.mit.edu/~minsky/E1/eb1.html

Olthof, A., Sutton, J.E., Slumskie, S.V., Daddetta, J. A, & Roberts, W.A. (1999). In search of the cognitive map: Can rats learn an abstract pattern of rewarded arms on a radial maze? Journal of Experimental Psychology, 25, 352-362.

Saffral, J. R. (2002). Constraints on statistical language learning. Journal of Memory and Language, 47, 172-196.

Wagman, M. (1991). Artificial intelligence and human cognition. New York: Praeger Publishers.

Wolfram, S. (2002). A new kind of science. Champaign: Wolfram Media.

Searle, J. R. (1982), The Chinese Room Revisited, Behavioral and Brain Sciences, 5, 345-348.

Westerman, D. L., et al. (2002). The attribution of perceptual expectancy in recognition memory: the role of expectation. Journal of Memory and Language, 47, (607-618)


Team Members

Team Mail

Sponsoring Teacher(s)