Symbolic Supercomputer for Artificial Intelligence and Cognitive Science Research
Funded by: Office of Naval Research, Defense University Instrumentation Program
Principal Investigator: Ken Forbus
Overview
Supercomputers are invaluable in many areas of science and engineering. Unfortunately, artificial intelligence and cognitive science research has not benefited from supercomputing because traditional supercomputers are primarily aimed at numerical simulation. We believe that we can change that by building a symbolic supercomputer, a high-end cluster machine carefully configured to support the creation and modeling of intelligent systems. We have created a symbolic supercomputer, to support research being conducted on two ONR projects:
- The artificial intelligence project,
Qualitative Reasoning for Intelligent Agents (PI: Kenneth D. Forbus), is
exploring the use of qualitative modeling to provide capabilities for
intelligent software agents, including common sense reasoning.
- The cognitive science project, Analogical Learning and Case-Based Instruction (PI: Dedre Gentner, co-PI: Forbus), is exploring the roles of analogy and similarity in learning and reasoning from examples, to help provide the science base for better education and training.
We purchased a Linux Networx Evolocity cluster, with 67 nodes and a gigabit switch (see below). Each node has two 3.2 Ghz Pentium Xeon processors, with 4GB of RAM and 80GB of disk. This machine, called mk2, became operational on 8/25/04. We are already using it to support the projects above.
Back to Symbolic Supercomputing page | Back to Projects page | Back to QRG Home Page