Toward the parallel inference machine

Abstract

This paper describes an approach to the parallel inference machine (PIM) which provides a parallel programming environment to support AI applications. The importance of parallel software systems, especially parallel operating systems which control job allocation and resource management, has been recognized by many architecture researchers. However, research on these software systems is still in immatured status. This causes many problems for parallel inference machine research. In Japan's fifth generation computer systems project, parallel software research is now being emphasized and plans are conducted more systematically. The work is being pursued with one eye on parallel hardware research using a dedicated tool called a multi-PSI system consisting of several personal sequential inference machines. This paper describes research and development plans for the parallel inference machine in conjunction with the parallel software research.