Some Critical Factors that Limit the Effectiveness of Machine Intelligence Technology in Military Systems Applications

摘要

Machine intelligence technology is not the product of a recent breakthrough in computer system design as many have been led to believe. The fundamental concepts, and many of the underlying problems in the design of today's intelligent machines, were recognized fifty years ago, and earlier. Research efforts in diverse areas, some spanning decades, have culminated in the current state of the technology. However, numerous and often perplexing methodological problems currently challenge system development and thwart the widespread and successful application of the technology in critical areas. The issues discussed in this paper arise from the failure of MI technology to confront and effectively solve many of the complex problems inherent in military systems applications. Most critical is the fact that performance criteria for intelligent machines are vague and inadequate, or completely unspecified in some instances. It is often impossible to ascertain from the reported data exactly what a system is supposed to do and how well its functions are performed. The problem of performance assessment is one of several difficult issues discussed in this report.