From the time of the first embryonic computer developments in the late 1940s and early 1950s, DOD has encouraged and sponsored AI and HCI R&D in the laboratories of industry, academia, and government. Advancements in AI and HCI have allowed DOD to accelerate work toward its goal of improving national security while reducing the risk faced by individuals in hostile environments.

This goal has manifested itself in strong support for and pressure to advance technologies that

• Permit total replacement of humans in hostile environments by "satisfactory" perceiving, reasoning, and acting mechanical-electronic surrogates;
• Permit combinations of robots and humans to carry out needed functions in hostile environments with minimal or much reduced risk to the human;
• Permit remote control of robots or human-robot systems in hostile environments to substantially reduce the risk to the human;
• Improve dramatically via education and training the capability of humans to perform satisfactorily in hostile environments, principally via simulation, emulation, and modeling processes and procedures; and
• Expedite the design, development, and manufacture of products, systems, and platforms of high quality in a timely manner and at least cost via concurrent engineering, computer-aided design (CAD), computer-aided manufacturing (CAM), flexible computer-integrated manufacturing (FCIM), and other automated design and production technologies.

The hardware, software, reasoning, and representation technologies that make up AI and HCI have played a significant role in approaching this strategic DOD goal. Relevant technology advances include

• Autonomous, unmanned vehicles (AUVs),
• Expert systems,
• Distributed product design,
• "Smart" weapons and platforms,
• "Lights out" factories,
• Real-time space asset command and control (C²),
• SIMNET and other performance-enhancing simulation, and
• Real-time remote sensor, weapons, and platform management.

With the termination of the Cold War, and the resulting reduction in budget, force levels, and manpower, DOD has begun placing greater emphasis on technologies such as those just mentioned in order to leverage remaining resources.

NRL should be aware that two factors will influence the type of investment required for AI and HCI, as follows:

• AI research is more labor intensive than capital intensive, and HCI research is

• more capital intensive than labor intensive.
• AI is dependent upon high-performance computer systems, which fortunately are rapidly decreasing in price and operating costs. Industry competition should continue to drive these costs down.

Industrial interest in AI and HCI can be measured in terms of their use in industrial operations and the extent of external (usually government) funding of industrial programs.

AI practitioners and users would concur that AI has an apparently solid future in academia and a bright, expanding future in applications. Today AI is pervasive in academic institutions, although its applications are primarily in industry and government. Well-known and widespread examples of AI applications include the following: expert systems, multisystem analysis, natural language processing, pattern recognition, robotic control, and computer vision.

HCI is rapidly acquiring all of the attributes of a scientific specialty area of computer science and, according to many experts, will shortly be perceived as such. The hardware resource base for HCI is shared fairly equally between academia and industry.

The skill base for AI and HCI is currently shared fairly equally between academia and industry. Government is primarily an HCI customer rather than a research source. The diversity of the required skill base for AI and HCI makes it difficult for a single research organization to sustain the minimum skill requirements. One attractive option is for NRL to form long-term arrangements with other research groups to pool skill bases in order to meet programmatic needs. This pooling is a cost-effective means for obtaining more current skills than are often available in a government laboratory with a stable professional population and a low attrition rate. Fortunately, the growing success of multimedia networking has essentially negated any justification for "owning" all needed skills at one facility. NRL is encouraged to enter into such strategic alliances via multimedia networking to maintain the necessary diversity in its AI and HCI skill base. It is further encouraged to establish joint projects in as many research or application areas as is feasible. However, the laboratory is cautioned that cooperative networking programs need to be explicitly defined, with formal, agreed-on objectives.

The panel made a number of observations and reached a number of general recommendations regarding AI and HCI:

• NRL's research program in AI and HCI should be goal-oriented, with specific, substantive content, and not simply aimed at maintaining an "awareness" of the AI and HCI fields. The AI and HCI strategic planning process therefore would be more akin to a corporate model for planning than to an academic model.
• In support of the Navy's many unique and varied operational requirements, NRL should support and promote AI as a strong research component. Linking the strategic planning process to known Navy needs for AI and HCI technology, particularly during this period of restructuring and downsizing when operational leverage is paramount, will allow NRL to better assess its research and programmatic strengths alongside those of other defense research centers and laboratories.
• NRL management should manage and support AI as a recognized scientific specialty area within computer science. In doing so, it should be recognized that AI is still immature and is more "science" than engineering in nature.
• NRL should accelerate its HCI programs and support capital investment in HCI as a key research area for a defense laboratory.

• HCI is dependent upon customized electronic devices and components such as high-resolution displays and interactive electronics and often requires teleoperating and video devices. Customized HCI equipment rather than high computer power epitomizes HCI needs. Budgets for both AI and HCI should reflect the need for periodic equipment replacement and modernization due to the rapid obsolescence of supporting equipment.