Artificial Intelligence

Artificial intelligence (AI) describes methods for creating computational systems that perform tasks associated with learning, reasoning, perception, prediction or decision-making. AI is a broad field rather than a single technology: symbolic methods, statistical learning, neural networks and other approaches address different kinds of problems.

Approaches

AI systems can be built using explicit rules, machine learning from examples, or hybrid architectures. Capabilities depend on the task, the available information and how the system is evaluated; apparent fluency or confident output alone does not establish reliability.

Physical systems

When AI is linked to sensors, robotics hardware and control mechanisms, decisions may influence real-world objects. This raises challenges involving timing, energy limits, environmental uncertainty, and human safety. Physical AI is a useful topic for exploring these connections.

Evaluation and governance

Evaluating AI requires attention to accuracy, robustness, context and possible harms. NIST’s AI Risk Management Framework organizes practices for governing, mapping, measuring and managing AI risks, rather than endorsing any particular AI product.

See also

• Physical AI
• AI Safety
• Embodied AI
• Autonomous AI Agents