Glossary Terms

Immersive technology is any technology that extends reality to create presence and immersion, using visual, auditory, and other sensory information. Source: XRSI Standard XR Definitions (XR-001) Read more

Immersive training uses virtual and augmented reality to create hands-on learning environments, improving skill acquisition and knowledge retention while reducing the cost and risk associated with conventional training methods. Source: Unity Glossary Read more

An inertial measurement unit (IMU) combines accelerometers, gyroscopes, and sometimes magnetometers to report orientation with very low latency, typically under 2 milliseconds. That speed is what allows head tracking to update the view before a user perceives any delay. Source: Unity Glossary Read more

Inside-out tracking determines a headset’s position using cameras or sensors mounted on the headset itself, so no external equipment needs to be installed in the room. Source: XRSI Standard XR Definitions (XR-001) Read more

Interpretability concerns the meaning of an AI system’s output in the context of its designed purpose, and the extent to which cause and effect can be observed within the system. Where explainability addresses how a system works, interpretability addresses what its output actually means. Sources: FDA Digital Health and AI Glossary; NIST AI Resource Center… Read more

Interpupillary distance (IPD) is the space between a user’s pupils. Matching a headset’s lens spacing to it, usually through an adjustment on the device, is what allows comfortable stereoscopic viewing and a correct sense of scale. Source: Unity Glossary Read more

A language model is an approximate description of the patterns and regularities present in natural language, used to make predictions about language fragments it has not previously seen. Source: NIST AI Resource Center Glossary Read more

A large language model (LLM) is an AI model trained on very large text datasets to learn the relationships between words in natural language. These models use deep-learning methods and training datasets that can run to multiple terabytes. Sources: FDA Digital Health and AI Glossary; NIST AI Resource Center Glossary Read more

In artificial intelligence, learning is the procedure by which a program improves its performance by gaining knowledge. Source: NIST AI Resource Center Glossary Read more

A locked model produces the same output every time the same input is applied and does not change with use, because its parameters and configuration cannot be updated. Source: FDA Digital Health and AI Glossary Read more