Generative AI
Also called: genAI
Models that produce new content ... text, images, or structured output ... rather than only classifying or scoring an input.
Generative models create output rather than selecting a label. In medicine this covers note drafting, message replies, report impressions, patient education material, and synthetic data for training.
The regulatory distinction from classification matters. A model that outputs a probability of hemorrhage is straightforward to evaluate against a reference standard. A model that outputs a paragraph has no single correct answer to score against.
Where This Gets Misread
Because there is no reference standard for generated text, evaluation falls back on human review, and human review of fluent, plausible output is unreliable under time pressure. The absence of a scoring method is not a temporary gap ... it is intrinsic to the output type.