The word "model" now triggers thoughts of AI, but clinicians have used pharmacokinetic models for decades — differential equations with physically meaningful parameters, no training data required. This episode unpacks the fundamental differences between classical mathematical models, machine learning models, and algorithms, using drug dosing as a concrete example. We explore how theory-first PK models (built from known physiology) compare to data-first neural networks, why extrapolation risk differs dramatically between them, and how the FDA regulates each category differently. A practical taxonomy for anyone buying, building, or relying on "models" in 2026.
Episode #114847 — open it directly at myweirdprompts.com/114847