GPT
A well-known family of decoder-only (autoregressive) transformers, pre-trained by predicting the next word in a sequence and used to generate text. Fine-tuned to a domain through transfer learning.
A clinic uses a GPT-style model to draft replies to patient portal messages, which a nurse reviews before sending. Because the model writes each reply one word at a time, choosing among likely next words, the same message can produce differently worded drafts on two runs, and a draft can state a follow-up interval or a dose that sounds routine and appears nowhere in the chart. Ask the supplier which parts of the record the model sees for each draft, and train reviewers to check every number and date in a draft against the chart before it goes out.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →