As a doctor, I would never use this with a 3rd party LLM provider. Imho this needs to be restricted to a self-hosted and self-administrated instance in my hospital. It doesn't get more sensitive than anamnesis.
At one of the FAANGs I work, it's a relatively smart autocomplete. Saw some examples of autogenerated e2e tests. I think it's quite decent as a language tutor.
But besides that, agree, not many good examples, and it's been a year since gpt-4 release.
