Large Language Models Answer Medical Questions Accurately, but Can’t Match Clinicians’ Knowledge | Synapse