
Can AI Be Harmful? A Conversation with MIT’s Dr. Marzyeh Ghassemi
NEJM AI Grand Rounds
The Threat of Inquiry
Radiologists don't often get a patient's self-reported race when they get a chest x-ray. They'll get maybe some indication, 32 year old nail check for fracture but not all of the patient's medical records or demographics. The more obvious it was from the model who was black and white, the worse the disparity was in the models predictions. And so what we think is that sometimes this side information that's being learned by models is really harmful to improving overall clinical accuracy of models.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.