Business
Artificial intelligence systems in healthcare are often praised for their accuracy, speed, and access to vast medical knowledge. But what happens when AI knows the facts—and still gets patient safety wrong?
In this inaugural episode of Surely You Are Joking, Mr. Kamber, we examine why AI models frequently fail at medication safety, not because of missing data, but because of flawed reasoning. From incorrect drug–drug interaction logic to unsafe dosage recommendations, the episode explores how pattern recognition without true clinical understanding can lead to dangerous outcomes.
We unpack the difference between factual recall and clinical reasoning, explain why large language models struggle with causal inference, and discuss how hallucinations, overconfidence, and context loss can turn reliable information into unsafe medical advice. The episode also highlights real-world implications for hospitals, clinicians, regulators, and AI developers.
This conversation sets the tone for the podcast: questioning AI performance where the stakes are highest—and reminding us that intelligence is more than prediction.
Because in healthcare, being almost right is not good enough.