New Book Exposes Risks of Unverified AI in Healthcare

In "AI Snake Oil," Princeton University researchers Sayash Kapoor and Eric Topol examine the risks associated with the use of probabilistic models in healthcare without thorough validation. They argue that flawed predictive analytics, such as those used in Epic's sepsis model and various COVID-19 predictions, can lead to harmful outcomes if decision-makers do not fully understand the underlying data and governance. The authors advocate for careful scrutiny of these models, emphasizing that while predictive analytics has potential benefits, improper application can compromise patient safety.

Opens on LinkedIn