The study suggests a drop in skill over just three months. This is a very short period, especially for clinicians with over 27 years of experience on average. It raises the question of whether a true loss of skill could happen so quickly or there were other factors which the study did not measure

The number of colonoscopies performed nearly doubled after the AI tool was introduced, going from 795 to 1382. It’s possible that this sharp increase in workload, rather than the AI itself, could have led to a lower detection rate. 2/3

It is critically important to thoroughly evaluate AI models, going beyond the usual performance tests (such as AUC). We need to watch out for unexpected "shortcuts" models might take, as they could pop up when we least expect them and can lead to unfair or biased care.

doi.org/10.1136/bmjh...

This ability of AI models creates a significant concern for fair medical decisions, especially since vital signs are fundamental in healthcare. If AI can pick up on these subtle differences linked to ethnicity, it might accidentally make existing health inequalities worse.

This was true, even when we control for things like a person's age, gender, why they were admitted to the hospital, or how serious their illness is.

Thanks very much for the kind words @qmulse.bsky.social !

You should have experience with methods such as Autoencoders, GANs or Diffusion Models. Knowledge of clinical data is an advantage.

This is a unique opportunity to collaborate with the world-leading institutions both in London as well as USA, in one of the top ranked Universities for research.