To be honest, I'm not sure if I understand what is actually claimed here (it seems that they trained the model on their own, and claim that the problem is in th

To be honest, I'm not sure if I understand what is actually claimed here (it seems that they trained the model on their own, and claim that the problem is in the dataset?) but isn't the more sensible explanation that human doctors were overdiagnosing them?