Toward safer ophthalmic artificial intelligence via distributed validation on real-world data.

Given the criticality of validation to the algorithm pipeline, there is an urgent need for developers, machine learning researchers, and end-user clinicians to devise a consensus approach, allowing for the rapid introduction of safe, equitable, and clinically valid machine learning implementations.