AI image models can now reliably report how unsure they are

A new research paper describes a way for AI image recognition models to more accurately and efficiently report how confident they are in their predictions. This could make it safer to use these models in sensitive areas like medicine, where knowing the AI's certainty is critical.

AI models often make predictions without clearly stating how sure they are. This is a problem in fields like medical diagnosis, where a wrong prediction has serious consequences. This new method makes it faster and more accurate for these models to state their uncertainty. This could open up new uses for AI in critical applications that were too risky before.