…hallucinate – aka confidently making stuff up…
So the AI is acting more like a human. 🤖
I think that this shouldn’t be immediately disqualifying, as long as an actual person is using the outputs as an additional tool in the toolbox and are fully aware of the limitations.
“If an AI model outputs misleading diagnostic information, questions arise as to whether liability should fall on the AI developer for potential shortcomings in training data, the healthcare provider for over-reliance on opaque outputs, or the institution for inadequate oversight,” the authors say.
This seems like a question with a straightforward answer: the healthcare provider and institution should bear all liability for their mistakes. Blaming the AI developer should not be an option.