Great article thanks so much … I have read that radiology dataset is lacking in female representation do you think that could have amplified this issue?
I hadn't heard about that, but I'm sure it will make the AI predictions better when it's trained on better data. We're seeing that in coding. I was specifically struck by the design of 'human-in-the-loop' systems in the radiologist use case. When the AI was uncertain it made the humans slower and overall seems to have blunted the advantage of using AI. It shows there is much to learn on how to build systems that leverage AI effectively!
Great article thanks so much … I have read that radiology dataset is lacking in female representation do you think that could have amplified this issue?
Hi Michael,
I hadn't heard about that, but I'm sure it will make the AI predictions better when it's trained on better data. We're seeing that in coding. I was specifically struck by the design of 'human-in-the-loop' systems in the radiologist use case. When the AI was uncertain it made the humans slower and overall seems to have blunted the advantage of using AI. It shows there is much to learn on how to build systems that leverage AI effectively!