Read a whitepaper about AI training data bias, now I double check everything

I was skimming through a McKinsey report from last month about how training sets skew results in healthcare AI, and it mentioned a study out of Johns Hopkins where the model misdiagnosed 12% of Black patients due to uneven data. That got me thinking about my own use of AI tools for drafting emails or summarizing meeting notes, they can carry the same blind spots. I've started asking for the source or the reasoning behind any weird output, not just blindly accepting it. Has anyone else caught an AI tool making a questionable call based on something it was trained on?

0 comments

0 Comments

No comments yet

Be the first to share your thoughts on this discussion.