But who gets to define "human-level performance"?
Because right now, behind that curtain sit very limited benchmark datasets that in no way suffice as a proxy for human performance even if they are often taken at face value by those outside the practitioner community.
@Abebab Surprised not to see @propublica Machine Bias as one that raised a lot of awareness, and then others I *hope* are classic:
Fairness & Abstraction in Sociotechnical Systems
Towards a Critical Race Methodology in Algo Fairness
(realizing these are fairness-focused)
Finally giving in and rejoining Twitter...sorry to all my college friends (hi yall long time no see) bc it's mostly going to be artificial intelligence stuff๐ค apparently my field actually uses this thing?