If you are evaluating an AI system this quarter, ask one question first.
Who decided the right answers, and what could they see when they decided?
Part 1 of 5. Code, data pipeline and the blind judging tool:
https://t.co/ymXGQLZwoQ
I threw out 22 of my own labels because I had seen the answer first.
Context: I spent August measuring how accurately AI reads real Pakistani customer complaints. 61% on real ones. 95% on test items written in-house.
How that number got made, and why 22 labels had to go?
Part 1 of 5. Code, data pipeline and the blind judging tool:
https://t.co/ymXGQLZwoQ
#ai #founders #telecom #fintech #digitaltransformation
Some items in the judging queue are ones the models already agree on. They are mixed in and look identical to the rest.
They exist to catch the case where every model is confidently wrong about the same thing.
@nihal_farid Thanks for the feedback, I am not poet just write some lines to express my feelings. But I do believe you need to have some emotional sensitivity to write something.
@nihal_farid I often think abt People who were once very close to me but are no longer alive. Whenever I think abt them, the remeberance is very intense, emotional and crushing. Heart tends to sink, lips glued and breath runs out.
@AAsifAfridii Man that’s so sad. The unfortunate reality is this country is not for avg. law abiding citizens like us. It is for the khakis, the Sarkaris, the Elite & even terrorists. Reason & logic don’t impress inhabitants of this country. Many have left this country & I gave up hope too!