> we think one to two out of every 10 of these actions may have been mistakes
May have been a mistake? Reminds me of RTO and the subjective feeling of being more productive in the office. They have the feeling they may have made mistakes and base their new policy on that feeling.
I think what they are saying there is the press release interpretation of experiments showing a false positive rate of 10-20%, with error bars wide enough that stating a percentage gives too many significant figures. But the definition of FP is necessarily fuzzy; if you can perfectly identify them as FP at scale then you have built a better classifier and you no longer have the FP problem. So any statement about FP rates necessarily needs to be couched in uncertainty.
I don't think it's malicious wordsmithing where they are mis-representing the internal data, though I don't have the data to confirm.
May have been a mistake? Reminds me of RTO and the subjective feeling of being more productive in the office. They have the feeling they may have made mistakes and base their new policy on that feeling.