I agree and would personally find it surprising if they do filter out in-progress work. Not necessarily out of malicious intent but expert level conversations are likely the best training data they have available and is a continually growing source of real world application data. If anything I would expect them to upsample these individuals in training.
I agree and would personally find it surprising if they do filter out in-progress work. Not necessarily out of malicious intent but expert level conversations are likely the best training data they have available and is a continually growing source of real world application data. If anything I would expect them to upsample these individuals in training.