It occurs to me as an ML engineer that if a list of banned words were to be used to scrutinize proposals for Federal funding, then if everyone used as many as possible, most ML/AI approaches would barf at the low-SNR problem and be unable to find what they were seeking.