The author implicitly advocates censorship ("curb online harassment or hate speech") and then transitions into speaking about inherent bias in data sets and algos, but does not mention the risk of bias in censorship algos.
Reminds me of Dave Chapelle's bit on MLK getting a sneaker endorsement deal.
https://youtu.be/zSNcSEWFK8U
