MIT Technology Review @techreview
The models were enormously costly to create—both environmentally and financially; they were often trained on the toxic and abusive language of the internet. They’d come to dominate research in language AI, elbowing out promising alternatives. — PolitiTweet.org
