Yes, of course. Caliskan et. al (2017) showed this comparing implicit bias in psychological a/b experiments to bias present in language models trained on differ

Yes, of course. Caliskan et. al (2017) showed this comparing implicit bias in psychological a/b experiments to bias present in language models trained on different corpora (e.g., news, twitter, common crawl).