Manually migrating this from here because the answers were unsatisfactory: https://datascience.stackexchange.com/questions/26079/is-it-legal-to-scrape-youtube-videos-for-training-data
Can a machine learning researcher scrape youtube or use youtube videos/snippets of youtube videos to train a model? Would this violate 1) the creator's copyrights, 2) google's distribution/use rights?
This question does not concern datasets Google has freely released, but rather any other publicly available videos.