Scientists warn of AI collapse (Sabine Hossenfelder) [video]

Once AIs can reliably use computers, the AIs can also check directions about using computers to learn the credibility of various posters. It will lead to their being very good at correcting and authoring help documents.

AIs will soon know who is who across many forums and user names. They will learn each user's strengths and weaknesses. This will lead to them being even better at filtering training data.