@jeff @anildash @simon @amyhoy @ian @sayrer @jwz what would it mean for this tech to “be good”? Its entire *purpose* is to be a bullshit fountain. Like, at minimum, to be able to provide reliable outputs, the training data would need to be editorially flagged as “true” or “false”, not an undifferentiated slurry of Internet Words, which is such a monumental undertaking as to make the whole effort no longer cost-effective.