He did give a second talk a few years later explaining how the purely cognitive aspect may be handled by ML, and the total system probably going to end up being

He did give a second talk a few years later explaining how the purely cognitive aspect may be handled by ML, and the total system probably going to end up being a mix. But his fundamental opposition to ML for things related to safety was, IMHO absolutely correct.