“Early AI "doomers" who were prescient about AGI have failed to update their views based on the empirical realities of modern deep learning.”