I agree that there is no convincing evidence the Singularity is so near, but I don't agree with his central argument concerning the necessity of software engineering breakthrough requiring accurate models of human cognition. The AI and ML learning successes we've had the last decade aren't based on developing some kind of imperative algorithm for evaluating a specific problem domain. We don't need one for cognition either; we just need a machine that can learn how to learn better. That is really all the singularity is; unbounded recursive self-improvement.