False. Criticism and optimization overlap highly in the nature of the reasoning but it is possible to construct an optimizer that is not capable of criticism. If you really want one. I'm not sure what exactly Harry's intended human-level practical point is---some applications apply even though the claim is technically false but some may not.