Hubert Dreyfus was known as one of the most vocal pessimists about the possibility of artificial intelligence and one of the most strident philosophical critics of the theoretical underpinings of AI research programs in the 1960s and '70s. Though derided at the time, by the 1990s AI researchers had largely come to agree with Dreyfus' critiques of symbolic AI, and had advanced toward new approaches that worked from the assumption of cognition as "embedded and embodied."

The most recent commentary from Dreyfus on AI I could find is an essay from 2007 titled "Why Heideggerean AI Failed and How Fixing It Would Require Making It More Heideggerean." In it, he identifies three then-current AI research programs working in a non-Cartesian / post-symbolic vein: "Rodney Brooks' behaviorist approach at MIT, Phil Agre's pragmatist model, and Walter Freeman's dynamic neural model." He goes on to critique each of these approaches as still relying on one or several of the faulty philosophical assumptions (biological, psychological, epistemological, and ontological) that he had identified in his initial critiques.

I'm curious how the field of AI research has advanced in the past decade plus, and how Dreyfus' criticisms either apply or don't apply to current research. Does Dreyfus' critique imply the inevitable impossibility of "strong AI" tout court, or are there approaches to the problem which are able to avoid making the assumptions he identifies (or demonstrate those assumptions as not in fact being faulty)?