From Computation to Consciousness [video]

Papers and talks about "consciousness" are usually very bad. This is better than most, but not great. The beginning is a rehash of Philosophy 101, but after a while he gets into neural models of visual processing in the brain which can be implemented in software. That part is quite good. That alone would have been a good talk.

Around 13 minutes in, after mentioning and dismissing linguistic representations, he starts talking about simulation-type representations of the external world. This is progress. Historically, there's been an obsession with linguistic (or worse, predicate-calculus) models. Simulation-type representations, ones which can answer "what happens next" or "what if" questions, But he doesn't develop that concept further.

Then, around 20 minutes in, he gets off on dreams and meditation, and "how is it possible to be conscious without a self", and "free will is a social notion", none of which is helpful. By minute 26, he's off on psychedelic experiences and love. Around minute 30, he's off on dualism, questions like "is the mind part of the universe", and "could we build a mind in Minecraft". Then there's solipsism, and "are we living in a simulation", and whether it matters.

"Preference criteria for suitable encoding", at minute 34, starts to explore computational metaphysics, the nature of reality, and other deep subjects. Shallowly.

There's some good stuff in here, but it's not sorted out between what's actually known, what's conjectured, and what's speculated. If you stop watching at minute 20, you'll have seen all the good stuff. The philosophical excess baggage can be skipped.

I would have liked to see more on simulation-type representations of the real world. That's where "common sense" is probably implemented. (A useful definition of common sense, from a robotics or animal perspective, is that it is about getting through the next ten seconds or so of life without screwing up big-time. As a practical matter, we really need to get a good handle on that to build useful robots for unstructured environments. Google is trying to do this for automatic driving, apparently as a big-data problem.)

Is there a way to easily get rid of the German translations? I am German but unfortunately the translators are not very good and it is - at least for me - harder to listen to this language mix than to the English talk.

What an insightful and refreshing lecture. Thanks for sharing this. The theory about why the mind creates a sense of self, in order to facilitate higher learning, is something I can relate to and helps me understand.

Wait he just claimed the universe is computable because it produces regular data we can encode. Is that necessarily the case? Can't you claim that our senses just produce regular data that we can encode?