The visionary inventor, technologist, and futurist Ray Kurzweil now works at Google. He is heading a research team trying to crack artificial general intelligence seemingly by building a model of the neocortex. Google offers him resources to do this which are unmatched. Kurzweil recently gave a talk at Google I/O about his thoughts on the matter which you can watch here.
Rays talk doen’t give away much about the tech he is developing so I thought I’d buy his most recent book How To Create A Mind and see how he intends to achieve his goal. Ray did early and ground breaking work in text and speech recognition and the technologies he developed are used these days in things such as Apples SIRI. Rays states in his book that the neocortex is a vast hierarchical pattern recognition system and I don’t think anyone would contradict this. Further to this he sites communication with Henry Markham whos work has shown that there are repeated patterns of connectivity in the brain with assemblies of 12 or so neurons connected in a Lego like way. These assemblies are presumed to be individual pattern recognisers, and a hierarchical system of these would support hierarchies of feature detectors leading all the way to abstract concepts. However, I find the implicit belief that all the brain does is pattern recognition rather naive. Questions such as: how does pattern recognition relate to thought processes, behaviour or creativity? are not even approached.
Whist his book continues to discuss philosophical matters such as ‘what is the similarity between a computer and a brain?’ ,’do we have free will?’, and ‘what is the notion of identity?’, not only does he only cover old ground here but he also does not in any way relate them to his design thesis. His main and concluding thoughts seem to be with regard to the Law of Accelerating Returns, which states that once something becomes an information technology then it is subject to exponential price/performance enhancement. The problem Ray is faced with is that the mechanics of thought are presently unknown and therefore not as yet an information technology and so are not subject to this law. Unfortunately, Ray has not really suggested much on how we might tackle understanding and emulating the mechanics of thought, or perhaps he does know but Google want him to keep it under wraps.
Thank you for your essays, they are very well done. I particularly liked the piece on synchrony.
Re hierarchies and the neocortex, I don’t think everyone is in accord on this. There is a story here and links to some divergent points of view.
We look forward to your excellent posts.