I did a year long independent study under Minsky my senior year at Harvard in 1969-70. He basically gave me a long reading list and made me promise to read every word of it and never bother him again. I did both. 1969 was the year Minsky supposedly showed that connectionist architectures were a dead end (his paper's was actually narrower than that, but its effect on government funding of connectionism was not). But despite Minsky's supposed condemnation of connectionism, the reading list he gave me included a fair amount about neural nets, and was largely about brain science (which I viewed as connectionist). The most important single paper I read in the reading list was a brief 4-5 page mimeograph of a typewritten article by Minsky on K-Line theory. (I assumed for years that K-Line was Minsky's idea, but Deb Roy of MIT's Media Lab told me in the early 2000's that someone beside Minsky originated it.) K-Line theory is a very simple, but powerful, model of a mind. It assumes the mind constantly receives sequential sensory information, records that sequential information, and as it does so it matches that information against past recorded sequences. As it does, it activates the more closely matching recordings, and generates from them, I assumed through generalization, the probable implications of the currently received information. For example, this let me understand in 1969 how it would actually be rather easy for a massively parallel K-Line machine to generate a sense of intuition, something that a considerable number of people in AI viewed as totally unexplainable for decades after that time. The other most important information I got from Minsky's reading list was just what a powerful, massively parallel, complex supercomputer the human brain was. In 1969 Minsky had been quoted as saying computers would be more intelligent than humans within several years. Because I understood what a supercomputer the brain was, and because my interpretation of K-Line theory required a computer with computational power somewhat approaching that of the brain, I believed then that machines with the intellect of humans would not be made until the power of computers was somewhere very roughly in the teraopp ranges, something that was many years away at that time. But I agreed with Minsky that artificial intelligence was totally doable, once the necessary computing power arrived -- which it very arguably has now.
So, I owe a great intellectual debt to Minsky. On Tue, Jan 26, 2016 at 3:21 AM, Keyvan Mir Mohammad Sadeghi < [email protected]> wrote: > https://www.csail.mit.edu/node/2682 > > -- > Keyvan Mir Mohammad Sadeghi > MSc AI > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/8630185-a57a74e1> | Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
