On Fri, Mar 22, 2013 at 4:55 PM, YKY (Yan King Yin, 甄景贤) <[email protected]> wrote: > > This is a somewhat improved version: > https://genifer.googlecode.com/files/logica-universalis%2823-Mar-2013%29.pdf > > I was trying to map all logic formulas to conceptual space, but at the end I > stated there is still one unsolved problem =( > > I'm beginning to want to combine ideas in neural networks with logic, to make > it more efficient. Pure logic seems too powerful but also too slow...
It is tempting to represent knowledge for AI using some kind of formal logic. First, it seems you can express most ideas in it. Second, there are algorithms for deduction, reasoning, and arithmetic that are vastly more efficient on a computer than what goes on in the brain when we solve such problems in our heads. Why would you ever use a 10 petaflop neural network to add numbers at the rate of 1 digit per second with a 5% error rate? Of course this approach has never led to a solution to AI. We are still stuck on the same problems we faced in the 1960's. Language is vastly complex, but humans seem to have little trouble understanding it even when it is full of ungrammatical sentences and missing or misspelled words. Maybe some test cases would help. What is the problem you are trying to solve? What resources do you have, in terms of a knowledge base and computing power? -- -- Matt Mahoney, [email protected] ------------------------------------------- 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
