--- "YKY (Yan King Yin)" <[EMAIL PROTECTED]> wrote: > I think AGI will need some innovative algorithms. For the logic-based > paradigm, we'd need things like: > > 1. merging probabilistic logic and fuzzy logic -- never been done before, > well.. except by Ben and Pei ;) > 2. belief revision algorithms -- current ones are unfeasible for large > KBs. The exact solution is known to be strictly harder than NP > 3. abductive reasoning, inductive learning, etc, are also very inefficient > currently > > True, we can build an AGI using purely existing algorithms, but that would > be a low-quality AGI with a lot of inadequacies.
I doubt this is where the bottleneck lies. KR approaches such as Cyc, NARS, and your proposal are highly abstract representations of human knowledge and thought; that tiny subset of what the brain actually does that can be implemented efficiently on a computer. It is an extension of the idea that humans can add numbers or play chess, so let's model it and do it better on a machine. It has nothing to do with AGI. Here is the problem you need to solve: write a program that translates natural language into Cycl, Narsese, or whatever language you will program your system in. I think you will find that you need to solve AI as a subproblem first, and that solution will obsolete the problem you were trying to solve in the first place. -- Matt Mahoney, [EMAIL PROTECTED] ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e