On Tue, Feb 26, 2008 at 6:10 PM, Ben Goertzel <[EMAIL PROTECTED]> wrote: > > Knowing how to carry out inference can itself be procedural knowledge, > > in which case no explicit distinction between the two is required. > > > > -- > > Vladimir Nesov > > Representationally, the same formalisms can of course be used for both > procedural and declarative knowledge. > > The slightly subtler point, however, is that it seems that **given finite > space > and time resources**, it's far better to use specialized > reasoning/learning methods > for handling knowledge that pertains to carrying out coordinated sets of > action > in space and time. > > Thus, "procedure learning" as a separate module from general inference.
Ben, You are talking about optimization, and in this sense different mutually-convertible representations are equivalent to one common representation and only add in performance and tweakability. My take on Pei's classification was that 'declarative' and 'procedural' knowledge refer to different styles of processing and different capabilities. At this point I see 'procedural' knowledge as a useful metaphor to generalize all types of processing in AGI system. It provides an intuitive take on limited resources issue: only limited amount of processing is allowed at each time, so system learns to deal with constraints, learning reasoning procedures that work under these constraints (and together), and spending no runtime on those that don't. It enables the same flexibility with streamed I/O (natural language?) as with internal reasoning processes. 'Hard' (semantic) knowledge corresponds to stable procedures, and weak episodic knowledge, that requires many relevant cues to line up to access, gradually transforms into semantic form if used sufficiently, without need to trace this process explicitly (which Novamente needs to do, so far as I understand based on few available documents). Plus, I like to think about learning as a kind of imitation, and procedural imitation seems more direct. It's "substrate starting to imitate (adapt to) process with which it interacted" as opposed to "a system that observes a process, and then controls inference to reason about doing that kind of thing". -- Vladimir Nesov [EMAIL PROTECTED] ------------------------------------------- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
