Re: [agi] Webs vs Nets
On Sat, Oct 11, 2008 at 7:38 AM, Mike Tintner [EMAIL PROTECTED]wrote: As I understand the way you guys and AI generally work, you create well-organized spaces which your programs can systematically search for options. Let's call them nets - which have systematic, well-defined and orderly-laid-out connections between nodes. That is simply incorrect ... the connections between nodes/terms/concepts/whatever are chaotic and self-organized and disorderly, within OpenCogPrime, NARS, or any of a load of other AGI systems. And then you have some cognitive processes that try to build order out of the chaos and create links imposing some fragmentary order ... which won't last long unless actively maintained [roughly: as some folks build directories of parts of the Web...] There is a large body of study of the connection statistics of large networks, and some (but less) study of the dynamics of connection stats in large networks. -- Ben G --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Webs vs Nets
Ben, Some questions then. You don't have any spaces or frames as such within your systems? (what terms would you use/prefer here BTW?) Everything is potentially connected to everything else? Perhaps you can give some example from say your pet-in-a-virtual-world (or anything else). It doesn't have a frame say re fetching, or some other activity? How can it connect, as you can connect on the Web, from say the domain of fetching and balls to any other domain ? Like hide-and-seek? Or conversation? (Or, on the Web itself, to planets in a solar system). It won't have ordered hierarchies, say, re animals (...mammals...humans etc? Another feature of the webs vs nets distinction. Webs it seems to me are *multi-domain* of their very nature. .( A domain, for me, consists of a set of elements which behave according to consistent rules - e.g. chess pieces which move in set ways on a board).So webs are composed of diverse and often contradictory domains and rules. Your sex web, for example, will have a whole variety of domains, religious, literary, different moralities, etiquette, fashion, pornographic, fantasy etc offering contradictory rules about whom you can and can't have sex with, and how, and when - and for what reasons Ditto our language webs consist of radically conflicting rules about how we can and can't speak, construct sentences, use words, spell, mix different conventions, accents, tones etc. etc. Do your spaces/domains exist similarly with conflicting rules? You don't need to keep updating them for consistency? Your system can, for example, survive with conflicting rules of logic - Nars-ian and PLN - as your own brain can? I suspect IOW there *are* important distinctions to be drawn explored here. And my first attempt here may be rather like my first attempt at defining programs a long time ago, which failed to distinguish between sequences and structures of instructions - and was then pounced on by AI-ers. On Sat, Oct 11, 2008 at 7:38 AM, Mike Tintner [EMAIL PROTECTED] wrote: As I understand the way you guys and AI generally work, you create well-organized spaces which your programs can systematically search for options. Let's call them nets - which have systematic, well-defined and orderly-laid-out connections between nodes. That is simply incorrect ... the connections between nodes/terms/concepts/whatever are chaotic and self-organized and disorderly, within OpenCogPrime, NARS, or any of a load of other AGI systems. And then you have some cognitive processes that try to build order out of the chaos and create links imposing some fragmentary order ... which won't last long unless actively maintained [roughly: as some folks build directories of parts of the Web...] There is a large body of study of the connection statistics of large networks, and some (but less) study of the dynamics of connection stats in large networks. -- Ben G -- agi | Archives | Modify Your Subscription --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Webs vs Nets PS
I guess the obvious follow up question is when your systems search among options for a response to a situation, they don't search in a systematic way through spaces of options? They can just start anywhere and end up anywhere in the system's web of knowledge - as you can in searching the Web itself? Presumably they must search among well-defined spaces, otherwise how could you have been having this argument about combinatorial explosion with Richard et al? A web, I guess, by definition - (I'm tossing this out as I go along) - can't be systematically searched, and there can be no combinatorial explosion. At worst, you can surf for too long :). --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Webs vs Nets PS
On Sat, Oct 11, 2008 at 9:46 AM, Mike Tintner [EMAIL PROTECTED]wrote: I guess the obvious follow up question is when your systems search among options for a response to a situation, they don't search in a systematic way through spaces of options? They can just start anywhere and end up anywhere in the system's web of knowledge - as you can in searching the Web itself? Presumably they must search among well-defined spaces, otherwise how could you have been having this argument about combinatorial explosion with Richard et al? Some OpenCog/Novamente MindAgents search systematically, some search unsystematically... A web, I guess, by definition - (I'm tossing this out as I go along) - can't be systematically searched, and there can be no combinatorial explosion. At worst, you can surf for too long :). The World Wide Web can certainly be systematically searched... and Web search algorithms are susceptible to combinatorial explosions... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Webs vs Nets
The OpenCog Atomspace --- its knowledge-base of nodes and links --- is totally free-form without any overarching structures imposed by the programmer However, hierarchies or frames can of course exist as structures within this free-form pool of nodes and links In building a particular app using OpenCog, one can opt to build in hierarchies and frames and such (via creating XML files containing appropriate nodes/links and importing them) or one can start from a blank slate and let the whole structure emerge as it will... Ben G On Sat, Oct 11, 2008 at 9:38 AM, Mike Tintner [EMAIL PROTECTED]wrote: Ben, Some questions then. You don't have any spaces or frames as such within your systems? (what terms would you use/prefer here BTW?) Everything is potentially connected to everything else? Perhaps you can give some example from say your pet-in-a-virtual-world (or anything else). It doesn't have a frame say re fetching, or some other activity? How can it connect, as you can connect on the Web, from say the domain of fetching and balls to any other domain ? Like hide-and-seek? Or conversation? (Or, on the Web itself, to planets in a solar system). It won't have ordered hierarchies, say, re animals (...mammals...humans etc? Another feature of the webs vs nets distinction. Webs it seems to me are *multi-domain* of their very nature. .( A domain, for me, consists of a set of elements which behave according to consistent rules - e.g. chess pieces which move in set ways on a board).So webs are composed of diverse and often contradictory domains and rules. Your sex web, for example, will have a whole variety of domains, religious, literary, different moralities, etiquette, fashion, pornographic, fantasy etc offering contradictory rules about whom you can and can't have sex with, and how, and when - and for what reasons Ditto our language webs consist of radically conflicting rules about how we can and can't speak, construct sentences, use words, spell, mix different conventions, accents, tones etc. etc. Do your spaces/domains exist similarly with conflicting rules? You don't need to keep updating them for consistency? Your system can, for example, survive with conflicting rules of logic - Nars-ian and PLN - as your own brain can? I suspect IOW there *are* important distinctions to be drawn explored here. And my first attempt here may be rather like my first attempt at defining programs a long time ago, which failed to distinguish between sequences and structures of instructions - and was then pounced on by AI-ers. On Sat, Oct 11, 2008 at 7:38 AM, Mike Tintner [EMAIL PROTECTED]wrote: As I understand the way you guys and AI generally work, you create well-organized spaces which your programs can systematically search for options. Let's call them nets - which have systematic, well-defined and orderly-laid-out connections between nodes. That is simply incorrect ... the connections between nodes/terms/concepts/whatever are chaotic and self-organized and disorderly, within OpenCogPrime, NARS, or any of a load of other AGI systems. And then you have some cognitive processes that try to build order out of the chaos and create links imposing some fragmentary order ... which won't last long unless actively maintained [roughly: as some folks build directories of parts of the Web...] There is a large body of study of the connection statistics of large networks, and some (but less) study of the dynamics of connection stats in large networks. -- Ben G -- *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] Nothing will ever be attempted if all possible objections must be first overcome - Dr Samuel Johnson --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Webs vs Nets
Ben, Thanks. But you didn't reply to the surely central-to-AGI question of whether this free-form knowledge base is or can be multi-domain - and particularly involve radically conflicting sets of rules about how given objects can behave - a central feature of the human brain and its knowledge base, I would argue. I haven't thought this through, but my first thought is that such a multi-domain structure lends itself v. strongly to the cross-domain thinking that remains a problem for AGI. Ben:The OpenCog Atomspace --- its knowledge-base of nodes and links --- is totally free-form without any overarching structures imposed by the programmer However, hierarchies or frames can of course exist as structures within this free-form pool of nodes and links In building a particular app using OpenCog, one can opt to build in hierarchies and frames and such (via creating XML files containing appropriate nodes/links and importing them) or one can start from a blank slate and let the whole structure emerge as it will... Ben G On Sat, Oct 11, 2008 at 9:38 AM, Mike Tintner [EMAIL PROTECTED] wrote: Ben, Some questions then. You don't have any spaces or frames as such within your systems? (what terms would you use/prefer here BTW?) Everything is potentially connected to everything else? Perhaps you can give some example from say your pet-in-a-virtual-world (or anything else). It doesn't have a frame say re fetching, or some other activity? How can it connect, as you can connect on the Web, from say the domain of fetching and balls to any other domain ? Like hide-and-seek? Or conversation? (Or, on the Web itself, to planets in a solar system). It won't have ordered hierarchies, say, re animals (...mammals...humans etc? Another feature of the webs vs nets distinction. Webs it seems to me are *multi-domain* of their very nature. .( A domain, for me, consists of a set of elements which behave according to consistent rules - e.g. chess pieces which move in set ways on a board).So webs are composed of diverse and often contradictory domains and rules. Your sex web, for example, will have a whole variety of domains, religious, literary, different moralities, etiquette, fashion, pornographic, fantasy etc offering contradictory rules about whom you can and can't have sex with, and how, and when - and for what reasons Ditto our language webs consist of radically conflicting rules about how we can and can't speak, construct sentences, use words, spell, mix different conventions, accents, tones etc. etc. Do your spaces/domains exist similarly with conflicting rules? You don't need to keep updating them for consistency? Your system can, for example, survive with conflicting rules of logic - Nars-ian and PLN - as your own brain can? I suspect IOW there *are* important distinctions to be drawn explored here. And my first attempt here may be rather like my first attempt at defining programs a long time ago, which failed to distinguish between sequences and structures of instructions - and was then pounced on by AI-ers. On Sat, Oct 11, 2008 at 7:38 AM, Mike Tintner [EMAIL PROTECTED] wrote: As I understand the way you guys and AI generally work, you create well-organized spaces which your programs can systematically search for options. Let's call them nets - which have systematic, well-defined and orderly-laid-out connections between nodes. That is simply incorrect ... the connections between nodes/terms/concepts/whatever are chaotic and self-organized and disorderly, within OpenCogPrime, NARS, or any of a load of other AGI systems. And then you have some cognitive processes that try to build order out of the chaos and create links imposing some fragmentary order ... which won't last long unless actively maintained [roughly: as some folks build directories of parts of the Web...] There is a large body of study of the connection statistics of large networks, and some (but less) study of the dynamics of connection stats in large networks. -- Ben G -- agi | Archives | Modify Your Subscription agi | Archives | Modify Your Subscription -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] Nothing will ever be attempted if all possible objections must be first overcome - Dr Samuel Johnson -- agi | Archives | Modify Your Subscription --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription:
Re: [agi] Webs vs Nets
On Sat, Oct 11, 2008 at 12:27 PM, Mike Tintner [EMAIL PROTECTED]wrote: Ben, Thanks. But you didn't reply to the surely central-to-AGI question of whether this free-form knowledge base is or can be multi-domain - and particularly involve radically conflicting sets of rules about how given objects can behave - a central feature of the human brain and its knowledge base, I would argue. Yes, in both NARS and PLN, the knowledge in the network can be multi-domain ... and can contain conflicting rule sets ... there is no requirement of global consistency... Ben G --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com