Re: [agi] OpenCog Prime complex systems [was MOVETHREAD ... wikibook and roadmap ...
Hi... About OCP and Eliezer ... This is another topic that was bound to come up! OpenCogPrime is the design and approach of myself and a number of my colleagues ... but it's not Eliezer's design or approach Eliezer and I have many points of agreement, many points of disagreement, and many points of ongoing debate and discussion ... and of course, considerable mutual respect for one another as thinkers I'm not going to try to put words in his mouth ... but I will say some fairly obvious things on this theme... 1) for sure, we would both prefer a provably beneficial AGI system. 2) He is more optimistic than I am that a provably beneficial (or provably Friendly, etc.) AGI system is a feasible goal. 3) we both agree that it would be dangerous to allow an AGI system whose ethical nature was poorly understood to achieve a high level of intelligence and/or practical power 4) I tend to be of the opinion that a useful theory of AGI ethics is more likely to come out of a combination of theory and experimentation (experimentation with AGi systems w/ general intelligence below that of an adult human, but much greater than that of existing AI programs), than out of pure armchair theorizing... 5) I am more optimistic than Eliezer about the AGI potential of OpenCogPrime type systems -- Ben G On Sat, Aug 2, 2008 at 2:31 AM, Terren Suydam [EMAIL PROTECTED] wrote: What I don't understand is how Eliezer Yudkowski, whose focus is on Provably Friendly AGI, could be happy with a design that emphasizes emergence as the explanatory bridge that crosses the gap between design and self-awareness. Ben, is this a point of contention between you two or does Eliezer endorse your approach? Specifically, how could the concept of Friendliness be programmed in, in prior fashion, when the relevant structures (self, other) must emerge after the software's been written? Terren --- On *Fri, 8/1/08, David Hart [EMAIL PROTECTED]* wrote: It is intended that correct and efficient learning methodologies will be influenced by emergent behaviors arising from elements of interaction (beginning at the inter-atom level) and tuning (mostly at the MindAgent level), all of which is carefully considered in the OCP design (although not yet explicitly and thoroughly explained in the wikibook). -dave -- *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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
Re: [agi] OpenCog Prime complex systems [was MOVETHREAD ... wikibook and roadmap ...]
David Hart wrote: On 8/2/08, *Richard Loosemore* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: Thus: in my paper there is a quote from a book in which Conway's efforts were described, and it is transparently clear from this quote that the method Conway used was random search: I believe this statement misinterprets the quote and severely underestimates the amount of thought and design inherent in Conway's invention. In my option, the stochastic search methodologies (practiced mainly by his students) can be considred 'tuning/improvement/tweaking' and NOT themselves part of the high-level conceptual design. But, this topic is a subjective interpretation rabbithole that is probably not worth pursuing further. Back on the topic of OpenCog Prime, I had typed up some comments on the 'required methodologies' thread that were since covered by Ben's **interactive learning** comments, but my comments may still be useful as they come from a slightly different perspective (although they require familiarity with OCP terminology found in the wikibook, and I'm sure Ben will chime in to correct or comment if necessary): 'Teaching' [interactive learning] should be included among those words loaded with much future work to be done. 'Empirical studies done on a massive scale' includes teaching, and does not necessarily imply using strictly controlled laboratory conditions. Children learn in their pre-operational and concrete-operational stages using their own flavor of 'methodological empirical studies' which the teaching stages of OCP will attempt to loosely recreate with proto-AGI entities within virtual worlds in a variety of both guided (structured) and free-form (unstructured) sessions. The complex systems issue comes into play when considering the interaction of OCP internal components (expressed in code running in MindAgents) that modify structures of atoms (including maps, which are themselves atoms that encapsulate groups of atoms to store patterns of structure or activity mined from the atomspace) with each other and with the external world. A key point to consider about MindAgents is that the result of their operation is a proxy for the action of atoms-on-atoms. The rules that govern some of these inter-atom interactions are analogous to the rules within cellular automata systems, and are subject to the same general types of manipulations and observable behaviors (e.g. low-level logical rules, various algorithmic manipulations like GA, MOSES, etc, and higher-level transformations, etc.). It is intended that correct and efficient learning methodologies will be influenced by emergent behaviors arising from elements of interaction (beginning at the inter-atom level) and tuning (mostly at the MindAgent level), all of which is carefully considered in the OCP design (although not yet explicitly and thoroughly explained in the wikibook). The complex systems issue does not come into play in only that location. Or rather, there is no basis on which you can say that it only occurs there. More generally, this does not address the questions that I asked. Was it meant to? Richard Loosemore --- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
RE: [agi] OpenCog Prime wikibook and roadmap posted (moderately detailed design for an OpenCog-based thinking machine)
Ben, Thanks for the large amount of work that must have gone into the production of the wikibook. Along with the upcoming PLN book (now scheduled for Sept 26 according to Amazon) and re-reading The Hidden Pattern, there should be enough material for a diligent student to grok your approach. I think it will take some considerable time for anybody to absorb it all, so don't be too discouraged if there isn't a lot of visible banter about issues you think are important; we all come at the Big Questions of AGI from our own peculiar perspectives. Even those of us who want to believe may have difficulty finding sufficient common ground in viewpoints to really understand your ideas in depth, at least for a while. If there's one thing I'd like to see more of sometime soon, it would be more detail on the early stages of your vision of a roadmap, to help focus both analysis and development. Great stuff! --- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
RE: [agi] OpenCog Prime wikibook and roadmap posted (moderately detailed design for an OpenCog-based thinking machine)
I would like to second the thank you. You posted a lot more than I expected and I really appreciate it (and intend to show it by thoroughly reading all of it and absorbing it before commenting). Mark - Original Message - From: Derek Zahn To: agi@v2.listbox.com Sent: Friday, August 01, 2008 3:41 PM Subject: **SPAM** RE: [agi] OpenCog Prime wikibook and roadmap posted (moderately detailed design for an OpenCog-based thinking machine) Ben, Thanks for the large amount of work that must have gone into the production of the wikibook. Along with the upcoming PLN book (now scheduled for Sept 26 according to Amazon) and re-reading The Hidden Pattern, there should be enough material for a diligent student to grok your approach. I think it will take some considerable time for anybody to absorb it all, so don't be too discouraged if there isn't a lot of visible banter about issues you think are important; we all come at the Big Questions of AGI from our own peculiar perspectives. Even those of us who want to believe may have difficulty finding sufficient common ground in viewpoints to really understand your ideas in depth, at least for a while. If there's one thing I'd like to see more of sometime soon, it would be more detail on the early stages of your vision of a roadmap, to help focus both analysis and development. Great stuff! -- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
Re: [agi] OpenCog
Mike Dougherty wrote: On Dec 28, 2007 1:55 PM, Richard Loosemore [EMAIL PROTECTED] wrote: Mike Dougherty wrote: On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. ... considering the proliferation of AGI frameworks, it would appear that any other framework is pretty easy to build, no? ok, I'm being deliberately snarky - but if someone wrote about your own work the way you write about others, I imagine you would become increasingly defensive. You'll have to explain, because I am honestly puzzled as to what you mean here. I am not a published computer scientist. I recognize there are a lot of brains here working at a level beyond my experience. I was only pointing out that using language like just as easy to build to trivialize your system could be confrontational. It may not deliberately offend anyone, either because they are also not concerned about this nuance or they discount your attitude as a matter of course. Well, no: I think for anyone who understood what I was saying, no attitude would have been seen there. None intended, certainly: it was just a simple boring statement of fact, not a trivialization of anyone. I think with slightly different sentence constructions your ideas would be better received and sound less condescending. That's all I was saying on that. I mean framework in a very particular sense (something that is a theory generator but not by itself a theory, and which is complete account of the domain of interest). As such, there are few if any explicit frameworks in AI. Implicit ones, yes, but not explicit. I do not mean framework in the very loose sense of bunch of tools or bunch of mechanisms. hmm... I never considered framework in that context. I thought framework referred to more of a scaffolding to enable work. As such, a scaffolding makes a specific kind of building. Though I can see how it can be general enough to apply the technique to multiple building designs. As for the comment above: because of that problem I mentioned, I have evolved a way to address it, and this approach means that I have to devise a framework that allows an extremely wide variety of Ai systems to be constructed within the framework (this was all explained in my paper). As a result, the framework can encompass Ben's systems as easily as any other. It could even encompass a system built on pure mathematical logic, if need be. I believe I misunderstood your original statement. This clarification makes more sense. Oh, nobody expects it to arise automatically - I just want the system-building process to become more automated and less hand-crafted. Again, I agree this is a good goal - but isn't it akin to optimizing too early in a development process? Sure, there are well-known solutions to certain classes of problem. Building a sloppy implementation to those solutions is foolish when there are existing 'best practice' methods. Is there currently a best practice way to achieve AI? Jeepers, no!! There are narrow solutions to little issues that can be optimized, which arguably cannot be added to each other in any way, let alone integrated into a full AGI, let alone be optimal in a full AGI. I think we are having this discussion because of a confusion about context. All of this is about the particular program of research that I have adopted. Within that context, there is no premature optimization going on: in fact, exactly the opposite. It is the most extreme form of not optimizing too early that you could possibly think of. Let me preemptively agree that we should all continuously strive to implement better practices than we may currently be comfortable with - we should be doing that anyway. (how can we build self-improving systems if we are not examples of such ourselves) My guess is that any system that is generalized enough to apply across design paradigms will lack the granular details required for actual implementation. On the contrary, that is why I have spent (am still spending) such an incredible amount of effort on building the thing. It is entirely possible to envision a cross-paradigm framework. With a different understanding of your use of framework I am less dubious of this position. Give me about $10 million a year in funding for the next three years, and I will deliver that system to your desk on January 1st 2011. Well, I'd love to have the cash on hand to prove you wrong. It would be a nice condition to have for both of us. There is, though, the possibility that a lot of effort could be wasted on yet another AI project that starts out with no clear idea of why it thinks that its approach is any better than anything that has gone before. Given the sheer amount of wasted effort expended over the last fifty years, I would be
Re: [agi] OpenCog
On 12/28/07, Jean-Paul Van Belle [EMAIL PROTECTED] wrote: IMHO more important than working towards contributing clean code would be to *publish the (required) interfaces for the modules as well as give standards for/details on the knowledge representation format*. I am sure that you have those spread over various internal and published documents (indeed, developing a system like Novamente or proposing a framework is impossible without those) but a cut-and-paste of the relevant sections are essential documentation for the framework. Also a concrete example of how a third-party module would slot into this framework would be mightily useful. I am raising this because many would-be AGI developers have to decide on an interface and KR standard even if they develop their own proprietory system - lots of mileage would be gotten from not having to reinvent the wheel. I agree that it would be a nice thing, but it requires people to have similar AGI architectures. Another problem is that it's very hard to get it started, but once it's started it would be easier for people to join since they only have to focus on their own module. YKY - 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=8660244id_secret=80253985-7285bc
Re: [agi] OpenCog
On Dec 28, 2007 4:17 AM, Ed Porter [EMAIL PROTECTED] wrote: Richard, You are entitled to your reservations about OpenCog, but others, like me, are entitled to our enthusiasms about it. You are correct that OpenCog starts with a certain approach, but I think it is an approach that has a lot of promise, and if it has fatal limitations, hopefully OpenCog will help us learn about them, so either the system can be improved, or replaced by a better approach. If you have another approach, I wish you good luck with it. I can't be too enthusiastic about OpenCog yet because I know next to nothing about it, despite all these 'executive' publications and stray papers about Novamente. Let's wait and see. -- Vladimir Nesovmailto:[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=8660244id_secret=79854254-d72e0c
Re: [agi] OpenCog
OpenCog is definitely a positive thing to happen in the AGI scene. It's been all vaporware so far. I wonder what would be the level of participation? Also I think it's going to increase the chance of a safe takeoff, by exposing users and developers gradually to AGI. But we also need to have some security measures. I look forward to seeing it! YKY - 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=8660244id_secret=79861421-6f527c
Re: [agi] OpenCog
Benjamin Goertzel wrote: I wish you much luck with your own approach And, I would imagine that if you create a software framework supporting your own approach in a convenient way, my own currently favored AI approaches will not be conveniently explorable within it. That's the nature of framework-building. Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. Richard Loosemore - 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=8660244id_secret=79873601-00cc5e
Re: [agi] OpenCog
On Dec 28, 2007 5:59 AM, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: OpenCog is definitely a positive thing to happen in the AGI scene. It's been all vaporware so far. Yes, it's all vaporware so far ;-) On the other hand, the code we hope to release as part of OpenCog actually exists, but it's not yet ready for opening-up as some of it needs to be extracted from the overall Novamente code base, and other parts of it need to be cleaned-up in various ways... Much of the reason for yakking about it months in advance of releasing it, was a desire to assess the level of enthusiasm for it. There are a number of enthusiastic potential OpenCog developers on the OpenCog mail list, so in that regard, I feel the response has been enough to merit proceeding with the project... I wonder what would be the level of participation? Time will tell! -- Ben - 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=8660244id_secret=79870666-e314ea
Re: [agi] OpenCog
On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. ... considering the proliferation of AGI frameworks, it would appear that any other framework is pretty easy to build, no? ok, I'm being deliberately snarky - but if someone wrote about your own work the way you write about others, I imagine you would become increasingly defensive. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. I do believe an academic formalism for discussing AGI would be valuable to allow different camps to identify their similarity/difference in approach and implementation. However, I do not believe that AGI will arise automatically from meta-discussion. My guess is that any system that is generalized enough to apply across design paradigms will lack the granular details required for actual implementation. I applaud the effort required to succeed at your task, but it does not seem to me that you are building AGI as much as inventing a lingua franca for AGI builders. I admit in advance that I may be wrong. This is (after all) just a friendly discussion list and nobody's livelihood is being threatened here, right? - 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=8660244id_secret=79882049-5a2bf8
Re: [agi] OpenCog
IMHO more important than working towards contributing clean code would be to *publish the (required) interfaces for the modules as well as give standards for/details on the knowledge representation format*. I am sure that you have those spread over various internal and published documents (indeed, developing a system like Novamente or proposing a framework is impossible without those) but a cut-and-paste of the relevant sections are essential documentation for the framework. Also a concrete example of how a third-party module would slot into this framework would be mightily useful. I am raising this because many would-be AGI developers have to decide on an interface and KR standard even if they develop their own proprietory system - lots of mileage would be gotten from not having to reinvent the wheel. =Jean-Paul -- Research Associate: CITANDA Post-Graduate Section Head Department of Information Systems Phone: (+27)-(0)21-6504256 Fax: (+27)-(0)21-6502280 Office: Leslie Commerce 4.21 On 2007/12/28 at 14:59, in message [EMAIL PROTECTED], Benjamin Goertzel [EMAIL PROTECTED] wrote: On Dec 28, 2007 5:59 AM, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: OpenCog is definitely a positive thing to happen in the AGI scene. It's been all vaporware so far. Yes, it's all vaporware so far ;-) On the other hand, the code we hope to release as part of OpenCog actually exists, but it's not yet ready for opening-up as some of it needs to be extracted from the overall Novamente code base, and other parts of it need to be cleaned-up in various ways... Much of the reason for yakking about it months in advance of releasing it, was a desire to assess the level of enthusiasm for it. There are a number of enthusiastic potential OpenCog developers on the OpenCog mail list, so in that regard, I feel the response has been enough to merit proceeding with the project... I wonder what would be the level of participation? Time will tell! -- Ben - 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/?; - 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=8660244id_secret=79895084-0bd555
Re: [agi] OpenCog
On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Benjamin Goertzel wrote: I wish you much luck with your own approach And, I would imagine that if you create a software framework supporting your own approach in a convenient way, my own currently favored AI approaches will not be conveniently explorable within it. That's the nature of framework-building. Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. I don't believe it is possible to create a framework that both a) is unbiased regarding design type b) makes it easy to construct AGI designs Just as different programming languages are biased toward different types of apps, so with different AGI frameworks... -- Ben - 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=8660244id_secret=79885135-d592af
Re : [agi] OpenCog
http://gbbopen.org/ - Message d'origine De : Benjamin Goertzel [EMAIL PROTECTED] À : agi@v2.listbox.com Envoyé le : Vendredi, 28 Décembre 2007, 15h14mn 10s Objet : Re: [agi] OpenCog On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Benjamin Goertzel wrote: I wish you much luck with your own approach And, I would imagine that if you create a software framework supporting your own approach in a convenient way, my own currently favored AI approaches will not be conveniently explorable within it. That's the nature of framework-building. Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. I don't believe it is possible to create a framework that both a) is unbiased regarding design type b) makes it easy to construct AGI designs Just as different programming languages are biased toward different types of apps, so with different AGI frameworks... -- Ben - 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/?; _ Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr - 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=8660244id_secret=79906306-182ce3
Re: [agi] OpenCog
Benjamin Goertzel wrote: On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Benjamin Goertzel wrote: I wish you much luck with your own approach And, I would imagine that if you create a software framework supporting your own approach in a convenient way, my own currently favored AI approaches will not be conveniently explorable within it. That's the nature of framework-building. Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. I don't believe it is possible to create a framework that both a) is unbiased regarding design type Nobody says unbiased. b) makes it easy to construct AGI designs Then you have not been paying attention :-) (because I know for a fact that I have said this to you in the past ) I am specifically targetting the problem of making it easier. In my environment your Novamente system would be harder to implement than a system that is better suited to my framework, BUT the point of all the effort I am making is that your system would be (e.g.) ten times easier to build than it is now, whereas my type of AGI design would be (e.g.) a thousand times easier to build than it would be if I had to hand craft it using the currently available tools. Either way, it would be easier. Richard Loosemore - 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=8660244id_secret=80022516-3d8694
Re: [agi] OpenCog
Mike Dougherty wrote: On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. ... considering the proliferation of AGI frameworks, it would appear that any other framework is pretty easy to build, no? ok, I'm being deliberately snarky - but if someone wrote about your own work the way you write about others, I imagine you would become increasingly defensive. You'll have to explain, because I am honestly puzzled as to what you mean here. I mean framework in a very particular sense (something that is a theory generator but not by itself a theory, and which is complete account of the domain of interest). As such, there are few if any explicit frameworks in AI. Implicit ones, yes, but not explicit. I do not mean framework in the very loose sense of bunch of tools or bunch of mechanisms. And in my comment to Ben, I said any other in reference to a particular AI system, not referring to frameworks at all. As for the way I write about others' work. I don't understand. I have done a particular body of research in AI/cognitive science, and as a result I have published a paper in which I have explained that there is a very serious problem with the methodological foundations of all current approaches to AI. As a result I am obliged to point out that many things said about AI fall within the scope of that problem. This is not personal nastiness on my part, just a consequence of the research I have done. Should anyone become defensive or offended by that? Not at all. So I am confused. As for the comment above: because of that problem I mentioned, I have evolved a way to address it, and this approach means that I have to devise a framework that allows an extremely wide variety of Ai systems to be constructed within the framework (this was all explained in my paper). As a result, the framework can encompass Ben's systems as easily as any other. It could even encompass a system built on pure mathematical logic, if need be. This is not a particularly dramatic statement. My purpose is to create a description language that allows us to talk about different types of AGI system, and then construct design variations autonmatically. I do believe an academic formalism for discussing AGI would be valuable to allow different camps to identify their similarity/difference in approach and implementation. However, I do not believe that AGI will arise automatically from meta-discussion. Oh, nobody expects it to arise automatically - I just want the system-building process to become more automated and less hand-crafted. My guess is that any system that is generalized enough to apply across design paradigms will lack the granular details required for actual implementation. On the contrary, that is why I have spent (am still spending) such an incredible amount of effort on building the thing. It is entirely possible to envision a cross-paradigm framework. Give me about $10 million a year in funding for the next three years, and I will deliver that system to your desk on January 1st 2011. I applaud the effort required to succeed at your task, but it does not seem to me that you are building AGI as much as inventing a lingua franca for AGI builders. Not really. I don't want a lingua franca as such, I just need the LF as part of the process of addressing the complex systems problem. I admit in advance that I may be wrong. This is (after all) just a friendly discussion list and nobody's livelihood is being threatened here, right? No, especially since few people are being paid full time to work on AGI projects. There is, though, the possibility that a lot of effort could be wasted on yet another AI project that starts out with no clear idea of why it thinks that its approach is any better than anything that has gone before. Given the sheer amount of wasted effort expended over the last fifty years, I would be pretty upset to see it happen yet again. Richard Loosemore - 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=8660244id_secret=80020995-5b8a2d
Re: [agi] OpenCog
On Dec 28, 2007 1:55 PM, Richard Loosemore [EMAIL PROTECTED] wrote: Mike Dougherty wrote: On Dec 28, 2007 8:28 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Actually, that would be a serious miusunderstanding of the framework and development environment that I am building. Your system would be just as easy to build as any other. ... considering the proliferation of AGI frameworks, it would appear that any other framework is pretty easy to build, no? ok, I'm being deliberately snarky - but if someone wrote about your own work the way you write about others, I imagine you would become increasingly defensive. You'll have to explain, because I am honestly puzzled as to what you mean here. I am not a published computer scientist. I recognize there are a lot of brains here working at a level beyond my experience. I was only pointing out that using language like just as easy to build to trivialize your system could be confrontational. It may not deliberately offend anyone, either because they are also not concerned about this nuance or they discount your attitude as a matter of course. I think with slightly different sentence constructions your ideas would be better received and sound less condescending. That's all I was saying on that. I mean framework in a very particular sense (something that is a theory generator but not by itself a theory, and which is complete account of the domain of interest). As such, there are few if any explicit frameworks in AI. Implicit ones, yes, but not explicit. I do not mean framework in the very loose sense of bunch of tools or bunch of mechanisms. hmm... I never considered framework in that context. I thought framework referred to more of a scaffolding to enable work. As such, a scaffolding makes a specific kind of building. Though I can see how it can be general enough to apply the technique to multiple building designs. As for the comment above: because of that problem I mentioned, I have evolved a way to address it, and this approach means that I have to devise a framework that allows an extremely wide variety of Ai systems to be constructed within the framework (this was all explained in my paper). As a result, the framework can encompass Ben's systems as easily as any other. It could even encompass a system built on pure mathematical logic, if need be. I believe I misunderstood your original statement. This clarification makes more sense. Oh, nobody expects it to arise automatically - I just want the system-building process to become more automated and less hand-crafted. Again, I agree this is a good goal - but isn't it akin to optimizing too early in a development process? Sure, there are well-known solutions to certain classes of problem. Building a sloppy implementation to those solutions is foolish when there are existing 'best practice' methods. Is there currently a best practice way to achieve AI? Let me preemptively agree that we should all continuously strive to implement better practices than we may currently be comfortable with - we should be doing that anyway. (how can we build self-improving systems if we are not examples of such ourselves) My guess is that any system that is generalized enough to apply across design paradigms will lack the granular details required for actual implementation. On the contrary, that is why I have spent (am still spending) such an incredible amount of effort on building the thing. It is entirely possible to envision a cross-paradigm framework. With a different understanding of your use of framework I am less dubious of this position. Give me about $10 million a year in funding for the next three years, and I will deliver that system to your desk on January 1st 2011. Well, I'd love to have the cash on hand to prove you wrong. It would be a nice condition to have for both of us. There is, though, the possibility that a lot of effort could be wasted on yet another AI project that starts out with no clear idea of why it thinks that its approach is any better than anything that has gone before. Given the sheer amount of wasted effort expended over the last fifty years, I would be pretty upset to see it happen yet again. Considering the amount of wasted effort in every other sector that I have experience with, I think you should keep your expectations low. Again, I would like to be wrong. - 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=8660244id_secret=80057282-a98eae
RE: [agi] OpenCog
Richard, You are entitled to your reservations about OpenCog, but others, like me, are entitled to our enthusiasms about it. You are correct that OpenCog starts with a certain approach, but I think it is an approach that has a lot of promise, and if it has fatal limitations, hopefully OpenCog will help us learn about them, so either the system can be improved, or replaced by a better approach. If you have another approach, I wish you good luck with it. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Thursday, December 27, 2007 7:19 PM To: agi@v2.listbox.com Subject: [agi] OpenCog Ed Porter wrote: OpenCog: A Software Framework for Integrative Artificial General Intelligence by Dave Hart and Ben Goertzel says Contingent upon funding for OpenCog proceeding as planned, we are targeting 1H08 for our first official code release, to be accompanied by a full complement of documentation, tools, and development support Is there any show of support from people on the AGI and OpenCog lists that might help the funding effort, such as making small contributions or writing emails to any of the major potential contributors, that might help persuade them of the need, importance, and desire in the AGI community for this effort? I am sorry, but I have reservations about the OpenCog project. The problem of building an open-source AI needs a framework-level tool that is specifically designed to allow a wide variety of architectures to be described and expressed. OpenCog, as far as I can see, does not do this, but instead takes a particular assortment of mechanisms as its core, then suggests that people add modules onto this core. This is not a framework-level approach, but a particular-system approach that locks all future work into the limitations of the initial core. For example, I have many, many AGI designs that I need to explore, but as far as I can see, none of them can be implemented at all within the OpenCog system. I would have to rewrite OpenCog completely to get it to meet my needs. Richard Loosemore - 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/?; - 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=8660244id_secret=79755652-91da20attachment: winmail.dat
Re: [agi] OpenCog
Loosemore wrote: I am sorry, but I have reservations about the OpenCog project. The problem of building an open-source AI needs a framework-level tool that is specifically designed to allow a wide variety of architectures to be described and expressed. OpenCog, as far as I can see, does not do this, but instead takes a particular assortment of mechanisms as its core, then suggests that people add modules onto this core. This is not a framework-level approach, but a particular-system approach that locks all future work into the limitations of the initial core. For example, I have many, many AGI designs that I need to explore, but as far as I can see, none of them can be implemented at all within the OpenCog system. I would have to rewrite OpenCog completely to get it to meet my needs. Hi Richard, To be sure, OpenCog is not intended to be equally useful for all possible AGI approaches. To provide something equally useful for all AGI approaches, one would need to make something extremely broad -- basically, one would need to make a highly general-purpose operating-system and/or programming-language, rather than a specific software framework. OpenCog is designed to support a certain family of AGI designs, but is not designed to conveniently support all possible AGI designs. Definitely, there is room in the world for more than one AGI framework. As an example the CCortex platform seems like it may be a good framework within which to build biologically realistic NN based AGI systems (note, this is based on their literature only, I've never tried their system). I wish you much luck with your own approach And, I would imagine that if you create a software framework supporting your own approach in a convenient way, my own currently favored AI approaches will not be conveniently explorable within it. That's the nature of framework-building. -- Ben - 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=8660244id_secret=79828215-b4b8b5