Re: [agi] Soar vs Novamente
Just some quick comments. It appears to me that perhaps the primary topic in question is an ability to generalize or abstract knowledge to varieties of situations. I would say that for the most part Soar is very good at *representing* and *using* composable (and therefore generalized) knowledge representations, but it is not so far Soar's strong suit to *create* such knowledge representations. There has been a bit of research in the past to get Soar to do inductive learning, and those efforts have currently shifted a bit to "stepping outside" the standard Soar model and integrating in capabilities for reinforcement learning and episodic learning. However, these efforts are in early stages. For the most part when we want nice generalized knowledge in Soar (which is often, when we are trying to build robust cognitive models or intelligent agents), we engineer the abstractions and knowledge representations directly into the system.One strength of Soar (in my opinion) is that it encourages "composable" knowledge representations that can rapidly "assemble themselves" (again with the proper hard-coded engineering) into wide varieties of actions or solutions to problems. So for example, rather than having 1000 different schemas for opening different kinds of doors, or one monolithic high-level schema, the typical approach in Soar would be to engineer independently the various small steps that can compose into a variety of door-opening schemas, and then layer on top of those low-level actions a hierarchy of potential situations (or partial situations) in which the various steps would be appropriate to execute. Done "correctly", this can lead to a robust reasoning system that can easily switch its behavior as the environment changes.However, there is a big caveat here. Although I claim (and believe) that Soar encourages the development of such robust models, it does not *require* you to represent your knowledge that way. It is certainly easy to build brittle systems in Soar, containing knowledge that is not abstracted well. An engineer has to do the work of finding the right abstractions, which it sounds to me like where some of the focus is in Novamente. Once you have some reasonable abstractions, though, Soar provides a good engine for representing the knowledge in modular and efficient ways.Randy JonesBen Goertzel [EMAIL PROTECTED] wrote: One of the key ideas underlying the NM design is to fully integratethe top-down (logical problem solving and reasoning) based approachwith the bottom-up (unsupervised, reinforcement-learning-basedstatistical pattern recognition) based approach.SOAR basically lies firmly in the former camp...-- BenOn 7/12/06, Yan King Yin <[EMAIL PROTECTED]> wrote: (From a former Soar researcher) [...] Generally, the bottom-up pattern based systems do better at noisy pattern recognition problems (perception problems like recognizing letters in scanned OCR text or building complex perception-action graphs where the decisions are largely probabilistic like playing backgammon or assigning labels to chemical molecules). Top-down reasoning systems like Soar generally do better at higher level reasoning problems. Selecting the correct formation and movements for a squad of troops when clearing a building, or receiving English instructions from a human operator to guide a robot through a burning building. [...] Doug From what I read, Soar also deals with (or has provisos to deal with) sensory processing, otherwise it wouldn't be the "unified cognitive architecture" as Allen Newell has intended it to be. The difference in emphasis between Novamente on perceptual learning and Soar on top-down reasoning, may be real but ideally it should not be accepted prima facie . IMO these 2 emphases should be integrated seamlessly. YKY To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]---To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]Thank YouJames Ratcliffhttp://falazar.com Do you Yahoo!? Get on board. You're invited to try the new Yahoo! Mail Beta. To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
Re: [agi] Soar vs Novamente
Soar, like other cognitive architectures (such as ACT-R), is not designed to directly deal with domain problems. Instead, it is a high-level platform on which a program can be built for a specific problem. On the contrary, Novamente, like other AGI systems (such as NARS), is designed to directly deal with domain problems. To work well, usually it needs to be trained with domain-specific knowledge, but such a training process is fundamentally different from a programming process. To me, many other differences, such as the role of learning, follow from the above difference between program to work and learn to work. The current issue of AI Magazine (http://www.aaai.org/Library/Magazine/vol27.php#Summer) is highly relevant to this discussion. Especially, the articiles by Langley, Cassimatis, and JonesWray provide good introductions and discussions about cognitive architectures. Pei On 7/13/06, Ben Goertzel [EMAIL PROTECTED] wrote: Thanks, Randy. This is very well put. Yes, one of the key things missing in rule and logic based AI systems like SOAR is the learning of new representations to match new situations and problems. Interestingly, this is also one of the key things missing in evolutionary learning as conventionally implemented. My colleague Moshe Looks has been working on a modified approach to evolutionary learning that involves automatically learning new representations for new problems; it is called MOSES and is being written for integration into Novamente as well as for standalone use. Some information on MOSES is here if you're curious: http://metacog.org/doc.html -- Ben On 7/13/06, James Ratcliff [EMAIL PROTECTED] wrote: Just some quick comments. It appears to me that perhaps the primary topic in question is an ability to generalize or abstract knowledge to varieties of situations. I would say that for the most part Soar is very good at *representing* and *using* composable (and therefore generalized) knowledge representations, but it is not so far Soar's strong suit to *create* such knowledge representations. There has been a bit of research in the past to get Soar to do inductive learning, and those efforts have currently shifted a bit to stepping outside the standard Soar model and integrating in capabilities for reinforcement learning and episodic learning. However, these efforts are in early stages. For the most part when we want nice generalized knowledge in Soar (which is often, when we are trying to build robust cognitive models or intelligent agents), we engineer the abstractions and knowledge representations directly into the system. One strength of Soar (in my opinion) is that it encourages composable knowledge representations that can rapidly assemble themselves (again with the proper hard-coded engineering) into wide varieties of actions or solutions to problems. So for example, rather than having 1000 different schemas for opening different kinds of doors, or one monolithic high-level schema, the typical approach in Soar would be to engineer independently the various small steps that can compose into a variety of door-opening schemas, and then layer on top of those low-level actions a hierarchy of potential situations (or partial situations) in which the various steps would be appropriate to execute. Done correctly, this can lead to a robust reasoning system that can easily switch its behavior as the environment changes. However, there is a big caveat here. Although I claim (and believe) that Soar encourages the development of such robust models, it does not *require* you to represent your knowledge that way. It is certainly easy to build brittle systems in Soar, containing knowledge that is not abstracted well. An engineer has to do the work of finding the right abstractions, which it sounds to me like where some of the focus is in Novamente. Once you have some reasonable abstractions, though, Soar provides a good engine for representing the knowledge in modular and efficient ways. Randy Jones Ben Goertzel [EMAIL PROTECTED] wrote: One of the key ideas underlying the NM design is to fully integrate the top-down (logical problem solving and reasoning) based approach with the bottom-up (unsupervised, reinforcement-learning-based statistical pattern recognition) based approach. SOAR basically lies firmly in the former camp... -- Ben On 7/12/06, Yan King Yin wrote: (From a former Soar researcher) [...] Generally, the bottom-up pattern based systems do better at noisy pattern recognition problems (perception problems like recognizing letters in scanned OCR text or building complex perception-action graphs where the decisions are largely probabilistic like playing backgammon or assigning labels to chemical molecules). Top-down reasoning systems like Soar generally do better at higher level reasoning problems. Selecting the correct formation and movements for a squad of troops when clearing a
Re: [agi] Soar vs Novamente
(From a former Soar researcher) [...] Generally, the bottom-up pattern based systems do better at noisy pattern recognition problems (perception problems like recognizing letters in scanned OCR text or building complex perception-action graphs where the decisions are largely probabilistic like playing backgammon or assigning labels to chemical molecules). Top-down reasoning systems like Soar generally do better at higher level reasoning problems. Selecting the correct formation and movementsfor a squad of troops when clearing a building, or receiving English instructions from a human operator to guide a robot through a burning building. [...] Doug From what I read, Soar also deals with (or has provisosto deal with) sensory processing, otherwise it wouldn't be the unified cognitive architecture as Allen Newell has intended it to be. The difference in emphasis between Novamente on perceptual learning and Soar on top-down reasoning, may be real but ideally it should not be accepted prima facie . IMO these 2 emphasesshould be integrated seamlessly. YKY To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
[agi] Soar vs Novamente
(From a former Soar researcher) I don't have the time to get involved in a big discussion board, but just in case nobody else replies I thought I'd send you a couple of sentences. Soar at it's core is a pretty simple beast. It's a very high performance production rule system with built in support for goal hierarchies, operatorsand learning. This is placed within a strong theory of how to build and organize large complex AI systems. It represents all knowledge symbolically, which seems like a big difference from Novamente which appears to build in probabilistic reasoning at a more primitive level. One of Soar's main strengths is its longevity--something of an existence proof for its value. It has been around for 20+ years now and still has a very active research community associated with it. It's been used in a vast range of different projects and has some very notable successes, such as systems used to control tactical fighter aircraft in large scale military simulations. There's also a company (http://www.soartech.com/) that is largely based around building AI systems using Soar. In evaluating it I'd say Soar's specialty is problems that require integrating large amounts of complex knowledge from multiple sources. If you're just trying to solve one specific problem (e.g. finding a best plan to get from A to B) then a general architecture isn't the best choice. You're better with a tool that does just the one thing you want--like apure planner in that case. But if you're interested in integrating lots of knowledge together Soar is a good choice. I've not used Novamente so I can't say how well it stacks up. From a quick reading it seems like Novamente has perhaps more of a "bottom-up" approach to knowledge and reasoning as they talk about patterns emerging from the environmental data. That's a lot closer to the neural network/connectionist/GA school of thought than Soar which is more of a classic, top-down reasoning system with high level goals decomposed into steadily smaller pieces. Generally, the bottom-up pattern based systems do better at noisy pattern recognition problems (perception problems like recognizing letters in scanned OCR text or building complex perception-action graphs where the decisions are largely probabilistic like playing backgammon or assigning labels to chemical molecules). Top-down reasoning systems like Soar generally do better at higher level reasoning problems. Selecting the correct formation and movementsfor a squad of troops when clearing a building, or receiving English instructions from a human operator to guide a robot through a burning building. I don't know if any of that helps and I may have misplaced Novamente in the scheme of things -- I've just scanned that work briefly. Doug (Former Soar researcher).James Ratcliff [EMAIL PROTECTED] wrote: Yan, I had heard of it, but had yet to read up on it, after breifly reading a bit here, the main pages, and the first tutorial, I am duly impressed with its abilities. Though leary of having to download and work with a large complex package it apepars to be. Have you or anyone else downloaded and played with the application suite, or have any more insights into its working that we may compare contrast it with the Novamente project?Ref Site: http://sitemaker.umich.edu/soarI have also invited a person from Soar to join the discussion.One goal of mine is to produce a very simplistic web interface, similar to the uses of Open Mind Common Sense, that is easy to get in, edit, and possibly use the agent, and add to the knowledge bases, and possibly open it up to a large section of the internet for supervised learning input.James RatcliffYan King Yin [EMAIL PROTECTED] wrote: On 7/12/06, James Ratcliff [EMAIL PROTECTED] wrote: This is essential. If a long term plan would be made only formulated in terms of (very concrete) microlevel concepts there would be a near-infinity of possible plans, and plan descriptions would be enormously long, and would contain a lot of counterfactuals, because a lot of details are not known yet (causing another combinatiry explosion). If you wanted to go to Holland and made a plan like: move leg up, put hand on phone, turn left etc etc Planning would be unfeasible. Instead you make a more abstract plan, like: order ticket, go to airport, take plane, go to hotel. You formulate it on the right level of abstraction. And during the execution of the high level plan(go to Holland) it would cause more concrete plans (go to airport), that would cause more concrete plans(drive in car), and so on until the level of physical body movement is reached (step on brake). Each level of abstraction is tied to a certain time scale. A plan, and a prediction have a certain (natural) life time that is on the time scale of their level of abstraction. One thing I have been working on in these regards is the use of a 'script system' [] Hi James, have you looked at Soar?