Presentation is loading. Please wait.

Presentation is loading. Please wait.

© 2004 Soar Technology, Inc.  July 13, 2015  Slide 1 Thinking… …inside the box Soar Workshop Presentation Presented on 10 June 2004 by Jacob Crossman.

Similar presentations


Presentation on theme: "© 2004 Soar Technology, Inc.  July 13, 2015  Slide 1 Thinking… …inside the box Soar Workshop Presentation Presented on 10 June 2004 by Jacob Crossman."— Presentation transcript:

1 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 1 Thinking… …inside the box Soar Workshop Presentation Presented on 10 June 2004 by Jacob Crossman Large Long Term Declarative Memory: Issues and Ideas

2 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 2 Outline  Purpose: Understand current thinking with regard to long term declarative memory Understand possible “engineering solutions” Understand where current research on the Soar architecture might be applicable  Motivation: three projects at SoarTech CIANC – ontology integration HLSR – type system and compiler “bookkeeping” BINAH – spatio-temporal reasoning  Explore the issues  Propose some solutions  Allow some time for discussion  This is probably a good topic for a breakout session

3 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 3 Issue Summary  In practice, many types of knowledge Are easiest to represent declaratively Are easiest to maintain in declarative form Can be generated and used in large quantities  Examples Ontological and Type information Auto-generated knowledge Temporal knowledge  In Soar, encoding large quantities of declarative knowledge tends to be Time consuming and difficult to get right OR Inefficient  Technical issues: How to store declarative knowledge How to retrieve declarative knowledge How to integrate learning with declarative knowledge

4 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 4 Problem I: Ontology Integration  Description: Integrate knowledge from an ontology (e.g. Standard Upper Merged Ontology) into Soar so the agent can reason over it  Motivation: Make it easier to modify declarative information (e.g. the types of targets an agent may see in a mission) Structure declarative information so that it may be more easily reasoned over  Example Projects: CIANC (this workshop) Related Examples: HLSR, BINAH  Issues: Ontologies can be very large, and can overwhelm agent memory (e.g. 200MB)  Current Solution: Store data in working memory Write operators to reason over the ontology on demand Still a work in progress

5 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 5 Problem II: Compiled Knowledge  Description: Knowledge is encoded in an abstraction higher than Soar code and compiled into Soar code  Motivation: Speed development by Reducing the details a developer must encode Reducing the expertise a developer must have to build agents In some cases, allowing experts to encode knowledge  Example Projects: HLSR (this workshop) Related Examples: TAQL, AGILE, BINAH, HERBAL, Redux  Issues: Ontological/type information can become very large Many high-level language structures require some declarative representation to track their state (e.g. goals) Meta-reasoning structures  Current Solution: Store data in working memory Manage declarative memory context with productions – compiler hides details Still work in progress

6 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 6 Problem III: Temporal Reasoning  Description: memory is continuously updated with knowledge about previous beliefs  Motivation: Allow reasoning over time; in particular, let an agent reason over what it believed and did in the past  Example Projects: BINAH (this workshop) Related Examples: Episodic Memory, AGILE, HLSR  Issues: Chunking seems awkward for this process Declarative memory grows quickly and linearly as the time the agent executes The best way to access this knowledge is not yet clear  Current Solution: Store data in working memory Encode Soar-based “compacting” and “forgetting” algorithms Build language to query this knowledge (e.g. BINAH, HLSR) Still a work in progress

7 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 7 Categorization/Evaluation of Possible Solutions  Theory Computer Science Psychologically Plausible  Efficiency Fast Slow  Memory Size Small Large  Solution Form Long-term solution “Engineering” solution (aka “hack”)  Scalability Limited Unlimited  Architecture Integration Integrated External  Interaction Mechanism Passive Retrieval Active Retrieval  Programmability Fast/Easy Slow/Difficult  Maintainability Easy Difficult  Compatibility Backward Compatible Not Backward Comp.

8 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 8 Typical Soar Technology, Inc. Priorities 1.Programmability  How fast can we build the system?  Result of project time constraints 2.Architecture Integration  External integration is time consuming  Internal solutions require only Soar developer 3.Efficiency  Load time  debug time  development time  Sometimes execution time is important 4.Memory Size  We have agents that are 200MB large!  Typical – 10-20MB 5.Maintainability

9 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 9 Typical Solution: Use Working Memory  Store everything in working memory (aka, treat WM as LT declarative memory) Use elaboration productions to load data OR Use input link  Characteristics Integrated solution Passive retrieval Engineering solution (“Hack”)  Advantages Easy to program and debug Fairly efficient (to a point) Easily maintained Backward Compatible  Disadvantages No compelling theory (in CS or Cognition) Not scalable Large (sometimes very large) memory sizes

10 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 10 Soar Solution: Productions and Data Chunking  “Listen to the architecture” and store all long term memory as productions Encode productions such that declarative knowledge is available based on cues Use data chunking to learn more declarative knowledge  Characteristics Integrated Solution Mixed passive/active retrieval Long-term solution  Advantages Theoretically sound (from Soar standpoint) Efficient Scalable (because of RETE) Backward compatible  Disadvantages Hard to program Difficult to maintain and debug Large memory sizes

11 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 11 Idea I: Integrate with a Database  Store long term declarative knowledge in an external database Use operators to cue retrieval Use I/O links to interact with DB  Characteristics External solution Active retrieval Mixed engineering/long-term solution  Advantages Moderately easy to program Easy to maintain Better managed memory sizes Scalable Backward compatible  Disadvantages Really has nothing to do with psychological or Soar theory Inefficient

12 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 12 Idea II: Use “virtual” working memory  Still store everything in working memory, but modify kernel to create “tiered” memory like an OS Store rarely accessed WM on disk Store often accessed WM in RAM  Characteristics Integrated solution Passive retrieval Engineering solution (“Hack”)  Advantages Easy to program Easy to maintain Better managed memory sizes Scalable Probably backward compatible  Disadvantages Not psychologically plausible Less efficient

13 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 13 Idea IIIa: Engineering Driven LT Declarative Memory  Develop a new LT declarative memory for Soar focusing on engineering issues such as efficiency, accuracy, programmability  Characteristics Internal to architecture Passive/active retrieval (depends on approach) Long-term solution  Advantages (anticipated) Easy to program Easy to maintain Better managed memory sizes Scalable Efficient  Disadvantages Probably not psychologically plausible Might not be backward compatible  Key Impact: Shift Soar further toward an AI architecture vs. a cognitive architecture

14 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 14 Idea IIIb: Cognition Driven LT Declarative Memory  Develop a new LT declarative memory for Soar focusing on issues of cognition such as retrieval timing, interference, and failures (e.g. similar to ACT-R)  Characteristics Internal to architecture Passive/active retrieval (depends on approach) Long-term solution  Advantages (anticipated) Easy to program Easy to maintain Better managed memory sizes Scalable Efficient  Disadvantages Probably psychologically implausible Might not be backward compatible  Key Impact: Shift Soar further toward a cognitive architecture vs. an AI architecture

15 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 15 Conclusions  For Soar Technology, the issue of large declarative memory in Soar Impacts many of our projects Grows more important every year  We believe that this issue has a broad impact on anyone building knowledge-rich agents  We would like to leverage what others are doing in the Soar community, if applicable  We (at SoarTech) will seek solutions Short term: engineering solutions such as integration with a database Long term: search for more permanent integrated solutions – but will the result still be Soar?  Important: If a solution is hard to program or maintain, it won’t get used (at least, not broadly) Therefore, any complex solutions should be abstracted through some higher level representation or process

16 © 2004 Soar Technology, Inc.  July 13, 2015  Slide 16 Questions/Comments


Download ppt "© 2004 Soar Technology, Inc.  July 13, 2015  Slide 1 Thinking… …inside the box Soar Workshop Presentation Presented on 10 June 2004 by Jacob Crossman."

Similar presentations


Ads by Google