From OntologPSMW

Jump to: navigation, search
[ ]
    (1)
Session CYC
Duration 1.5 hour
Date/Time 02 Feb 2022 17:00 GMT
9:00am PST/12:00pm EST
5:00pm GMT/6:00pm CET
Convener Ravi Sharma
Track Disaster Landscape

Contents

Pandemics and Other Disasters     (2A)

Agenda     (2B)

  • 12:00 EST to 12:45 EST Doug Lenat, CEO CyCorp, on Representation and Reasoning Lessons Learned in Building Cyc     (2B1)
    • The CyCorp architecture and applications provide deep insights into intelligence, reasoning, ontology, knowledge bases and natural languages.     (2B1A)
    • Doug Lenat was awarded the biennial Computers and Thought Award in 1977 for his pioneering work in symbolic machine learning and automated discovery. Dr. Lenat was a professor at CMU and Stanford, and left academia to found the Cyc project in 1984, which is still proceeding at full speed in the form of Cycorp, the 50-person company he is the CEO of. He also holds the distinction of being the only individual to have served on the Scientific Advisory Boards of both Microsoft and Apple.     (2B1B)
  • 12:45 EST to 13:30 EST Discussion     (2B2)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

[11:40] RaviSharma: welcome everyone to an exciting talk from Doug Lenat CEO CyCorp     (2E1)

[11:57] Donna Fritzsche: Hi Everyone!     (2E2)

[12:25] Mike Bennett: Yay for combining symbol and statistical AI!     (2E3)

[12:25] Douglas R. Miles: Question: Has "Explainable AI" funding been interested in CYC?     (2E4)

[12:42] BobbinTeegarden: @Doug: Is the answer 'holonic' and you move your perspective (context) around until you get best fit?     (2E5)

[12:51] BobbinTeegarden: Are you moving to contextual nesting, like lisp?     (2E6)

[12:58] RaviSharma: Doug - By compiling the list of mistakes and publishing them would be a great guide for future ontology developers and users!     (2E7)

[12:58] SteveRay: I distinguish ML applications from symbolic applications by recognizing that ML successes tend to be automating perception while symbolic systems automate cognition. Do you agree?     (2E8)

[12:59] RaviSharma: What are the differences in Watson and CyC and also Search engines?     (2E9)

[13:00] RaviSharma: Does your Gene example imply a close tie with bio-genetic-ontologies?     (2E10)

[13:02] Andrew Dougherty: Are you working on an Intelligent Tutoring System that teaches CYC usage to people using CYC? What about doing Mathematical Knowledge Management in CYC by bridging like Lean and other formal mathematical libraries like Archive of Formal Proofs, etc? Are you taking on remote unpaid interns?     (2E11)

[13:03] Rachuri: Thank you very much for the inspiring presentation (as well as for the question on Judea Pearl). Will the presentation/recording be made available?     (2E12)

[13:08] janet singer: My answer in the chat to the question re the video was yes, it will be on the meeting page https://bit.ly/346GFIz     (2E13)

[13:08] Ram D. Sriram: What do you see is the role of CYC in the Open Knowledge Network initiative at NSF?     (2E14)

[13:08] Mike Bennett: Do you define 'context' as a nexus of concepts and / or instances of concepts, or is it defined in some other way?     (2E15)

[13:09] RaviSharma: Ron Ross Q from chat: I'd like a quick overview of the future mistakes to be made, if not already mentioned.     (2E16)

[13:11] Pierre Chaminade: Synergy, as you described, seams really far from an epistemic engine. Without that, it seams that we will always be limited by the structure of the KB and I doubt that the space of problems solved by those tools will go outside what we can already solve. What is your take on that ?     (2E17)

[13:13] RaviSharma: Janet - thanks for recording location     (2E18)

[13:25] Douglas R. Miles: I think one of a true interesting litmus test milestone for Cyc is to write the rules of chess.. give it any number of hours it needs.. And then end up playing a kick butt game of chess.. in other words could spin up some sort sort of cache.. Since most interesting problems in inference are going to be at least as complex as a chess game     (2E19)

[13:26] John Sowa: Important point about Microsoft's natural language: They were 99% syntactic with very little semantics.     (2E20)

[13:26] John Sowa: I could summarize that point in 2 minutes.     (2E21)

[13:27] Ram D. Sriram: In health care how can you avoid the problems that IBM Watson encountered?     (2E22)

[13:27] Douglas R. Miles: what my litmus test helps define is in ways the cache is able to find the smallest state     (2E23)

[13:28] janet singer: @Ken It would be great to continue this discussion if Doug is available for a future hour, maybe after another one of the scheduled Wednesday sessions?     (2E24)

[13:29] Douglas R. Miles: 2 weeks from now there is a "Panel" open     (2E25)

[13:29] John Sowa: Mike, important point about context: everybody who tried to define context failed.     (2E26)

[13:30] Gary BC: We are out of time but it would have been interesting in light of the right vs left brain frame to ask Doug about the NELL experience which starts with KB building using some forms of "ML".     (2E27)

[13:31] Mike Bennett: Thank John. I think it can be done - we've been looking at this at OMG. Was keen to hear how Doug formally defines Context.     (2E28)

[13:32] ToddSchneider: Doug, natural language processing is important, though costly. Perhaps the injection of context into the processing will produce better results.     (2E29)

[13:33] John Sowa: Nobody who has ever defined context has succeeded.     (2E30)

[13:33] Mike Bennett: Can we ask more about Context at that future session? Seems to impinge on a lot of things we've considered today.     (2E31)

[13:33] Pierre Chaminade: @dmiles, are you up for a call after this ?     (2E32)

[13:34] SteveRay: Nice talk and discussion. Thanks!     (2E33)

[13:34] ToddSchneider: natural language processing is important only if communication with humans is needed.     (2E34)

[13:34] BobbinTeegarden: @John as close as we get at OMG is using who what when where why and how... it seems to point us in the right direction ... would use your 'ist'...     (2E35)

[13:34] Gary BC: Chris Mungall and Justin Reese are scheduled for next week.     (2E36)

[13:34] AlexShkotin: Thank you.     (2E37)

[13:34] John Sowa: There is one possible general definition of context: whatever is necessary to understand a sentence.     (2E38)

[13:35] Pierre Chaminade: Good bye everyone, that was a great experience !     (2E39)

[13:35] Mike Bennett: That's not really a definition though, that's a problem statement (scope).     (2E40)

[13:35] RaviSharma: john we will address your Comments next time,     (2E41)

[13:36] BobbinTeegarden: @John more on that...?     (2E42)

[13:36] RaviSharma: thanks a lot everyone.     (2E43)

[13:36] BobbinTeegarden: @John ... a 'microcontext'?     (2E44)

[13:36] John Sowa: Mike, that is what I said: it's impossible to give a definition of context that is independent of the context.     (2E45)

[13:37] John Sowa: I'll send a note about context to Ontolog Forum.     (2E46)

[13:38] BobbinTeegarden: @John ouroboros...     (2E47)

[13:38] RaviSharma: John but nothing can be meaningfully expressed without context- agree.     (2E48)

[13:38] BobbinTeegarden: @John ... or Lispian recursion?     (2E49)

[13:39] RaviSharma: Thanks John that would be valuable.     (2E50)

[13:40] Mike Bennett: @John that's an interesting assertion. I'll take that back to our little group and give it some more thought.     (2E51)

[13:40] ToddSchneider: 'Context' would appear to behave in a self-referential manner.     (2E52)

[13:41] RaviSharma: context creates environment and creates the base on which to further any activity- dialog or communication, especially among communicating parties!     (2E53)

[13:41] BobbinTeegarden: @Todd self-referential ... holonic?     (2E54)

[13:41] Mike Bennett: I remain to be convinced - pragmatically I think it can be simplified to 2 levels but that's just engineering pragmatics. I wish we'd had time to learn how Cc defines context since Doug said they had a formal definition of it.     (2E55)

[13:42] RaviSharma: Also one has to understand or identify if context changes during a discourse or engagement     (2E56)

[13:42] ToddSchneider: Bobbin, possibly. I'm not aware of research into 'self-referentiality'.     (2E57)

[13:42] RaviSharma: Context is like mini situational awareness.     (2E58)

[13:44] ToddSchneider: Ravi, 'awareness' of a situation is different from the situation or context of the situation one may be aware of.     (2E59)

[13:45] Andrew Dougherty: I don't know if this is known, but Contexts can inherit axioms via the genlMt(Mt1,Mt2), ist(P) means P follows, istAsserted(P) means it is asserted in the context. Constants can bear two different sets of assertions in different contexts. There are dimensions to contexts, such as temporal.     (2E60)

[13:45] RaviSharma: Bobbin I would say it has to be correlated with dialog parties otherwise it is a monolog and to be understood even in self-reference on has to define at least time space topic etc at minimum? Epoch is what I call it!     (2E61)

[13:45] BobbinTeegarden: Doesn't perspective scope around to different contexts...?     (2E62)

[13:47] BobbinTeegarden: @Ravi if a dialogue is a perspective/context, yes..     (2E63)

[13:47] RaviSharma: Todd - context can help you be almost aware of what is going on or about in a situation or meeting or presentation or book?     (2E64)

[13:47] Andrew Dougherty: *ist(C,P), istAsserted(C,P).     (2E65)

[13:48] BobbinTeegarden: *ist -- yes, can we get JohnSowa to talk more about ist (again) next time?     (2E66)

[13:48] RaviSharma: I was giving examples of context for What? a dialog bet people is an example in which context is helpful. In a drama the anchor explains context among different sections?     (2E67)

[13:49] BobbinTeegarden: How do you model context?!     (2E68)

[13:50] BobbinTeegarden: Isn't context multidimensional?     (2E69)

[13:50] RaviSharma: If I write a book, preface is supposed to provide context?     (2E70)

[13:51] RaviSharma: Yes Bobbin it is a great comment thanks     (2E71)

[13:53] RaviSharma: Doug - if you want to see these Qs, we can send you chat transcript or you can see it on meeting page?     (2E72)

[13:54] ToddSchneider: The common definitions of 'context' usually make reference to an ability to understand (presumably something in that context). Perhaps it may be useful to focus on 'context' with respect to 'understanding'? Or narrow it down to 'how is context related to communication'?     (2E73)

[13:55] RaviSharma: For the next interaction with Doug Lenat, I would not only like to learn how different users would be able to learn how to use CyC but also examples like healthcare or supply chain we could not get to today?     (2E74)

[13:55] RaviSharma: how much of LISP one has to know for example?     (2E75)

[13:57] janet singer: Re Johns context as what is necessary to understand a sentence, Winograd and Flores emphasized dialogue that articulates the basis of a representation and adjusts as needed to a new situation. To understand is to make a commitment to do this reinterpretation     (2E76)

[13:57] BobbinTeegarden: @Ravi I would think the important thing to know about LISP is that it's infinitely recursive... turtles all the way down, and UP     (2E77)

[13:58] RaviSharma: Todd if we do not specify any context in the beginning or even situation or publication, it will take longer for target to understand the background of event or dialog etc! Hence implication is some kind of context before we begin anything?     (2E78)

[13:58] BobbinTeegarden: @Janet yes! and contexts that morph are even harder.     (2E79)

[14:00] RaviSharma: Yes I had a course 50 years ago!     (2E80)

[14:01] RaviSharma: Janet - you explained it nicely and scholarly as well.     (2E81)

[14:03] RaviSharma: Bobbin - my course Prof had developed LISP - name Harold Macintosh, Florida 1963-65     (2E82)

[14:03] janet singer: The sense of social action commitment that could span morphing contexts seems completely at odds with formal ontological commitment     (2E83)

[14:04] RaviSharma: Bobbin - in many meetings I have said that ontology has to be extended to include uncertainties, dynamics, strengths and types of relationships etc.     (2E84)

[14:06] Mike Bennett: This kind of stuff is all covered under pragmatics,which considers the interpretation (semiotics) of a symbol in a given context.     (2E85)

[14:06] RaviSharma: Janet - As the history or topics evolve, contexts change! Change is the rule rather than exception, some models we follow are convenient as static!     (2E86)

[14:06] janet singer: One could picture a system that would include the abducers Doug mentioned working together with the human reasoner.     (2E87)

[14:09] janet singer: @Mike, Yes. And both pragmatics and semiotics have gotten inadequate attention in the focus on semantics     (2E88)

[14:09] RaviSharma: My favorite context (as function) is a tuple and one of the parameters is time!     (2E89)

[14:10] RaviSharma: Mike Bennett - thanks for semiotics!     (2E90)

[14:10] ToddSchneider: Ravi, humans, at least those that are not unconscious, and most organisms have some recognition or (sensory) awareness of their surroundings. Such awareness is implicit.     (2E91)

[14:11] RaviSharma: how are these related in epistemology     (2E92)

[14:12] RaviSharma: Todd - yes even if I am cognitive, I need to set my expectation in say reading a book, as to what it is about? context!     (2E93)

[14:19] janet singer: Charles Morris suggested the three perspectives on rules within semiotics as syntactics, semantics, and pragmatics in his Foundations of a Theory of Signs. But he emphasized they are not strictly separable     (2E94)

[14:20] janet singer: The rules of each kind provide context for the others     (2E95)

[15:04] RaviSharma: Janet - thanks     (2E96)

Resources     (2F)

Previous Meetings     (2G)


Next Meetings     (2H)

















































































































































































































































































































































































































This is a cached copy of the requested page, and may not be up to date.

Sorry! This site is experiencing technical difficulties.

Try waiting a few minutes and reloading.

(Can't contact the database server: Can't connect to local MySQL server through socket '/var/run/mysqld/mysqld.sock' (2) (localhost))


You can try searching via Google in the meantime.
Note that their indexes of our content may be out of date.