Re: [unrev-II] The Middle Layer - Information Is What Confuses Us

From: Jack Park (
Date: Fri Jul 07 2000 - 10:30:10 PDT

  • Next message: Frode Hegland: "[unrev-II] Comments on David Gelernters Manifesto"

    Folks, Joe is refering to a sketch I made yesterday at the Unrev meeting at
    SRI. The sketch was a depiction of the three layer architecture. To recall:
    The bottom layer is comprised of the written record of human discourse. Raw
    data, Peirce's possibility space.

    The middle layer is the *lens* to which Joe refers. Its function is to serve
    as what I call a *universal ontology* to which and from which information
    taken from any view or context can reside or be viewed. This is Peirce's
    Actuality space, though I am beginning to learn that I may not have that
    right. The lens is to be comprised of a dense knowledge representation
    layer, one that, in all likelihood, we do not yet know the nature of. This
    layer will probably be one of an evolutionary nature. Howard Liu suggests it
    might start with really simple representations (not unlike those of a
    child), which evolve over time to be more refined, more complex, and perhaps
    more generalized and specialized in order to serve the many contexts

    The upper layer is, in my original words, a kind of API (application
    programming interface) to the OHS, which I originally viewed as an external
    entity. I have called this layer Peirce's Probability Space. Doug clarified
    my view: an OHS is the upper layer, plus the lower 2 layers. The lower 2
    layers comprise the DKR. In reality, there could be one DKR serving one or
    very many upper layers. The upper layer is the user at a workstation
    interacting with the DKR.

    Joe speaks to the middle layer, the lens. The visualization that Mary
    Keeler gave to the group was an image of a single lens. Howard Liu and I
    were talking about a fly's eye, an object with very many tiny lenses, Joe's
    multifaceted lens. We're speaking in metaphors here. To my mind, it
    matters not which kind of lens you choose (single or multi), the metaphor
    still works. The lens puts a kind of focus on what's really *out there* in
    the bottom layer. How you achieve that focus can be implemented many ways.
    Joe brings up that subject.

    One lense, with the top layer being a user-taylored secondary lens. That's
    one approach. The user constructs a personal view through the big lens,
    selecting those objects desired (recall the library search on all books by
    Joe Bag'O'Donuts).

    Multiple lenses, each prefocussed onto a particular point of view, with the
    user just navigating (by parametric search) to the appropriate lens.

    Those are the two candidates that I can think of now. My take on these
    candidates is that I favor the first candidate. Construct the middle layer
    such that it can potentially satisfy any view request. Were I to implement
    the second approach, odds are that I would miss a needed view. I think I am
    saying that the first candidate -- a single lens middle layer -- is general
    enough that it can evolve to satisfy any view requested. Indeed, it's the
    request for views, the user interaction that makes the system grow. What
    could satisfy the biological learning metaphor more soundly?

    To go slightly nonlinear, consider this: I happen to think that the API is
    the key to the whole Unrev II project. Standardize the way you interact
    with a DKR and anyone can build the OHS layer, or the DKR layer in any way
    they wish.

    Thanks for the comments, Joe.
    BTW: I'm going to see if I can get permission to put a gif of that sketch up
    on a web site. Film at 11...

    From: <>

    > I liked your diagram showing the content and 'lens' which forms vews of
    > the content and structure. I think placing the 'lens' only in the DKR
    > caused a question of where functionality resides.
    > This middle layer of providing a view of DKR content actually resides in
    > both the DKR and the OHS. The DKR 'lens" (multiple multifaceted) forms
    > DKR content and structure views while the OHS 'lens' forms presentation
    > views of the content delivered to it by the DKR.
    > Of course the OHS may have added tools to form views of DKR content and
    > structure independent of the DKR 'lens(es)' capability.

    Let Internet Call Manager notify you of phone calls while you're surf!
    Never feel rushed to get off-line again. Includes Caller ID and Web
    voicemail. And only $5 per month. Start your FREE trial today!

    Community email addresses:
      Post message:
      List owner:

    Shortcut URL to this page:

    This archive was generated by hypermail 2b29 : Fri Jul 07 2000 - 10:35:08 PDT