No restraint on creativity

sowa@watson.ibm.com
Message-id: <199111271147.AA08498@venera.isi.edu>
Date: Wed, 27 Nov 91 06:42:30 EST
From: sowa@watson.ibm.com
To: GINSBERG@t.stanford.edu
Cc: INTERLINGUA@isi.edu, KR-ADVISORY@isi.edu, SRKB@isi.edu
Subject: No restraint on creativity

Matt,

When you quoted those two lines from my previous note,

> You can argue all you want about how standards might inhibit certain
> lines of research, but the alternative is even more stultifying.

you omitted the immediately preceding line, which said that unless
we propose something better, "the winner by forfeit will be SQL2."

I intended the phrase "the alternative" to refer to SQL2, which
would inhibit EVERY line of research, not to mention education,
business, engineering, manufacturing, and everything else that
depends on a free interchange of information.

But to be explicit, I don't believe that a logic-based standard
for information interchange would inhibit ANY line of research.
In fact, I liked your recent article in AI Magazine very much;
I didn't interpret it as being against standards, but as quite
a reasonable framework for standards.

Following is the kind of direction that I believe would accommodate
and facilitate EVERY line of research:

 1. At a minimum, the standard should include first-order logic.
    Every version of nonstandard reasoning includes classical FOL
    as a limiting case.  If you were given the complete design data
    for the Boeing 777 stated in FOL, I'm sure that you could use
    it as a basis for any kind of hypothetical or nonclassical
    reasoning that you might want to do.

 2. Since almost every AI system has a type hierarchy, I would like
    to see a typed or sorted version of logic.  If you want an untyped
    system, you could use the undefined type at the top of the hierarchy
    for every variable.

 3. One of my complaints about some of the proposed standards is with
    what they include more than with what they leave out.  Cuts, for
    example, do not belong in the basic core, although they might be
    included in something like Cyc's heuristic level.  But in that
    case, I would want the multiple reasoning engines that Cyc
    supports so that I could pick and choose which version of
    heuristics I want.

 4. Since we are talking about standards for describing systems,
    we need some metalanguage for talking about systems.  Again, I
    would prefer to see a classical first-order structure for that
    metalanguage, but where the individuals in the domain of discourse
    happen to be the elements of some other language.  Two layers of
    first-order logic give you much of the power of second-order logic,
    but without the complexities that arise from freely mixing the
    different kinds of quantifiers.  You could have a first-order
    theory of arithmetic, for example, but where you would have to
    go to the metalanguage to apply the axiom of induction whenever
    you wanted to define a new function.

 5. I very much like McCarthy's idea of treating contexts as first
    class objects.  You could allow the outer context to be a classical
    FOL that happens to talk about things inside some other context,
    which may happen to be propositions or even truth values.  The outer
    context could have classical truth values {T,F}, but it could talk
    about and define an interpreter for another context that happened
    to have truth values {T', U', F'}.  As long as you never identify
    T with T' or F with F', the inner context could be hopelessly
    inconsistent without affecting the integrity of the containing
    context or any other context embedded within it.  That is much
    like a well-designed operating system, where one user could get
    hung up in a loop without affecting the system itself or any
    other users.

 6. If you allow people to define any kind of theory they want, you
    have not restrained anybody's creativity.  But by providing a
    common framework, you could build libraries of plug-compatible
    theories, ontologies, and inference engines.  Of course, people
    might get into trouble if they plugged a new ontology into an old
    theory, but that is no worse that what might happen when they plug
    together the components of any computer or stereo system.  If you
    allow people to express their creativity freely, you have to give
    them the option of getting themselves into trouble -- as long as
    they don't hurt anyone else.

I'm not proposing this framework as the last word on how to design the
standards.  But I'm suggesting it as a way to support simple standards
for simple things, while accommodating almost anything new that may come
along.  I'm sure that there are other frameworks that could do as well
or better, and I'd like to see some discussion of alternatives.

John Sowa