Re: CHAT: programming langs

Gerald Koenig wrote:

> >But then what is the underlying, ultimate generic "language"?
>
> I'm reading my new copy of Wierzbicka's semantics book and she supposes
> a small closed set of less than 100 concepts that are universally found
> in spoken language. I suppose the same set will be found in the
> semantic assembly language of the function set that your computer
> inputs funnel to.  She is also working on a minimalist universal
> grammar, all empirically based. I'm putting the Wierzbickas into
> Nilenga-NGL and translating the tenses into vector tense.
I think that's likely to be a/the right way to go.

> >The two modes I can think of are "imperative" as in Perl/Python/etc,
> >or "declarative" as in Prolog and backward-chaining logic systems.
>
> Perhaps we can describe the languages as more or less deterministic.
> The desiderative would be then much less deterministic, it seems that
> such a language would be capable of writing itself to reach a goal.
Sounds like the old General Problem Solver, subgoals and heuristics.
Also I imagine semi-autonomous agents capable of "consulting" one-another,
which requires an interchange protocol or language. The "shareable ontology"
approach doesn't seem to have worked out.

> Just as "the wish is the father of the deed", such a desiderative
> language would have to be capable of self-programming.  I seem to be
> getting into one of those never never lands of artificial intelligence,
> and making no progress also. Statistics, chaos theory, attractors, huge
> databases, would seem to have to be parts of such capability.
In the usenet newsgroup news:comp.ai.nat-lang for the past 2 years,
a blind man in Hawaii (he surfs) has been discussing (heatedly at times)
a program and a whole linguistic theory for parsing English directly,
and having the computer learn its ontology that way. We have sufficient
hardware capability right now, and lack only the concept of how to use it!
I tried out his earlier version, and it *does* work. The later model does
a lot more; but he's handicapped by a lack of sufficient linguistic study.

> The trend of modeling the computer on its creator continues. Perhaps
> simple life forms will indwell in future computers to serve as
> "motivators" or "wanters". Like dog and master they will function as a
> team. I'd better stop here before even conculturalists start to wonder
> at my sanity.
It was just gettin' good there ...

--
As they say in Tepa: hike waipettu.