Re: AI, security, just wondering.

Re: AI, security, just wondering.

[Samuel Mize]

In article <36484B8C.84EEC61D@interact.net.au>,
Dizzy Casablanca  <Dizzy@interact.net.au> wrote:
>  Hello,

Hi!

>  I am wondering whether Ada has any AI applications at all ?  I realise
>that Ada is employed in various embedded scenarios, however - I was
>curious as to whether the language may be employed in some respect with
>AI concepts where autonomous embedded systems are required.

I know that there used to be an annual conference about AI and Ada,
called Aida[1].  I haven't tracked the field for some years.

>  Is Ada not a language which supports AI ?

Not directly.  They're all Turing machines, of course -- any
algorithm can be implemented in any language -- but AI tends to
study things like logic programming, self-adapting programs and
neural networks, which are clumsy to express in a procedural
language.

You'll probably hear that people are doing AI work in C.  Most are
not.  They're doing tool-smithing in C.  The people doing AI work
are those USING the tools built in C.  It's like saying that the
guy who sharpens scalpels is working in the field of brain surgery.

I dunno about your Java/security question.

Sam Mize

[1] It was at an eastern USA University, I think George Mason University.

-- 
Samuel Mize -- smize@imagin.net (home email) -- Team Ada
Fight Spam: see http://www.cauce.org/ \\\ Smert Spamonam

Re: AI, security, just wondering.

[Robert A Duff]

Dizzy Casablanca <Dizzy@interact.net.au> writes:

>   One last issue, - I have noticed that thre are compilers available to
> compile Java bytecode with Ada.  My knowledge is admittedly limited, but
> I am wondering if this doesn't introduce additional security issues...
> can Ada compiled into Java class files do things which java could not
> itself do, and get around the limitations on applets in browsers ?

No.  The Java Virtual Machine is supposed to deal with security in the
proper way.  It doesn't matter where the byte codes come from (Java
compiler, Ada compiler, or some hacker putting in bits by hand).  If
there are security holes, they're not the fault of the compiler that
produced the byte codes, or the language that's being compiled.

>...Or
> does the fact that ada is being used not change the security thing at
> all ?

Yes.

- Bob
-- 
Change robert to bob to get my real email address.  Sorry.

Re: AI, security, just wondering.

[Hans Marqvardsen]

Dizzy Casablanca wrote:
> 
>   Is Ada not a language which supports AI ?  I don't know if AI relies
> on specific data structures and types which are more effectively dealt
> with in other languages.  I understand that one can always
> port/interface one language to another and that Ada supplies various
> solutions (pragmas ?) for this.  So one could, I guess, port Ada to an
> AI system in another language... but can Ada itself be applied to AI
> problems ?
> 

You may wish to check ISBN 0-07-003350-1:
Artificial Intelligence with Ada, Louis Baker, McGraw-Hill 1989.


He discusses, and gives Ada83-code for:

  Backward-chaining expert system shell
  Forward-chaining expert system shell
  Frames database
  Natural language parser
  Augmented Transition Network Parser 

It looks interesting, but it's really not my field.

-- hans 

-- Hans

AI, security, just wondering.

[Dizzy Casablanca]

  Hello,

  I am just getting into Ada.  I started with Java as a first language
but upon discovering  Ada found her a much friendlier companion.

  I am wondering whether Ada has any AI applications at all ?  I realise
that Ada is employed in various embedded scenarios, however - I was
curious as to whether the language may be employed in some respect with
AI concepts where autonomous embedded systems are required.

  Is Ada not a language which supports AI ?  I don't know if AI relies
on specific data structures and types which are more effectively dealt
with in other languages.  I understand that one can always
port/interface one language to another and that Ada supplies various
solutions (pragmas ?) for this.  So one could, I guess, port Ada to an
AI system in another language... but can Ada itself be applied to AI
problems ?

  I am interested in this type of thing, and eventually hope to develop
the savvy to deal with it in source code.

  One last issue, - I have noticed that thre are compilers available to
compile Java bytecode with Ada.  My knowledge is admittedly limited, but
I am wondering if this doesn't introduce additional security issues...
can Ada compiled into Java class files do things which java could not
itself do, and get around the limitations on applets in browsers ?  Or
does the fact that ada is being used not change the security thing at
all ?  I have much to learn, so I am uncertain about these things.

  I am a *sigh* newbie... I guess... in this field at least... but this
is what I am wondering about, and this forum is the place to ask, I
guess... so please let me know.  Thanks

  Graeme Wallace.

dizzy@interact.net.au



$------------------------------------------------------------$
                                Have a Nice day.
 http://www.geocities.com/SouthBeach/Surf/4269/index.html
$------------------------------------------------------------$

Re: AI, security, just wondering.

[Dale Stanbrough]

hm@ddre.dk_nospam wrote:

"You may wish to check ISBN 0-07-003350-1:
 Artificial Intelligence with Ada, Louis Baker, McGraw-Hill 1989.
 
 
 He discusses, and gives Ada83-code for:
 
   Backward-chaining expert system shell
   Forward-chaining expert system shell
   Frames database
   Natural language parser
   Augmented Transition Network Parser 
 
 It looks interesting, but it's really not my field."



Does anyone know if the source code is available outside of the book?

Dale

Re: AI, security, just wondering.

[Jon S Anthony]

>    Backward-chaining expert system shell
>    Forward-chaining expert system shell
>    Frames database
>    Natural language parser
>    Augmented Transition Network Parser 

You can write anything in anything - theoretically.  But you would be
digging yourself a large hole for no good reason doing this sort of
stuff in Ada.  For example, you can write a reasonbly featured and
quite reasonably efficient Prolog engine in CL in something like 4 or
5 pages of (well structured and layed out) code.  An ATN is only about
a page or so (for examples of both see Paul Grahm's On Lisp).  I'm
talking about something that's actually useable and useful - not just
a simple toy.  As a reasonable estimate, doing the same thing in Ada
would be >> 10 times the effort (though maybe not quite 100 times).
What's more it wouldn't be much more efficient - if any.


-- 
Jon Anthony
Synquiry Technologies, Ltd. Belmont, MA 02478, 617.484.3383
"Nightmares - Ha!  The way my life's been going lately,
 Who'd notice?"  -- Londo Mollari

Re: AI, security, just wondering.

[dewarr]

In article <uf1zn88ijz.fsf@synquiry.com>,
  Jon S Anthony <jsa@synquiry.com> wrote:
> You can write anything in anything - theoretically.  But
you would be
> digging yourself a large hole for no good reason doing
this sort of
> stuff in Ada.  For example, you can write a reasonbly
featured and
> quite reasonably efficient Prolog engine in CL in
something like 4 or
> 5 pages of (well structured and layed out) code.  An ATN
is only about
> a page or so (for examples of both see Paul Grahm's On
Lisp).  I'm
> talking about something that's actually useable and
useful - not just
> a simple toy.  As a reasonable estimate, doing the same
thing in Ada
> would be >> 10 times the effort (though maybe not quite
100 times).
> What's more it wouldn't be much more efficient - if any.


This is complete nonsense in my view, it is perfectly
reasonable to program AI problems in Ada. Even if you
did wish to do it by first writing a prolog interpretor
(an extremely dubious proposition), it is absurd to say
that a prolog interpretor written in Ada would be much
more than 50 pages of code.

The efficiency claim is also without substance.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own    

Re: AI, security, just wondering.

[Niklas Holsti]

dewarr@my-dejanews.com wrote:
> 
> In article <uf1zn88ijz.fsf@synquiry.com>,
>   Jon S Anthony <jsa@synquiry.com> wrote:
> > You can write anything in anything - theoretically.  But
> > you would be
> > digging yourself a large hole for no good reason doing
> > this sort of
> > stuff in Ada.  For example, you can write a reasonbly
> > featured and
> > quite reasonably efficient Prolog engine in CL in
    [snip]
> > As a reasonable estimate, doing the same
> > thing in Ada
> > would be >> 10 times the effort (though maybe not quite
> > 100 times).
> > What's more it wouldn't be much more efficient - if any.
> 
> This is complete nonsense in my view, it is perfectly
> reasonable to program AI problems in Ada. Even if you
> did wish to do it by first writing a prolog interpretor
> (an extremely dubious proposition), it is absurd to say
> that a prolog interpretor written in Ada would be much
> more than 50 pages of code.
> 
> The efficiency claim is also without substance.

Regarding the "dubious proposition", I know of one non-toy Prolog
system written in Ada. Around 1985-1988, a research project at
the Department of Computer Science, University of Helsinki,
wrote a fairly innovative Prolog system to support research in
modularisation of logic programs and LP implementation
techniques. The Helsinki Prolog system was implemented in
VAX/VMS Ada and included an interactive programming environment.

I don't have at hand data on its size or speed, but I did not
hear any negative rumours at the time, although I was in fairly
close contact with the principals. I'm sure the whole thing was
more than 50 pages, but of course it contained a lot more than
just the interpreter.

There may be some conference or journal papers on this system,
perhaps by Pekka Kilpel�inen or Esko Ukkonen.

The really interesting issue IMHO would be to explore ways to
integrate logic programming and functional programming with Ada.
After all, if an Ada application just needs to do some Prolog
processing, current O/S and middleware tools make it fairly
easy to create a loose coupling between the Ada program and
some existing Prolog processor. Only a tighter coupling would
motivate integrating the Prolog processor with the Ada program.

Re: AI, security, just wondering.

[Jon S Anthony]

dewarr@my-dejanews.com writes:

> 
> In article <uf1zn88ijz.fsf@synquiry.com>,
>   Jon S Anthony <jsa@synquiry.com> wrote:
> > You can write anything in anything - theoretically.  But
> you would be
> > digging yourself a large hole for no good reason doing
> this sort of
> > stuff in Ada.  For example, you can write a reasonbly
> featured and
> > quite reasonably efficient Prolog engine in CL in
> something like 4 or
> > 5 pages of (well structured and layed out) code.  An ATN
> is only about
> > a page or so (for examples of both see Paul Grahm's On
> Lisp).  I'm
> > talking about something that's actually useable and
> useful - not just
> > a simple toy.  As a reasonable estimate, doing the same
> thing in Ada
> > would be >> 10 times the effort (though maybe not quite
> 100 times).
> > What's more it wouldn't be much more efficient - if any.
> 
> 
> This is complete nonsense in my view, it is perfectly
> reasonable to program AI problems in Ada. Even if you
> did wish to do it by first writing a prolog interpretor
> (an extremely dubious proposition), it is absurd to say
> that a prolog interpretor written in Ada would be much
> more than 50 pages of code.

The Prolog engine was merely an example and no it is not an
interpreter either. And of course my claim is that it would be rather
more than 50 pages of well structured code - not densed up for effect.
And even if it were not "much more", just being more than 10 times the
amount is a lot more than 10 times the effort (program effort is not
linear with program size - as I'm sure you well know).

Further, I have direct evidence here that the "claims" are far from
nonsense - having done a portion of our current work (in a reasonably
related area) first in Ada and then switching to CL.  Now I consider
myself a pretty good Ada programmer, and a pretty good CL programmer
and I did both and the actual measured difference is as I stated.
That's only one example, but it is at least real evidence and not
nonsense.


> The efficiency claim is also without substance.

Not according to my measurements.  This fact was the final reason for
why we switched from Ada to CL.


/Jon

-- 
Jon Anthony
Synquiry Technologies, Ltd. Belmont, MA 02478, 617.484.3383
"Nightmares - Ha!  The way my life's been going lately,
 Who'd notice?"  -- Londo Mollari

Re: AI, security, just wondering.

[Jean-Pierre Rosen]

Niklas Holsti a �crit dans le message <364F4B47.8F20F624@icon.fi>...
>The really interesting issue IMHO would be to explore ways to
>integrate logic programming and functional programming with Ada.
>After all, if an Ada application just needs to do some Prolog
>processing, current O/S and middleware tools make it fairly
>easy to create a loose coupling between the Ada program and
>some existing Prolog processor. Only a tighter coupling would
>motivate integrating the Prolog processor with the Ada program.

A few years ago, CR2A (a French Company) developped Adlog
(not Adalog!), a system to write Ada applications which included
a deductive part. Prolog clauses were described in the form of
aggregates, then passed to a Prolog interpreter, also written in
Ada.

The product disappeared due to a lack of customers. I don't
know if it could be revived if there is an interest.

----------------------------------------------------------------------------
                  J-P. Rosen (Rosen.Adalog@wanadoo.fr)
      Visit Adalog's web site at http://perso.wanadoo.fr/adalog

Re: AI, security, just wondering.

[Art Duncan]

Niklas Holsti <nholsti@icon.fi> wrote:

> ...
>
> The really interesting issue IMHO would be to explore ways to
> integrate logic programming and functional programming with Ada.
> ...

Jean-Pierre Rosen <rosen.adalog@wanadoo.fr> also wrote:

> A few years ago, CR2A (a French Company) developped Adlog
> (not Adalog!), a system to write Ada applications which included
> a deductive part.

At the risk of blowing my own horn, there was a paper in the recent SIGAda 98
Conference describing a way of integrating functional programming, lazy evaluation,
and infinite data structures with Ada.

I hope to have the first version of the library available at my web site 

        http://www.cs.rpi.edu/~duncan

by the end of this week. (The paper and some tutorial notes are there currently.)

There will eventually be three libraries (corresponding to three different implementation
approaches). I will post an announcement when each becomes available.

Regards,

Art Duncan
Adjunct Professor of Computer Science
Rensselaer Polytechnic Institute
Troy, NY 12180

duncan@cs.rpi.edu