C++ AI Blog

[Arthur T. Murray]

A C++ AI Weblog for implementing artificial intelligence is at
http://mentifex.virtualentity.com/cpp.html -- with C++ source.

Arthur T. Murray

 

[White Wolf]

A C++ AI Weblog for implementing artificial intelligence is at
http://mentifex.virtualentity.com/cpp.html -- with C++ source.

What is on your site does not really look C++. You use a non-existent
header (iomanip.h), you use typedef to create an existing type bool (with
another name, but what for?). Your code is full of unmanaged pointers, no
use of STL data structures (which in itself is not an error, but the use of
_any_ generic element is missing, even one written by yourself) etc. The
code looks like something from 10 years ago. I strongly suggest that before
you start putting an awful lot of manhours into an AI system you learn
enough to do it right. Otherwise you will make something which have have
unreasonably high maintenance.

This post is not meant to be discouragement, but a friendly warning/help to
give you the chance to not go down to a path which is clearly not the best
one.

 

[Arthur T. Murray]

What is on your site does not really look C++.

It was written by an engineer who took a course in C++ a year ago.

You use a non-existent header (iomanip.h),

The "iomanip.h" is supposed to be a header for the manipulation
of input/output. I don't know if the header is included in the
engineer's rather old (early) version of C++, or if he has
written it himself and did not give it to me yesterday
when he handed me the four pages of C++ code that I
proceeded to key into the uploaded "cpp.html" webpage.

you use typedef to create an existing type bool
(with another name, but what for?).

I do not know. I will have to ask the incognito engineer.

Your code is full of unmanaged pointers,
no use of STL data structures (which in itself is not an error,
but the use of _any_ generic element is missing, even one
written by yourself) etc.

Sorry; I don't know any C++ -- he did it as a favor for me

The code looks like something from 10 years ago.

At least it's a start, something for others to react against.
As of twelve hours after uploading, the "cpp.html" page
has gotten around fifty hits in response to this Usenet thread.
(And it is redundantly at three other domains on the Web :=)

I strongly suggest that before you start putting an awful lot
of manhours into an AI system you learn enough to do it right.
Otherwise you will make something which have have
unreasonably high maintenance.

http://mentifex.virtualentity.com/theory5.html is my main work.

Currently (as you may notice from the "See also..." links at
http://mentifex.virtualentity.com/cpp.html the plan is to
set up such "XYZ AI Weblogs" for about twenty "XYZ" programming
languages as a departure point for coding AI in each language.
Even if the very first code has some defects... THE RACE IS ON.

This post is not meant to be discouragement, but a friendly
warning/help to give you the chance to not go down to a path
which is clearly not the best one.

You are one of Nature's gentlemen, and I thank you kindly.
Again thanking "WW aka Attila",

Sincerely,

Arthur T. Murray

 

[White Wolf]

Arthur T. Murray wrote:
[SNIP]

It was written by an engineer who took a course in C++ a year ago.

Well, I hope he did not pay for that course.

The "iomanip.h" is supposed to be a header for the manipulation
of input/output.

Nope. It is a pre-standard header, from an ancient past.

I don't know if the header is included in the
engineer's rather old (early) version of C++, or if he has
written it himself and did not give it to me yesterday
when he handed me the four pages of C++ code that I
proceeded to key into the uploaded "cpp.html" webpage.

He did a usual C++ course, which teaches C++ from 10 years ago.

I do not know. I will have to ask the incognito engineer.

I would rather not ask him anything anymore. At least C++. He is probably
perfect in Java or SmallTalk.

Sorry; I don't know any C++ -- he did it as a favor for me

Khm. This favor stinks. ;-) He does not know much C++ either.

At least it's a start, something for others to react against.

If you want a flame war...

As of twelve hours after uploading, the "cpp.html" page
has gotten around fifty hits in response to this Usenet thread.
(And it is redundantly at three other domains on the Web :=)

Great. How many reactions did you get? Nowadays if people see code like
this (I mean serious C++ programmers) they will remember the URL just not to
use it by mistake.

http://mentifex.virtualentity.com/theory5.html is my main work.

Currently (as you may notice from the "See also..." links at
http://mentifex.virtualentity.com/cpp.html the plan is to
set up such "XYZ AI Weblogs" for about twenty "XYZ" programming
languages as a departure point for coding AI in each language.
Even if the very first code has some defects... THE RACE IS ON.

The code displayed on your page is *not* C++. And you know a gentleman does
not race if he knows in advance that he will win.

You are one of Nature's gentlemen, and I thank you kindly.

Oh. I was joking with gentleman and now you say it... I must be weakening.
;-)

IMHO if you try to search for AI and C++ on the internet (Google is your
friend [TM]) you will most probably find people who not only know AI but
also C++.

(Sidenote: to learn C++ to be an expert might take up to 3 years of
permanent, cutting edge C++ work.)

Anyway, good luck for your quest!

 

[llewelly]

It was written by an engineer who took a course in C++ a year ago.

A year ago!?! Was it taught by a charlatan or an ignoramus? Sorry,
that was harsh, give me a minute to calm down.

The "iomanip.h" is supposed to be a header for the manipulation
of input/output.

[snip]

"iomanip.h" dates from 1990-era C++ . The C++1998 standard mandates
<iomanip> (no .h suffix) Anyone still using iomanip.h is at least
5 years out of date. This and other evidence leave me feeling your
incognito engineer was taught C++ by someone not remotely
qualified. It is sad to hear of courses still so far behind the
times.


[snip]

Currently (as you may notice from the "See also..." links at
http://mentifex.virtualentity.com/cpp.html the plan is to
set up such "XYZ AI Weblogs" for about twenty "XYZ" programming
languages as a departure point for coding AI in each language.
Even if the very first code has some defects... THE RACE IS ON.

[snip]

Please don't pretend this code can be used any kind of fair or honest
language comparision.

 

[Kevin Goodsell]

llewelly wrote:

"iomanip.h" dates from 1990-era C++ . The C++1998 standard mandates
<iomanip> (no .h suffix) Anyone still using iomanip.h is at least
5 years out of date. This and other evidence leave me feeling your
incognito engineer was taught C++ by someone not remotely
qualified. It is sad to hear of courses still so far behind the
times.

Is this formatting really necessary? Personally, I would appreciate it
if you'd set your news client to format messages left-justified like
most usenet messages. It's a distraction to have it indented like that.

-Kevin

 

[David B. Held]

Currently (as you may notice from the "See also..." links at
http://mentifex.virtualentity.com/cpp.html the plan is to
set up such "XYZ AI Weblogs" for about twenty "XYZ"
programming languages as a departure point for coding
AI in each language.
Even if the very first code has some defects... THE RACE
IS ON.

[snip]

Please don't pretend this code can be used any kind of fair
or honest language comparision.

I don't think that's the point. I think Mr. Murray is convinced
that he really understands intelligence and AI well enough to
have conveniently provided a template whereby skilled
programmers can crank out new AIs like sweatshop children
cranking out Nikes. Perhaps he believes that by encouraging
programmers from different languages to "race" towards his
"goal", it will harness the competitive energy of said
programmers, achieving his ultimate aim of producing the
first real intelligent artifact. And all this will be accomplished
without him having learned a single programming language
or paying a single programmer to write a single line of code.

It's a clever idea...if you know a bunch of clueless and naive
programmers. But if you know such coders, they probably
aren't smart enough to write the AI anyway. And if they were,
you'd be wondering why they would do it for free, make it
publicly available, and do it in their spare time.

I don't think it's far from the mark to say that Mr. Murray is
an AI crackpot. Looking at his web site in depth shows
that he has a rather shallow view of AI, specifying things
like making the user press Return after certain inputs, and
leaving obviously difficult tasks like: "Create a sensory
modality for the AI" up to the programmer. If he had any
financial gain from this, I'd say he's the modern-day maker
of Stone Soup. But alas, it seems to me that all he has is
too much free time on his hands.

He seems to think that an AI can be built incrementally,
without really fully understanding how human cognitive
processes work in detail. I think a good 60-70 years of
less-than-spectacular AI projects indicates otherwise.
Of course, none of this really has to do with C++.

Dave

 

[llewelly]

Currently (as you may notice from the "See also..." links at
http://mentifex.virtualentity.com/cpp.html the plan is to
set up such "XYZ AI Weblogs" for about twenty "XYZ"
programming languages as a departure point for coding
AI in each language.
Even if the very first code has some defects... THE RACE
IS ON.

[snip]

Please don't pretend this code can be used any kind of fair
or honest language comparision.

I don't think that's the point.

On second thought, agreed. I saw 'RACE', and
comp.programming.contests, and jumped to conclusions. Call it a
knee-jerk reaction from too many ill-designed and poorly done
langauge comparisons. (Usually done by people who think they have
to prove their favorite langauge is 'better' than C++, by some
ill-chosen definition of 'better' .)

[snip]

He seems to think that an AI can be built incrementally,
without really fully understanding how human cognitive
processes work in detail. I think a good 60-70 years of
less-than-spectacular AI projects indicates otherwise.
Of course, none of this really has to do with C++.

[snip]

Agreed. (Though I'm curious where you got '60-70 years' - the only
computer I know of from 60 years ago was the British Colossus; I
don't think ENIAC was finished until 1947 or so.)

 

[David B. Held]

He seems to think that an AI can be built incrementally,
without really fully understanding how human cognitive
processes work in detail. I think a good 60-70 years of
less-than-spectacular AI projects indicates otherwise.
Of course, none of this really has to do with C++.

[snip]

Agreed. (Though I'm curious where you got '60-70 years' -
the only computer I know of from 60 years ago was the
British Colossus; I don't think ENIAC was finished until
1947 or so.)

I guess the first AI projects were run in people's heads.

Dave

 

[Y2KYZFR1]

A C++ AI Weblog for implementing artificial intelligence is at
http://mentifex.virtualentity.com/cpp.html -- with C++ source.

Arthur T. Murray

that is some pretty heinous code, and you were silly for "re-typing"
something that un-useable anyway, a short quick search of google would
have found at least half a dozen source code to color coded syntax
highlighted html programs.

sorry to here you this person wasted so much of your time.

 

[Arthur T. Murray]

http://mentifex.virtualentity.com/mind4th.html Mind.Forth Robot AI
is gradually being ported into a C++ AI Mind with source online at
http://mentifex.virtualentity.com/aicppsrc.html -- AI in progress.


A.T. Murray

 

[David B. Held]

http://mentifex.virtualentity.com/mind4th.html Mind.Forth
Robot AI is gradually being ported into a C++ AI Mind
with source online at
http://mentifex.virtualentity.com/aicppsrc.html -- AI in
progress.

I don't know who is porting your "AI", but you should
probably at least include a copyright notice on the code,
in case it actually does something. Otherwise, someone
might steal it, claim copyright, and sell it. If it's the same
"engineer" that contributed before, he still hasn't learned
modern C++. And it's pretty obvious that the whole
design is a little lacking, since the interface appears
disorganized and sloppy. That's what happens when you
code first and design later. Also, a better place to develop
the code is not in a weblog, but on a CVS server such
as Sourceforge.

Dave

 

[Tristan Miller]

Greetings.

http://mentifex.virtualentity.com/mind4th.html Mind.Forth Robot AI
is gradually being ported into a C++ AI Mind with source online at
http://mentifex.virtualentity.com/aicppsrc.html -- AI in progress.

This program is not written in C++, and as such will not compile on any
strictly-conforming compiler. You may want to look into fixing this, along
with some of the other more features of the program which, while not
strictly incorrect, are either dangerous or redundant (no header wrappers,
const int function parameters, definining your own boolean type when C++
already provides one, etc.).

 

[jmfbahciv]

Misunderstanding all you see, Tristan Miller wrote on 25 Sep 2003:

This program is not written in C++,
and as such will not compile on any
strictly-conforming compiler.
You may want to look into fixing this,

ATM:
Yesterday I conferred once again with the
Incognito Engineer (yes, he's an old Cogster)
who wrote the quasi-C++ code now online at
http://mentifex.virtualentity.com/17sep03a.html
and meanwhile at the other link above.

To him in his dotage it is indeed C++ code,
because he took a class in C++ a year ago.

I don't care what kind of code it is,
the aging Cogster is a lot smarter than I am,
and so I just want him to learn the Theory of Mind
http://mentifex.virtualentity.com/theory5.html
that is being implemented.

You're into computers, right?
Well, the old Cogster has been
into computers not since Roman days
(say, let's post this historicity
among the antiquated living dinosaurs
of and
stir up some life in that group)
but since as the Romans might say
"antequam tu natus esses a Tua Matre."

Have you ever heard of
John W. Mauchly and J. Presper Eckert?

Well, one of those two, I think it was
Eckert, no, maybe it was Mauchly, called up
my AI C++ confrere, mon fraire, to discuss
with this now Mentificoid (he's mine!)
some software that the Incognito Engineer
had sold through the pages of a magazine,
and now Mauchly or Eckert had bought a copy
and was trying to make it work.

Have you ever had, say, Alexander Graham Bell
call you up about some add-on device that you
sold for use in telephony? Well, that's what
it was like, to get such a phone call from
J. Presper Eckert or John W. Mauchly.

Come here, Miller, I want you. It took me
a year and a half to get the old Cogster
to even take the bait about working on C++ AI.
Now that he is somewhat hooked, I'm not going
to tell him that one Tristan Miller sez
that "[t]his program is not written in C++"
because I don't care if it's written in
TRAC by Calvin Mooers -- it's still AI.

along with some of the other more [?]
features of the program which,
while not strictly incorrect, are either
dangerous or redundant (no header wrappers,
const int function parameters, defining
your own boolean type when C++
already provides one, etc.). [...]

I can give you a list of names of people who did some good
work once in their lives. Then they got distracted and never
produced anything useful again. Is your blurb such a list?

I'm finding that quite a few change serious gears when they
discover they may be mortal.

Now, how about answering the poster's request which was fix
the code so it'll compile? In order to sell your claptrap,
you should be interested in people being able to run it.

/BAH

Subtract a hundred and four for e-mail.

 

[Arthur T. Murray]

Misunderstanding all you see, Tristan Miller wrote on 25 Sep 2003:

This program is not written in C++,
and as such will not compile on any
strictly-conforming compiler.
You may want to look into fixing this,

ATM:
Yesterday I conferred once again with the
Incognito Engineer (yes, he's an old Cogster)
who wrote the quasi-C++ code now online at
http://mentifex.virtualentity.com/17sep03a.html
and meanwhile at the other link above.

To him in his dotage it is indeed C++ code,
because he took a class in C++ a year ago.

I don't care what kind of code it is,
the aging Cogster is a lot smarter than I am,
and so I just want him to learn the Theory of Mind
http://mentifex.virtualentity.com/theory5.html
that is being implemented.

You're into computers, right?
Well, the old Cogster has been
into computers not since Roman days
(say, let's post this historicity
among the antiquated living dinosaurs
of and
stir up some life in that group)
but since as the Romans might say
"antequam tu natus esses a Tua Matre."

Have you ever heard of
John W. Mauchly and J. Presper Eckert?

Well, one of those two, I think it was
Eckert, no, maybe it was Mauchly, called up
my AI C++ confrere, mon fraire, to discuss
with this now Mentificoid (he's mine!)
some software that the Incognito Engineer
had sold through the pages of a magazine,
and now Mauchly or Eckert had bought a copy
and was trying to make it work.

Have you ever had, say, Alexander Graham Bell
call you up about some add-on device that you
sold for use in telephony? Well, that's what
it was like, to get such a phone call from
J. Presper Eckert or John W. Mauchly.

Come here, Miller, I want you. It took me
a year and a half to get the old Cogster
to even take the bait about working on C++ AI.
Now that he is somewhat hooked, I'm not going
to tell him that one Tristan Miller sez
that "[t]his program is not written in C++"
because I don't care if it's written in
TRAC by Calvin Mooers -- it's still AI.

along with some of the other more [?]
features of the program which,
while not strictly incorrect, are either
dangerous or redundant (no header wrappers,
const int function parameters, defining
your own boolean type when C++
already provides one, etc.). [...]

A.T. Murray

 

[Arthur T. Murray]

[...] Now, how about answering the poster's
request which was fix the code so it'll compile?

ATM:
The Incognito Engineer on Social Security is fixing the
http://mentifex.virtualentity.com/aicppsrc.html code
and running it on his admittedly old C++ version.

He (the Cogster) is hoping to obtain work in C++ AI;
I (Mentifex) only want to see AI take over the world.

Yesterday the not-yet-really-old @64 Cogster was
surprised to learn in our discussion that the AI
http://mentifex.virtualentity.com/mind4th.html Mind
stores each incoming word as a fresh new engram in the
http://mentifex.virtualentity.com/audstm.html auditory
short term memory. Being an engineer, he wuz expecting
there to be one central look-up table that held wurdz.
(OK, I will stop that right now.) Us neurotheoreticians
hold a more enlightened Dalai-Lama-esque view in which
we try to organize the C++ (or XYZ) AI software to be
a reasonable mentifacsimile of how the brain operates.

Cogster kept typing in the word "yes" as his test-case
for running the nascent, shall we say, emergent, AI
through its paces, and I said, "No, it is better to
use a three-word sentence of input."

I had to explain to the embedded systems engineer that
since we are trying to create a machine that thinks,
we need to use a complete three-word idea or thought
as our sample input to troubleshoot the AI as we code it.

The single word "yes" as input does not give the baby AI
enough to deal with and respond to. "Cats eat fish" or
some such other three-word input sets up a relationship
among an S-V-O set of subject, verb and object concepts.

Thus enlightened, Cogster and Mentifex withdrew to their
separate mountaintops. Tomorrow there will be mountains
between us; soon we will reconvene and try to fix the AI.

Anyone else who wants to run ahead of us slow-witted ones
is welcome to put better, faster, altius, citius C++ AI code
up on the Web and to invite the world to run away with it.

In order to sell your claptrap, you should be
interested in people being able to run it.

It's not for sale; it's for free.

A.T. Murray

 

[Greg Menke]

I can give you a list of names of people who did some good
work once in their lives. Then they got distracted and never
produced anything useful again. Is your blurb such a list?

I'm finding that quite a few change serious gears when they
discover they may be mortal.

Now, how about answering the poster's request which was fix
the code so it'll compile? In order to sell your claptrap,
you should be interested in people being able to run it.

/BAH

This one is hopeless. ATMurray never says anything fundamentally
useful, just variations on ads for the Mentifex thing. Note, I'm not
using the words him or it, as there is some question about ATMurray
being a human or an AI. For my part I'm undecided.

Gregm

 

[David B. Held]

[...]
Note, I'm not using the words him or it, as there is some
question about ATMurray being a human or an AI. For my
part I'm undecided.

I object to the 'I' part of calling him an AI.

Dave

 

[jmfbahciv]

This one is hopeless. ATMurray never says anything fundamentally
useful, just variations on ads for the Mentifex thing. Note, I'm not
using the words him or it, as there is some question about ATMurray
being a human or an AI. For my part I'm undecided.

I'm not. AI implementations aren't that dumb.

/BAH

Subtract a hundred and four for e-mail.

 

[Jim]

This one is hopeless. ATMurray never says anything fundamentally
useful, just variations on ads for the Mentifex thing. Note, I'm not
using the words him or it, as there is some question about ATMurray
being a human or an AI. For my part I'm undecided.

Gregm

Perhaps just a reminder that prescription medications shouldn't be missed.

Jim