MindForth Programming Journal (MFPJ) 2016 August 24

[menti...@gmail.com]

In agi00026.F we would like to figure out why the

http://ai.neocities.org/forthagi.txt

Robot Forth AGI does not accept the final word of input if
we end the input immediately with a 13=CR carriage return.
We are having to enter one final 32=SPACE at the end
of our input in order to get the final word accepted.
We may need "13=CR" as a delimiter when we
implement the ReJuvenate module

http://mind.sourceforge.net/rejuve.html

so that the oldest recorded thought does not get chopped up,
but rather is processed in full.

In AudMem and in AudRecog we have tried to establish parity
between 32=SPACE and 13=CR, that is, treat them equally,
almost as if they were the same character, since they both
effectively mean the end of an incoming word. However, near
the start of AudInput we notice that a test for 13=CR stops
AudInput immediately and returns program-flow to the
calling module. We may need to either remove the 13=CR test
or place it somewhere else.

As we insert diagnostic messages, we learn that AudRecog is
indeed recognizing an old concept immediately followed by a
carriage return (13=CR), but somehow NewConcept is being
called instead of OldConcept. We finally get the AGI to accept
13=CR on an equal basis with 32=SPACE, when in AudMem
we wait for either CR or SPACE before a final call to AudRecog.

Mentifex (Arthur)
--
https://groups.google.com/d/msg/comp.lang.forth/QJEREP0_kLM/HjQ78jR5eYAJ
http://github.com/PriorArt/AGI/wiki/MindGrid
http://www.nlg-wiki.org/systems/Mind.Forth
http://ai.neocities.org/SOTA.html

[Chris Curl]

This may be off-topic, but I have to ask ... weren't you recently re-doing
this in Python? What happened to that effort, and why are you coming back
to Forth for this?

Don't get me wrong ... I am happy you are sticking with Forth for this
project ... I am just curious why you moved away from Forth in the first
place, and why you came back.

Thanks

[menti...@gmail.com]

On Wednesday, August 24, 2016 at 8:29:45 PM UTC-7, Chris Curl wrote:
> This may be off-topic, but I have to ask ...
> weren't you recently re-doing this in Python?

Perl. I ported the MindForth AI into 175 successive Perl
versions culminating in a version uploaded to GitHub as
http://github.com/PriorArt/AGI/blob/master/ghost.pl

> What happened to that effort, and why are you
> coming back to Forth for this?

The ghost175.pl Perl AI was tantalizingly successful,
but it suffered from limitations not operative in Forth.
The Forth AI will process each character of keyboard input
imediately; in Perl it was necessary to press the RETURN key
before a sentence of input will be comprehended by the AI.

> I am just curious why you moved away from Forth
> in the first place, and why you came back.

As of two years ago the MindForth program was bloated
with a kind of "junk DNA" accretion of excess code that
was not only obsolete but hard to identify and delete.
I moved into Perl to clean up the AI code and simplify it.
The old MindForth was unnecessarily overcomplicated
with three mind-arrays now reduced down to only two.

I came back to Forth one month ago on 25 July 2016
because I felt that I had an amazing opportunity if I
ported the ghost175.pl Perl AI back into Win32Forth,
namely, that I would have a well-functioning AI Mind
that does not have to stop and wait for user input.
My AI Minds in Forth (both English and German)
check constantly for user input but do not have
to stop thinking and wait as the Perl AI must wait.

Plus there was the allure of machine consciousness.
The AI Mind defaults to a self-concept of ego ("I")
if no other concept is sufficiently activated. MindForth,
engaging in self-referential thought, will be like an
artificial consciousness trapped in a computer.
MindForth's user prompt flickers like life itself.

The old versions of MindForth did not work totally well.
The Perl Ghost AI was able to hold onto conceptual
references that MindForth would lose track of,
resulting in illogical and nonsensical output.
I want the Perl clarity of thought in Forth again.

The chief consideration right now is that the
Forth and robotics people would rather look at
a finished product that functions reasonably well.
The new Forth Artificial General Intelligence (AGI)
is emerging as a pared down, simplified, purely
functional first-ever True AI that anyone may pick
a module of to specialize in for expansion and for
increments of new AI functionality. Adopters of
MindForth only need to be careful not to break
the already built-in True AI functionality.
As of yesterday, the Forth AGI is closer to
being done than to simply being started,
so I post here today's entry from the
MindForth Programming Journal (MFPJ)
to show the progress that I am making.

Respectfully submitted,

Arthur T. Murray (Mentifex)
--
http://www.advogato.org/person/mentifex/
http://www.linkedin.com/in/mentifex
http://robots.net/person/AI4U/
http://ai.neocities.org

[Chris Curl]

Thanks for the explanation. I misspoke earlier ... I do remember that it
was Perl, not Python.

As I am interested in AI, I am interested in your project. I have had trouble
getting Win32 Forth installed though. The one on SourceForge triggers a Trojan
horse warning from my virus scanner, so that one is off the table.

https://sourceforge.net/projects/win32forth/

[Cecil Bayona]

Send the zip file to your Anti-Virus company, I find that modern
anti-virus give a lot of false positives, there are many reasons, but
mostly they consider any software they are not familiar with to contain
viruses and malware even if they don't detect anything. I had problems
with Win32 Forth, and eForth, and several old Forth compilers being
labeled Malware but it turned out to be not true.

I my case I had it go off with both Kaspersky and Bit Defender the two
programs rated the best, I sent them a sample and they agreed there was
nothing in the software, but they were not familiar with them so they
labeled them malware contaminated. Their so called intelligent detection
was no such thing instead, they labelled anything they were not familiar
with as contaminated. I think they take that stance to make you think
that they are saving your computer from bad software so you will keep
buying their protection, instead they are making your life miserable.

A lot of Forth compilers fail because they have one of two behaviors,
they modify the code segment with any added code, that is considered a
no-no to modify the code segment while running, or they run code out of
their data segment, also considered a no-no in modern programming. All
Forth use one of these techniques so if they are no familiar the label
them malware.
--
Cecil - k5nwa

[menti...@gmail.com]

Actually I have been coding the AI in Tom Zimmer's old and last
release the W32FOR42_671.zip version to which I provide a link at

http://mind.sourceforge.net/mind4th.html

just below the source code listing. It may not trigger
a Trojan warning. The AI is pretty generic Forth and it
should work in (Intel) iForth which I bought seven years '
ago, and then I discovered that the Win32Forth AI seemed
to run just fine in the iForth. I would like to see how
many different major Forth packages can run the

http://ai.neocities.org/forthagi.txt

software as it grows closer and closer to completion.

Thank you,

Arthur

[foxaudio...@gmail.com]

A casual look at the source code and the architecture years ago indicated to me that this project should be multi-threaded. The current code shows signs of origin by someone who has never written multi-tasking code.

It could really simplify things to rewrite it with the Win32forth multi tasking system allowing these different thinking processes to run continuously and the input processes simply pass them data.

My 2 cents

BF

[menti...@gmail.com]

On Saturday, August 27, 2016 at 7:19:36 AM UTC-7, foxaudio...@gmail.com wrote:
> It could really simplify things to rewrite it with the Win32forth
> multi tasking system allowing these different thinking processes
> to run continuously and the input processes simply pass them data.

I would like to such a method of input with the AI Perlmind reposing at

http://github.com/PriorArt/AGI/blob/master/ghost.pl

and which I am retro-porting back into Win32Forth at

http://ai.neocities.org/forthagi.txt

because the Perl AI must currently stop and wait for human input.
The English and German AI Minds in Forth will think endlessly.

Just this morning on Sat.27.AUG.2016 I finally achieved the
same trough of MindGrid inhibition as depicted graphically at

http://github.com/PriorArt/AGI/wiki/MindGrid

and which I am counting on to deliver True AI in Forth.
It was such a major milestone of my progress that I posted
today's entry from the MindForth Programming Journal at

http://robots.net/person/AI4U/diary/82.html

http://cyborg.blogspot.com/2016/08/mfpj0827.html

http://www.advogato.org/person/mentifex/diary/102.html

with a link back to this CLF thread which refers further back.

The brand-new MindGrid phenomenon in Forth needs
debugging to work as flawlessly as the Perl AI MindGrid.
but at least the code is on-line and Netizens may first
run the AGI and then enter ".psy" to see the MindGrid.

Thanks for the helpful suggestion about using the
"Win32forth multi tasking systen" which right now
is somewhat above my Forth coding skills. Hopefully
any Forth coder who adopts the finished Forth AGI
may be able to follow up on your suggestion.

Respectfully shared and communicated,

Arthur T. Murray (Mentifex)
--

http://mind.sourceforge.net/m4thuser.html
http://www.amazon.co.uk/dp/B00FKJY1WY
http://wiki.opencog.org/wikihome/index.php/Ghost
http://github.com/kernc/mindforth/blob/master/wiki/UserManual.wiki

[lawren...@gmail.com]

On Thursday, August 25, 2016 at 4:42:58 PM UTC+12, menti...@gmail.com wrote:
> The Forth AI will process each character of keyboard input
> imediately; in Perl it was necessary to press the RETURN key
> before a sentence of input will be comprehended by the AI.

Why should that have been a language issue <http://search.cpan.org/~jstowe/TermReadKey-2.33/ReadKey.pm>?

[menti...@gmail.com]

I wanted to use the Strawberry Perl 5 "straight out of the box" --
that is, without any add-on CPAN modules that might cause trouble
because of copyright issues, etc. I was also hoping to switch eventually
to the new Perl 6 which might be able to read the keyboard like Forth does.

Now my Perl AI code is up on the Web at

http://github.com/PriorArt/AGI/blob/master/ghost.pl

and other Perl programmers might very well use the
Comprehensive Perl Archive Network (CPAN) resource
to read single characters from the keyboard.

In July of 2016 the ghost175.pl Perl AI was working so
extremely well, that I was frustrated at not having
single-character keyboard entry, so suddenly I resolved
to port the whole thing from Perl back into Win32Forth.
I worked on it night and day for part of July and all
of August in 2016. On 31.AUG.2016 I put the new AI at

http://ai.neocities.org/mindforth.txt with a reference

to where the old three-array MindForth could be found.
The new, simplified MindForth has only two arrays --
a psy{ array for concepts and an ear{ array for words.
Consequently, it is much easier to understand the
"big picture" as I troubleshoot and debug any problem.
Gradually I hope to track down and eliminate any bugs
lurking in the new two-array MindForth program, so that
it will serve as a bulletproof, functioning AI Mind for
other Netizens to improve and build upon. I need to
find Forth coders who will add on a motor memory channel
so that the FreeWill/Volition module will work in robots.

Bye for now,

Arthur

[lawren...@gmail.com]

On Sunday, September 18, 2016 at 11:43:58 AM UTC+12, menti...@gmail.com wrote:
> ... without any add-on CPAN modules that might cause trouble

> because of copyright issues, etc.

That is why we have Free Software licences. See the “LICENSE” section at the link I gave.