IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

The Art of Computer Mindmaking

Arthur T. Murray uj797 at victoria.tc.ca
Mon Feb 8 08:45:39 EST 1999


Am 5. Februar 1999, kurz nach Mitternacht im Hotel Prager Fruehling.
Just by putting up a faulty and deficient REENTRY screen yesterday
at Vaierre, I obtained something to think about.
   Now I am beginning to realize that I can not go directly into
SHORTTERMMEM (STM).

Wowowow-oh-wow!  This Forth language is amazingly powerful.

I started off tonight by analyzing where REENTRY must cut into the
system.  From 26nov1994 Mind.rexx it seemd to be at SENSORIUM.

So I went into Mind.forth SENSORIUM and I saw that any change --
any new code -- would have to be inserted into AUDITION Scr #9.
But I knew that I wanted to bypass the keyboard, because the in-
put was coming from the artificial mind itself.  (This progress
is looking good for my 24th upload later today.)

It became clear that the way to bypass the keyboard input was to
conditionalize it.  Aparrently that need was why I put in the
variable "earshut" several months ago.

If "earshut" becomes true (-1) in SENSORIUM Scr #14, then the
Forthmind does not want to listen to any more input but wants to
think instead.

Even the MIND Scr #44 begins with "earshut" being set to "false"
so that the AI will listen to user input.

Just now I have been thinking that "earshut" is too negative a
word, and that perhaps we should change it to something positive,
like "allears" or "listening."  But then upon further inspection
I see that we already have the word "alert" in MIND Scr #44.
Probably it is time to change "alert" to "listening."  We could
use the word "obedient," but later on the robots might not like
that choice.

Anyway, I went into AUDITION Scr #9 and I conditionalized the key-
board entry, so that "earshut" had to be not true.  But "not shut"
is the same as "open," so I want a positive word like "listening."

The problem with "alert" is that we kind of want the robot to be
alert at all times, but maybe we should accept the idea that the
robot is not so alert while engaged in thought-generation.

After I conditionalized the keyboard entry, I figured that I would
need to perform a fetch-from-array to supply each re-entered
phoneme, but to my shock I discovered that the array-fetch had
already been done, even with a loop, in the SPEECH Scr #34, with
the result going as "unk" straight into REENTRY Scr #33, so I did
not need to do a fetch from the array ear{ }.  Just by shifting
the information flow into the AUDITION Scr #9, I was making the
wanted variable "unk" available to the sensorium routines.

When I next ran the Mind.forth program, I got a weird phenomenon
of only one character being re-entered at a time, but the effect
was galvanizing upon me, because I realized instantly that the
core of the REENTRY problem had been solved.

Now I went into AUDITION Scr #9 and I used "alert" instead of
"earshut," but I forgot to change to positive logic in the if-
clause, so I had a run-on screen problem and I had to reboot.
Luckily I had just saved the astounding state-of-the-program
on one of the four disks at hand.

Now with "alert" in AUDITION Scr #9 I have switched to positive
logic and the program is no longer going haywire.

I think that somewhere in the re-entry rationale I may need to
turn the "alert" status back on to "true."  Oh, it does happen
in THINK Scr #43, but it may need to happen sooner.

In LANG-UK Scr #42 I have set "alert" to zero (false), with a
comment, "The robot is thinking, not alert."  However, I was
getting a spurious "User:" prompt during the re-entry of the
robot thought.  I have tracked that spurious prompt down to
SENSORIUM Scr #14, where it is no longer true that only the
"User" will be doing things in SENSORIUM.

I have switched the "User:" prompt up one level in the hierarchy
to MIND Scr #44, just before the call to SENSORIUM Scr #14.  Now
the "User:" prompt should be appropriate because the SENSORIUM is
receiving user input.

Now I am saving the program to disk, in case anything goes wrong.

Oops, I forgot to take out the asterisk, so the first-ever line of
good reentry was interspersed with Hyman-Kaplan-style asterisks --
and also with reduplicated letters.
--
http://www.geocities.com/Athens/Agora/7256/m-forth.html Mind.forth



More information about the Neur-sci mailing list

Send comments to us at biosci-help [At] net.bio.net