24 Feb
2024
24 Feb
'24
12:36 p.m.
Has anyone used it or something contemporaneous?
Is it at all applicable to any degree to today's approach to AI/machine learning
tasks? I would like to perhaps eventually create a game, probably not chess, lilely
something simpler. The old expert system modeling paradigm seems to have largely if not
entirely fallen out of favor. From what I'm reading though TP seems to be geared for
that.
24 Feb
24 Feb
8:22 p.m.
I bought it when my first ship was in Bahrain, but never did anything with it. I think I
still have the manuals, but am not sure about the floppies. Maybe something for when I
retire.
Zane
Sent from my iPhone
On Feb 24, 2024, at 12:36 PM, Just Kant via cctalk
<cctalk(a)classiccmp.org> wrote:
Has anyone used it or something contemporaneous?
Is it at all applicable to any degree to today's approach to AI/machine learning
tasks? I would like to perhaps eventually create a game, probably not chess, lilely
something simpler. The old expert system modeling paradigm seems to have largely if not
entirely fallen out of favor. From what I'm reading though TP seems to be geared for
that.
25 Feb
25 Feb
5:35 a.m.
On Sat, Feb 24, 2024 at 08:36:35PM +0000, Just Kant via cctalk wrote:
Has anyone used it or something contemporaneous?
Not me, at least not yet. I am kind of wet dreaming about it, so maybe
one day.
Is it at all applicable to any degree to today's
approach to
AI/machine learning tasks? I would like to perhaps eventually create
a game, probably not chess, lilely something simpler. The old expert
system modeling paradigm seems to have largely if not entirely
fallen out of favor. From what I'm reading though TP seems to be
geared for that.
AI is many things, and one of them is Chad Geppetto. I think Prolog is
quite underappreciated and it might be more relevant (at least in some
cases) simply because, unlike Chad, one's code in Prolog will not
hallucinate. Unless this is exactly what you put into it. From my
point of view, Chad is close to useless, because I could not trust the
results. It would tell me a story in a second and I would then spent a
day or a week trying to verify veracity of it.
But Chad is a new fancy of some people. As long as it does not become
my problem, I see no problem.
As of games, I would first test the language. I do not think real-time
kind of speed/reaction was in Prolog's design criteria. I would try
OCaml for games - never used this either, but I have read they keep
their compiler stable and code written umpty (twenty?) years ago
compiles nowadays without changes. You may appreciate this when your
game turns fifteen years old. I have read about guy who first coded
his game in Java and gradually driven himself to a point when he could
no longer maintain it, so he rewrote in OCaml. I have read about other
guy, who wrote their tool in Python and for whatever reason (do not
remember) rewrote it on OCaml, again. They sounded satisfied with
their choices.
HTH
--
Regards,
Tomasz Rola
--
** A C programmer asked whether computer had Buddha's nature. **
** As the answer, master did "rm -rif" on the programmer's home **
** directory. And then the C programmer became enlightened... **
** **
** Tomasz Rola mailto:tomasz_rola@bigfoot.com **
1:15 p.m.
So the portions of code belonging to chatgpt which produce the hallucinations have been
isolated?
Which languages were used to build it?
2:20 p.m.
On Sun, Feb 25, 2024 at 3:15 PM Just Kant via cctalk
<cctalk(a)classiccmp.org> wrote:
So the portions of code belonging to chatgpt which
produce the hallucinations have been isolated?
It's a massive deep neural network, so you can't really isolate
anything. But there are parameters that you can use to tune it, like
how quickly it forgets earlier parts of a conversation, etc. and some
people speculate that they tweaked something like that which resulted
in the recent issues.
Which languages were used to build it?
One could say Python, but that mostly sits on top of C++, which then
invokes CUDA (or TPU or similar) code, but at the bottom it's all just
matrix multiplication.
Polog and other Logic Programming tools aren't applicable to Machine
Learning approaches which is all anyone is interested in for "AI"
these days. If you want to make a rules-based expert system, and you
know what all the rules are, then Prolog might still be a useful tool.
Turbo Pascal is even still available as its originators took it back
from Borland and made it into Visual Prolog for Windows which has a
free personal edition (the commercial license is only 100 euros too).
Also there's GNU Prolog if you just want to futz around with Prolog.
2:21 p.m.
Of course I mean Turbo Prolog there, sorry.
On Sun, Feb 25, 2024 at 4:20 PM Gavin Scott <gavin(a)learn.bio> wrote:
On Sun, Feb 25, 2024 at 3:15 PM Just Kant via cctalk
<cctalk(a)classiccmp.org> wrote:
So the portions of code belonging to chatgpt
which produce the hallucinations have been isolated?
It's a massive deep neural network, so you can't really isolate
anything. But there are parameters that you can use to tune it, like
how quickly it forgets earlier parts of a conversation, etc. and some
people speculate that they tweaked something like that which resulted
in the recent issues.
Which languages were used to build it?
One could say Python, but that mostly sits on top of C++, which then
invokes CUDA (or TPU or similar) code, but at the bottom it's all just
matrix multiplication.
Polog and other Logic Programming tools aren't applicable to Machine
Learning approaches which is all anyone is interested in for "AI"
these days. If you want to make a rules-based expert system, and you
know what all the rules are, then Prolog might still be a useful tool.
Turbo Pascal is even still available as its originators took it back
from Borland and made it into Visual Prolog for Windows which has a
free personal edition (the commercial license is only 100 euros too).
Also there's GNU Prolog if you just want to futz around with Prolog. 
3:41 p.m.
On 2/25/24 16:20, Gavin Scott via cctalk wrote:
Turbo Pascal is even still available as its originators took it back
from Borland and made it into Visual Prolog for Windows which has a
free personal edition (the commercial license is only 100 euros too).
Also there's GNU Prolog if you just want to futz around with Prolog.
I wrote a large application in Turbo Pascal in 1996 or so,
and it had been cast adrift since then. The computer it ran
on was getting old and I was worried it would die, and then
I discovered FPC (Free Pascal Compiler). It was designed to
port over Turbo Pascal and DEC Pascal programs, and took the
separate compilation directives like uses and $I. I was
able to get that program running on a Linux system in a few
days.
Jon
4:01 p.m.
He meant to say Prolog, not Pascal.
Regardless if you want to alleviate all the fuss and mess of running 16 bit wares on
modern h/w, just look for a 32 bit cast off. Many appropriate mobos can be had on epay for
a song.
Now no one I know wants to spend the next 40 years writing 16 bit apps. But if it
can't be (easily) done otherwise, it can be done on raw metal cheaply.
Sent with Proton Mail secure email.
On Sunday, February 25th, 2024 at 6:41 PM, Jon Elson via cctalk
<cctalk(a)classiccmp.org> wrote:
I wrote a large application in Turbo Pascal in 1996 or
so,
and it had been cast adrift since then. The computer it ran
on was getting old and I was worried it would die, and then
I discovered FPC (Free Pascal Compiler). It was designed to
port over Turbo Pascal and DEC Pascal programs, and took the
separate compilation directives like uses and $I. I was
able to get that program running on a Linux system in a few
days.
Jon
11:41 a.m.
35 years ago I got tasked to write a simple expert system in Turbo Prolog because I was
familiar with Turbo Pascal. The goal was an application to assist new members of the help
desk. I have vague recollections of having to define rules to evaluate answers to simple
questions. What I remember seems far removed from what I know of modern AI
implementations.
- Rod
On Feb 24, 2024, at 3:36 PM, Just Kant via cctalk
<cctalk(a)classiccmp.org> wrote:
Has anyone used it or something contemporaneous?
Is it at all applicable to any degree to today's approach to AI/machine learning
tasks? I would like to perhaps eventually create a game, probably not chess, lilely
something simpler. The old expert system modeling paradigm seems to have largely if not
entirely fallen out of favor. From what I'm reading though TP seems to be geared for
that.
303
days inactive
305
days old
8 comments
6 participants
participants (6)
-
Gavin Scott -
Jon Elson -
Just Kant -
Rod Bartlett -
Tomasz Rola -
Zane Healy