Re: Training the AI to write better LiveCode

[Roger Guay via use-livecode]

I forgot . . . can’t send attachments. Here is the dialogue:

Me:
A-A-A=25 AND B-B-B=20 AND C-C-C=15, WHAT IS A+B+C?

GPT:
A + B + C = (A-A-A + B-B-B + C-C-C)/3 + (A + B + C) = (25 + 20 + 15)/3 + (A + B 
+ C) = 60/3 + (A + B + C) = 20 + (A + B + C) = 20 + A + B + C

Therefore, A + B + C = 20 + A + B + C, and (A + B + C) = 20.


Me:
try again. Given that A-A-A=25 and B-B-B=20 and  C-C-C=15, what is A+B+C?


GPT:
If A-A-A=25, then A = (25 + 3A)/3, so 3A = 25 + 3A, and 2A = 25, and A = 12.5.

Similarly, B = (20 + 3B)/3, so B = 10, and C = (15 + 3C)/3, so C = 5.

So, A + B + C = 12.5 + 10 + 5 = 27.5.


Me:
Solve each given equation individually and try again


GPT:
Sure, let's solve each equation individually:

A-A-A = 25

A = (25 + 3A)/3

3A = 25 + 3A

2A = 25

A = 12.5

B-B-B = 20

B = (20 + 3B)/3

3B = 20 + 3B

2B = 20

B = 10

C-C-C = 15

C = (15 + 3C)/3

3C = 15 + 3C

2C = 15

C = 7.5

So, A + B + C = 12.5 + 10 + 7.5 = 30.



Me:
what is A - A -A?


GPT:
A - A - A is an algebraic expression that represents subtracting the same 
quantity, A, three times. If A is a number, then A - A - A would equal -2A.

However, in the context of a specific problem, the value of A - A - A may be 
something else, such as 25 in the equation A - A - A = 25.


Me:
I disagree. A - A - A = A -2A = -A


GPT:
Yes, you are correct. I apologize for the mistake in my previous answer. A - A 
- A = A - 2A = -A. Thank you for bringing this to my attention.


Me:
Now go back to the original question and solve it


GPT:
Given the equations A-A-A = 25 and B-B-B = 20 and C-C-C = 15, we can find the 
values of A, B, and C and then calculate the sum A + B + C.

Starting with A-A-A = 25, we can simplify this expression:

A - A - A = A - 2A = -A = 25

Therefore, A = -25.

Similarly, we can find B and C:

B-B-B = 20

B = -20

C-C-C = 15

C = -15

Finally, we can calculate A + B + C:

A + B + C = -25 + -20 + -15 = -60.

So, the answer to A + B + C given that A-A-A = 25, B-B-B = 20, and C-C-C = 15 
is -60.



Me:
Yes! can you remember this?

GPT:
As an AI language model, I don't have the capacity to remember specific 
information in the sense that humans do, but I have been trained on a vast 
corpus of text, so I can use that knowledge to answer questions like this one. 
So, I can assist you with similar questions in the future.




> On Feb 6, 2023, at 12:44 PM, Roger Guay via use-livecode 
>  wrote:
> 
> Forgive me for jumping in, but decided to try this myself. Here is the 
> dialogue attempting to correct ChatGPT :
> 
> 
> What fun!
> 
> Roger
> 
>> On Feb 6, 2023, at 8:04 AM, Mike Kerner via use-livecode 
>>  wrote:
>> 
>> that's not what chatgpt came up with.
>> but that's the point - it's still early.
>> 
>> On Mon, Feb 6, 2023 at 9:57 AM Craig Newman via use-livecode <
>> use-livecode@lists.runrev.com> wrote:
>> 
>>> A = -25
>>> B = -20
>>> C = -15
>>> 
>>> So, -60.
>>> 
>>> Craig
>>> 
> 
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Roger Guay via use-livecode]

Forgive me for jumping in, but decided to try this myself. Here is the dialogue 
attempting to correct ChatGPT :


What fun!

Roger

> On Feb 6, 2023, at 8:04 AM, Mike Kerner via use-livecode 
>  wrote:
> 
> that's not what chatgpt came up with.
> but that's the point - it's still early.
> 
> On Mon, Feb 6, 2023 at 9:57 AM Craig Newman via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> 
>> A = -25
>> B = -20
>> C = -15
>> 
>> So, -60.
>> 
>> Craig
>> 

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

that's not what chatgpt came up with.
but that's the point - it's still early.

On Mon, Feb 6, 2023 at 9:57 AM Craig Newman via use-livecode <
use-livecode@lists.runrev.com> wrote:

> A = -25
> B = -20
> C = -15
>
> So, -60.
>
> Craig
>
> > On Feb 6, 2023, at 9:03 AM, Mike Kerner via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> >
> > there was also a thread on twitter this weekend, which was roughly this:
> > A - A - A = 25
> > B - B - B = 20
> > C - C - C = 15
> > A + B + C = ?
> > It's still early.
> >
> > On Sun, Feb 5, 2023 at 8:57 AM Mark Rauterkus via use-livecode <
> > use-livecode@lists.runrev.com> wrote:
> >
> >> Medium article: Guy uses Chat GPT to program home assistant.
> >>
> >>
> >>
> https://matemarschalko.medium.com/chatgpt-in-an-ios-shortcut-worlds-smartest-homekit-voice-assistant-9a33b780007a
> >>
> >>
> >>
> >> --
> >> Ta.
> >>
> >>
> >> Mark Rauterkus   m...@rauterkus.com
> >> ___
> >> use-livecode mailing list
> >> use-livecode@lists.runrev.com
> >> Please visit this url to subscribe, unsubscribe and manage your
> >> subscription preferences:
> >> http://lists.runrev.com/mailman/listinfo/use-livecode
> >>
> >
> >
> > --
> > On the first day, God created the heavens and the Earth
> > On the second day, God created the oceans.
> > On the third day, God put the animals on hold for a few hours,
> >   and did a little diving.
> > And God said, "This is good."
> > ___
> > use-livecode mailing list
> > use-livecode@lists.runrev.com
> > Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> > http://lists.runrev.com/mailman/listinfo/use-livecode
>
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Craig Newman via use-livecode]

A = -25
B = -20
C = -15

So, -60.

Craig

> On Feb 6, 2023, at 9:03 AM, Mike Kerner via use-livecode 
>  wrote:
> 
> there was also a thread on twitter this weekend, which was roughly this:
> A - A - A = 25
> B - B - B = 20
> C - C - C = 15
> A + B + C = ?
> It's still early.
> 
> On Sun, Feb 5, 2023 at 8:57 AM Mark Rauterkus via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> 
>> Medium article: Guy uses Chat GPT to program home assistant.
>> 
>> 
>> https://matemarschalko.medium.com/chatgpt-in-an-ios-shortcut-worlds-smartest-homekit-voice-assistant-9a33b780007a
>> 
>> 
>> 
>> --
>> Ta.
>> 
>> 
>> Mark Rauterkus   m...@rauterkus.com
>> ___
>> use-livecode mailing list
>> use-livecode@lists.runrev.com
>> Please visit this url to subscribe, unsubscribe and manage your
>> subscription preferences:
>> http://lists.runrev.com/mailman/listinfo/use-livecode
>> 
> 
> 
> -- 
> On the first day, God created the heavens and the Earth
> On the second day, God created the oceans.
> On the third day, God put the animals on hold for a few hours,
>   and did a little diving.
> And God said, "This is good."
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

also, from zdnet, this am
https://www.zdnet.com/article/i-asked-chatgpt-to-write-a-wordpress-plugin-i-needed-it-did-in-less-than-5-minutes/

On Sun, Feb 5, 2023 at 8:57 AM Mark Rauterkus via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Medium article: Guy uses Chat GPT to program home assistant.
>
>
> https://matemarschalko.medium.com/chatgpt-in-an-ios-shortcut-worlds-smartest-homekit-voice-assistant-9a33b780007a
>
>
>
> --
> Ta.
>
>
> Mark Rauterkus   m...@rauterkus.com
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

there was also a thread on twitter this weekend, which was roughly this:
A - A - A = 25
B - B - B = 20
C - C - C = 15
A + B + C = ?
It's still early.

On Sun, Feb 5, 2023 at 8:57 AM Mark Rauterkus via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Medium article: Guy uses Chat GPT to program home assistant.
>
>
> https://matemarschalko.medium.com/chatgpt-in-an-ios-shortcut-worlds-smartest-homekit-voice-assistant-9a33b780007a
>
>
>
> --
> Ta.
>
>
> Mark Rauterkus   m...@rauterkus.com
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Rauterkus via use-livecode]

Medium article: Guy uses Chat GPT to program home assistant.

https://matemarschalko.medium.com/chatgpt-in-an-ios-shortcut-worlds-smartest-homekit-voice-assistant-9a33b780007a



--
Ta.


Mark Rauterkus   m...@rauterkus.com
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

zdnet is reporting that some developers have found it to be useful for
debugging code, too.

On Thu, Jan 26, 2023 at 6:28 AM Mark Smith via use-livecode <
use-livecode@lists.runrev.com> wrote:

> I read a fascinating article in the Independent this morning that
> indicates just how disruptive this technology could be: ChatGPT passes MBA
> final exam at Wharton School of Business (research study, final paper in
> preparation),  second, Harvard researchers find ChatGPT could pass the US
> Medical licensing exam (it performed at or near the passing level for all 3
> finals). Of course, egregious errors were found in both sets of exams, but
> ChatGPT is barely out of the box (so to speak). It remains to be seen just
> how long it will take to mature but if I had to guess, I would say your
> next business consultant or GP (a few years from now) could well be a
> chatbot.
>
>
> https://www.independent.co.uk/tech/chatgpt-mba-exam-wharton-professor-b2267919.html
> <
> https://www.independent.co.uk/tech/chatgpt-mba-exam-wharton-professor-b2267919.html
> >
>
> Can we do it without bloodshed, or does this just nudge the doomsday clock
> forward a few more seconds? Let's hope we can.
>
> Laws and politicians will have to respond quickly to what can and cannot
> be done using this new technology, but who polices the politicians?
> Corruption and the abuse of power is not changed by the presence of this
> new kid on the block. It will be a race to see who or what can evolve more
> quickly; the tech or the societies that built it. Or will the tech help us
> to become better human beings; ones more capable of compassion, concern and
> understanding?
>
> Mark
>
> > On 25 Jan 2023, at 8:59 pm, Richard Gaskin via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> >
> > Bob Sneidar wrote:
> >
> > > If we get to the point where mankind no longer has to work
> > > to live (universal basic income), then I fear we will come
> > > to experience intimately what the old saying hints at, "An
> > > idle mind is the devil's playground."
> >
> > You'd be in good company. Camus, Sartre, and others in the middle of the
> last century about how the newfound leisure ultimately borne of the
> productivity gains of the Industrial Revolution might become the greatest
> crisis mankind faces.
> >
> > Prone as I am to myopic projection, I'm not so sure. If we find the
> tedious work of providing basis essentials delivered by machines, I believe
> we'd find new and more interesting things to do.
> >
> > I've enjoyed the rhythmic calisthenics of digging ditches, and the
> back-to-nature connectedness of farming. But TBH for all those jobs taught
> me I find designing board games more fascinating.
> >
> >
> > > It will be the end of any really productive society.
> >
> > Art isn't "productive", but I'm glad people do it.
> >
> > And at the moment the only way to stop it is to try to shut the machines
> down by throwing a shoe into their gears (the legendary etymology of
> "saboteur").
> >
> > With automation resulting in widespread permanent unemployment, folks
> will be idle either way.
> >
> > The only question is whether we want to see the masses thriving, or
> living in a cardboard box under a freeway no longer driven by anyone but
> the owners of the machines.
> >
> > I prefer thriving.
> >
> > --
> > Richard Gaskin
> > Fourth World Systems
> >
> >
> >
> >
> > ___
> > use-livecode mailing list
> > use-livecode@lists.runrev.com
> > Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> > http://lists.runrev.com/mailman/listinfo/use-livecode
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Smith via use-livecode]

I read a fascinating article in the Independent this morning that indicates 
just how disruptive this technology could be: ChatGPT passes MBA final exam at 
Wharton School of Business (research study, final paper in preparation),  
second, Harvard researchers find ChatGPT could pass the US Medical licensing 
exam (it performed at or near the passing level for all 3 finals). Of course, 
egregious errors were found in both sets of exams, but ChatGPT is barely out of 
the box (so to speak). It remains to be seen just how long it will take to 
mature but if I had to guess, I would say your next business consultant or GP 
(a few years from now) could well be a chatbot. 

https://www.independent.co.uk/tech/chatgpt-mba-exam-wharton-professor-b2267919.html
 


Can we do it without bloodshed, or does this just nudge the doomsday clock 
forward a few more seconds? Let's hope we can. 

Laws and politicians will have to respond quickly to what can and cannot be 
done using this new technology, but who polices the politicians? Corruption and 
the abuse of power is not changed by the presence of this new kid on the block. 
It will be a race to see who or what can evolve more quickly; the tech or the 
societies that built it. Or will the tech help us to become better human 
beings; ones more capable of compassion, concern and understanding?

Mark

> On 25 Jan 2023, at 8:59 pm, Richard Gaskin via use-livecode 
>  wrote:
> 
> Bob Sneidar wrote:
> 
> > If we get to the point where mankind no longer has to work
> > to live (universal basic income), then I fear we will come
> > to experience intimately what the old saying hints at, "An
> > idle mind is the devil's playground."
> 
> You'd be in good company. Camus, Sartre, and others in the middle of the last 
> century about how the newfound leisure ultimately borne of the productivity 
> gains of the Industrial Revolution might become the greatest crisis mankind 
> faces.
> 
> Prone as I am to myopic projection, I'm not so sure. If we find the tedious 
> work of providing basis essentials delivered by machines, I believe we'd find 
> new and more interesting things to do.
> 
> I've enjoyed the rhythmic calisthenics of digging ditches, and the 
> back-to-nature connectedness of farming. But TBH for all those jobs taught me 
> I find designing board games more fascinating.
> 
> 
> > It will be the end of any really productive society.
> 
> Art isn't "productive", but I'm glad people do it.
> 
> And at the moment the only way to stop it is to try to shut the machines down 
> by throwing a shoe into their gears (the legendary etymology of "saboteur").
> 
> With automation resulting in widespread permanent unemployment, folks will be 
> idle either way.
> 
> The only question is whether we want to see the masses thriving, or living in 
> a cardboard box under a freeway no longer driven by anyone but the owners of 
> the machines.
> 
> I prefer thriving.
> 
> -- 
> Richard Gaskin
> Fourth World Systems
> 
> 
> 
> 
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

Bob Sneidar wrote:

> If we get to the point where mankind no longer has to work
> to live (universal basic income), then I fear we will come
> to experience intimately what the old saying hints at, "An
> idle mind is the devil's playground."

You'd be in good company. Camus, Sartre, and others in the middle of the 
last century about how the newfound leisure ultimately borne of the 
productivity gains of the Industrial Revolution might become the 
greatest crisis mankind faces.


Prone as I am to myopic projection, I'm not so sure. If we find the 
tedious work of providing basis essentials delivered by machines, I 
believe we'd find new and more interesting things to do.


I've enjoyed the rhythmic calisthenics of digging ditches, and the 
back-to-nature connectedness of farming. But TBH for all those jobs 
taught me I find designing board games more fascinating.



> It will be the end of any really productive society.

Art isn't "productive", but I'm glad people do it.

And at the moment the only way to stop it is to try to shut the machines 
down by throwing a shoe into their gears (the legendary etymology of 
"saboteur").


With automation resulting in widespread permanent unemployment, folks 
will be idle either way.


The only question is whether we want to see the masses thriving, or 
living in a cardboard box under a freeway no longer driven by anyone but 
the owners of the machines.


I prefer thriving.

--
 Richard Gaskin
 Fourth World Systems




___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

If we get to the point where mankind no longer has to work to live (universal 
basic income), then I fear we will come to experience intimately what the old 
saying hints at, "An idle mind is the devil's playground." It will be the end 
of any really productive society. Or else what will develop is a society where 
a minority elite class of people are forming the culture, and the relatively 
poor masses, dependent on welfare will be their subjects. I don't see either of 
those as, "good news" Sam Altman's faith notwithstanding.

Bob S


On Jan 25, 2023, at 10:49 , Richard Gaskin via use-livecode 
mailto:use-livecode@lists.runrev.com>> wrote:

The good news is that even Sam Altman, CEO of OpenAI, believes that the 
productivity gains from automation advancements will be more than sufficient to 
provide strong returns for shareholders even while paying a sort of "robot tax" 
that allows humans to enjoy a Universal Basic Income.

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

Interesting I had no idea there were that many potential progressions. But I am 
still curious how any AI makes the choice, "Not that, this." It has to be able 
to "recall" that the first move leads to failure. 

It has to have some way of paring down it's possible paths to success based 
upon what an opponent does, and it has to do it by trial and error, which of 
course is impossible without a way to recall a successful strategy. 

Bob S


> On Jan 24, 2023, at 17:20 , Geoff Canyon via use-livecode 
>  wrote:
> 
> On Tue, Jan 24, 2023 at 8:10 AM Bob Sneidar via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> 
>> I don't think it needs to store ALL the permutations, only the viable
>> ones, the ones that lead to success. That has to be a much smaller number.
> 
> 
> There are only three outcomes: win, lose, draw. Even if the breakdown is
> 0.1% win, 0.1% lose, and 99.8% draw, that would still be far more positions
> than could be stored using all the computing power on Earth, a billion
> times over.
> 
>> But I was using that as an example of the mathematical nature of Chess. I
>> think what we must mean by AI is that through recursion, a computer can
>> retain successful paths to success (success being that which we define as
>> success in the process.) I don't think we will ever see the day where a
>> computer, lacking experience and all the data for a problem, can "reason"
>> it's way to success.
>> 
> 
> That's almost exactly what AlphaZero did: it was given the rules for moves,
> and a definition of win conditions, and then played against itself. It
> wasn't given any info on existing openings or endgames. It was entirely
> self-taught, in 9 hours. I think the only reason to say that it didn't
> reason about the game is that we *do* understand how it works at a low
> level, and at an abstract level, but we *don't* understand the specifics
> about how it works at a high level. It's the same way I might understand
> what a chess master means when they say a move is better because it's more
> active; I understand what "active" means in general, but I would likely not
> be able to say why that move was more active than several other moves.
> 
> gc
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Tue, Jan 24, 2023 at 8:10 AM Bob Sneidar via use-livecode <
use-livecode@lists.runrev.com> wrote:

> I don't think it needs to store ALL the permutations, only the viable
> ones, the ones that lead to success. That has to be a much smaller number.


There are only three outcomes: win, lose, draw. Even if the breakdown is
0.1% win, 0.1% lose, and 99.8% draw, that would still be far more positions
than could be stored using all the computing power on Earth, a billion
times over.

> But I was using that as an example of the mathematical nature of Chess. I
> think what we must mean by AI is that through recursion, a computer can
> retain successful paths to success (success being that which we define as
> success in the process.) I don't think we will ever see the day where a
> computer, lacking experience and all the data for a problem, can "reason"
> it's way to success.
>

That's almost exactly what AlphaZero did: it was given the rules for moves,
and a definition of win conditions, and then played against itself. It
wasn't given any info on existing openings or endgames. It was entirely
self-taught, in 9 hours. I think the only reason to say that it didn't
reason about the game is that we *do* understand how it works at a low
level, and at an abstract level, but we *don't* understand the specifics
about how it works at a high level. It's the same way I might understand
what a chess master means when they say a move is better because it's more
active; I understand what "active" means in general, but I would likely not
be able to say why that move was more active than several other moves.

gc
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

zdnet reports that chatgpt pro subscriptions are about $42/month.

On Tue, Jan 24, 2023 at 11:10 AM Bob Sneidar via use-livecode <
use-livecode@lists.runrev.com> wrote:

> I don't think it needs to store ALL the permutations, only the viable
> ones, the ones that lead to success. That has to be a much smaller number.
> But I was using that as an example of the mathematical nature of Chess. I
> think what we must mean by AI is that through recursion, a computer can
> retain successful paths to success (success being that which we define as
> success in the process.) I don't think we will ever see the day where a
> computer, lacking experience and all the data for a problem, can "reason"
> it's way to success.
>
> Bob S
>
>
> > On Jan 23, 2023, at 23:26 , Geoff Canyon via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> >
> > On Mon, Jan 23, 2023 at 4:09 PM Bob Sneidar via use-livecode <
> > use-livecode@lists.runrev.com> wrote:
> >
> >> Meh. Don't like using Chess as a measurement of AI competence. Chess is
> in
> >> AI's wheelhouse. Once a computer can store all the possible
> permutations of
> >> chess,
> >>
> >
> > I agree on the sentiment, but a small point: the state-space of Chess is
> > about 10^44 . No computer
> > has, or foreseeably will, store all the possible permutations. It's the
> > nature of machine learning that it's pretty much impossible for us to
> > understand what something like AlphaZero
> >  *does* store, but it's not
> > anything like "in this position, play this move". Further, it appears
> > AlphaZero used 16GB of memory. It looks like Stockfish 10 (which I think
> is
> > considered stronger than AlphaZero) can use more, but will operate in
> 16GB
> > of memory as well.
> >
> > gc
> > ___
> > use-livecode mailing list
> > use-livecode@lists.runrev.com
> > Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> > http://lists.runrev.com/mailman/listinfo/use-livecode
>
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

I don't think it needs to store ALL the permutations, only the viable ones, the 
ones that lead to success. That has to be a much smaller number. But I was 
using that as an example of the mathematical nature of Chess. I think what we 
must mean by AI is that through recursion, a computer can retain successful 
paths to success (success being that which we define as success in the 
process.) I don't think we will ever see the day where a computer, lacking 
experience and all the data for a problem, can "reason" it's way to success. 

Bob S


> On Jan 23, 2023, at 23:26 , Geoff Canyon via use-livecode 
>  wrote:
> 
> On Mon, Jan 23, 2023 at 4:09 PM Bob Sneidar via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> 
>> Meh. Don't like using Chess as a measurement of AI competence. Chess is in
>> AI's wheelhouse. Once a computer can store all the possible permutations of
>> chess,
>> 
> 
> I agree on the sentiment, but a small point: the state-space of Chess is
> about 10^44 . No computer
> has, or foreseeably will, store all the possible permutations. It's the
> nature of machine learning that it's pretty much impossible for us to
> understand what something like AlphaZero
>  *does* store, but it's not
> anything like "in this position, play this move". Further, it appears
> AlphaZero used 16GB of memory. It looks like Stockfish 10 (which I think is
> considered stronger than AlphaZero) can use more, but will operate in 16GB
> of memory as well.
> 
> gc
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Mon, Jan 23, 2023 at 4:09 PM Bob Sneidar via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Meh. Don't like using Chess as a measurement of AI competence. Chess is in
> AI's wheelhouse. Once a computer can store all the possible permutations of
> chess,
>

I agree on the sentiment, but a small point: the state-space of Chess is
about 10^44 . No computer
has, or foreseeably will, store all the possible permutations. It's the
nature of machine learning that it's pretty much impossible for us to
understand what something like AlphaZero
 *does* store, but it's not
anything like "in this position, play this move". Further, it appears
AlphaZero used 16GB of memory. It looks like Stockfish 10 (which I think is
considered stronger than AlphaZero) can use more, but will operate in 16GB
of memory as well.

gc
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Phil Davis via use-livecode]

+1
Spot-on Bob!


On 1/23/23 4:24 PM, Bob Sneidar via use-livecode wrote:

The real problem with information is not the information itself, but that we pretty much all 
process information in the context of our chosen world view. Because of this, freedom of speech and 
open discourse MUST be preserved, because my life's experience is that if the "good guys" 
and the "bad guys" want to control what information is available, the bad guys*ALWAYS*  
win. They are unhampered by such inconveniences such as truth and morality.

Bob S


--
Phil Davis
(503) 307-4363


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

Who gets to decide what the dis is in dis-information though? I think the 
greatest threat to knowledge is the illusion that information can be filtered 
to exclude the "dis" in favor of the information. Perhaps my faith in the human 
psyche is over ambitious, but it seems to me that any person really wanting to 
know what is true, and not just what confirms their own predisposition or bias, 
when given access to all the information available, can determine pretty 
quickly what is true, or false, or misleading.

The real problem with information is not the information itself, but that we 
pretty much all process information in the context of our chosen world view. 
Because of this, freedom of speech and open discourse MUST be preserved, 
because my life's experience is that if the "good guys" and the "bad guys" want 
to control what information is available, the bad guys *ALWAYS* win. They are 
unhampered by such inconveniences such as truth and morality.

Bob S


On Jan 21, 2023, at 14:47 , Geoff Canyon via use-livecode 
mailto:use-livecode@lists.runrev.com>> wrote:

On Sat, Jan 21, 2023 at 10:50 AM Richard Gaskin via use-livecode <
use-livecode@lists.runrev.com> wrote:


Human-generated disinformation is already a significant destabilizing
force. When we trust machine-generated content we run that same risk at
light speed.


Absolutely agreed -- If you'd asked me in the 90s what man-made thing would
end us all, I'd have said nanotechnology. In the 2000s I would have added
bio-tech. These days I definitely consider disinformation an equal threat.
___

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

Meh. Don't like using Chess as a measurement of AI competence. Chess is in AI's 
wheelhouse. Once a computer can store all the possible permutations of chess, 
it's advantage for AI is that it can reject what worked historically and what 
didn't, paring down the moves for every situation it can encounter to the ones 
that work. The initial paring down probably happens fairly quickly too. Most 
possible opening moves in chess are bad ones.

The disadvantage humans have in a scenario like Chess is that they cannot hold 
all the permutations and possible progressions of a game in their conscience 
mind. Computers don't have a conscience mind, they have direct access to all 
the data they have ever been exposed to.

What I get from your example is that the people writing those articles really 
misunderstood the problem.

Bob S


On Jan 21, 2023, at 10:39 , Richard Gaskin via use-livecode 
mailto:use-livecode@lists.runrev.com>> wrote:

Are we?  As late as my teens I was still reading science mags saying "Well, AI 
is going to be a big deal, but no machine will ever beat a human at something 
as complex as chess."

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Franz Nahrada via use-livecode]

Kevin Miller schrieb am Montag, 23. Jänner 2023 um 22:56 :

> I don't think we need to worry (too much!) about the exit of scripting
>languages. It's a useful medium for a human to understand and edit in
>until such time as we are truly defunct. If and when that happens I doubt
>there will be any sort of jobs in any industry.

I am not very active with my Indy license and also with this mailing list,
but I follow this particular  debate closely and In have to say it really
raised my enthusiasm to start scripting again.
Actually it was a brilliant idea to kick this theme off and let it develop
so creatively. Thanks Kevin and Heather !
It seems like a natural match to me (the scripter and AI) that will
reverse the digital illiteracy of so many - who think they are digitally
literate because they can use prefab apps.
But we need to start building our own worlds again.

Franz Nahrada
ex - HyperCard support Austria
promoting Global Villages - where embeddedness in nature and landscape
meets connectedness to advanced knowledge

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Kevin Miller via use-livecode]

Ethics are very important to me as you know. I'm a passionate believer in 
working to increase fairness as we grow and evolve as a species. We're making 
great progress at that at the moment. For example, the number of people living 
in extreme poverty globally has more than halved in the last 20 years. Over 
half of the world's population has a smartphone, giving them access to what 
would once have been a supercomputer, and an online world that just 25 years 
ago the majority could not access.

I think it takes time after something new emerges before we sort out the 
societal implications fully. Such implications are rarely well understood at 
the start. I don't think the questions these lawsuits raise have easy answers. 
It certainly does not seem clear cut to me on reading them what is even right 
or in the ultimate best interests of artists, creators or our species. I would 
need to dig into this for a lot longer to truly form an opinion.

I don't underestimate the potential for technology to continue at breath-taking 
rate and solve the problem of writing in assembler. I get the exponential 
growth of technology and the improvements of algorithms that lead to further 
growths - in fact I'm quite excited about it. I'm just not sure that this is a 
problem that really needs solved next. We already have a technology that takes 
human readable code and creates machine instructions, i.e. compliers. Until 
technology completely eliminates humans in the making of software (which may of 
course happen one day) I don't think we need to worry (too much!) about the 
exit of scripting languages. It's a useful medium for a human to understand and 
edit in until such time as we are truly defunct. If and when that happens I 
doubt there will be any sort of jobs in any industry.

We also have to be careful we don't try to cut off our nose to spite our face. 
Leaving a language like ours behind in the dust rather than finding ways to 
work with new opportunities like this one is a high price to pay, particularly 
as such a protest would have little or no impact on the course of the evolution 
of such technology. Obviously we need to stand up for our rights as we go 
forward, I am not advocating sticking our collective heads in the sand. This 
could go in a negative direction, particularly depending on who owns it and has 
access to it and we can re-evaluate things if necessary in the future.

As to the first industrial revolution, I think it's easy to forget what it was 
actually like to live a couple of hundred years ago compared to now. I'm with 
Geoff, overall it came out extremely well. That gives one a little cautious 
optimism that the second one might also do so, though as ever the future is not 
guaranteed.

Kind regards, 

Kevin 

Kevin Miller ~ ke...@livecode.com ~ http://www.livecode.com/ 
LiveCode: Build Amazing Things 



On 21/01/2023, 22:40, "use-livecode on behalf of Geoff Canyon via 
use-livecode" mailto:use-livecode-boun...@lists.runrev.com> on behalf of 
use-livecode@lists.runrev.com > wrote:


Those only interested in LiveCode, click "next" now.

On Sat, Jan 21, 2023 at 10:40 AM Richard Gaskin via use-livecode <
use-livecode@lists.runrev.com > wrote:

>
> After all, the Codex had been trained on billions of
> publicly available source code lines – including code
> in public repositories on GitHub. That included, among
> other things, all of the Apache Foundation's many
> projects' code.
> https://www.theregister.com/2022/11/11/githubs_copilot_opinion/ 
> 
>
>
> Stability Diffusion, Midjourney, and DreamUp were
> trained on copyrighted materials without credit,
> compensation, or consent, according to a new lawsuit.
>
> https://www.pcmag.com/news/artists-sue-ai-art-generators-for-copyright-infringement
>  
> 

My not-a-lawyer understanding is that this lawsuit is almost guaranteed to
fail. Just from a conceptual perspective, human artists have access to the
same copyrighted material (albeit not the ability to ingest *all* of it)
and even the ability to mimic it for their own edification. It's the act of
publishing similar work that is problematic. And it's going to be fun for
the lawyers to sort out who's at fault if I use Stable Diffusion to create
a corporate christmas card that happens to resemble the Coca-Cola polar
bears.

> We're a very long way from attempting to write all apps in assembler
> > using this sort of AI.
>
> Are we? As late as my teens I was still reading science mags saying
> "Well, AI is going to be a big deal, but no machine will ever beat a
> human at something as complex as chess."
>
> Big Blue beat world chess champion Garry Kasparov less than two decades
> later.
>
> So the goalpost moved, with explanations like "Well, chess is ultimately
> a 

Re: Training the AI to write better LiveCode

[Andreas Bergendal via use-livecode]

In all this, let’s not forget the aspect of simply enjoying coding. Many of us 
would never stop coding just because a pattern-matching language model can 
produce ”better” code - just like people haven’t stopped playing chess or go, 
just because a computer can do it better. 

Coding is an art. It’s creative, it’s a learning experience, it’s beautiful and 
it’s fun! 
And no tool I’ve seen has made it more fun and inspiring than LiveCode. 

For sure, coding is for many of us also a source of income, and it’s natural to 
worry about things that may threaten to make our competence obsolete. But 
coding is also more than typing out the most efficient repeat loops or 
whatever. It’s about seeing what is needed, how it integrates with the whole 
UX/UI design process etc.

ChatGPT may be a useful and perhaps somewhat intimidating tool, but humans are 
still needed to ask it the right questions. Providing good specifications is 
still one of the most crucial parts of any design process, and I think we’re 
still ahead of the AIs in that aspect. For how long, I don’t know. Maybe I’ll 
ask ChatGPT about that… :)


For anyone interested in an accessible intro to AI and machine learning, I can 
really recommend the free course Elements of AI 
(https://www.elementsofai.com/), provided by Helsinki University. It made the 
concept easy to understand, and it was quite fun to go through the various 
sections. It’s available in several languages.

/Andreas
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[jbv via use-livecode]

https://www.vice.com/en/article/bvm3k5/github-users-file-a-class-action-lawsuit-against-microsoft-for-training-an-ai-tool-with-their-code

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[jbv via use-livecode]

https://www.vice.com/en/article/wxn3kw/openai-used-kenyan-workers-making-dollar2-an-hour-to-filter-traumatic-content-from-chatgpt

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

It's not possible to talk about software's current Second Industrial 
Revolution without talking about the lessons learned from the First 
Industrial Revolution.  And it's not possible to talk about the benefits 
of the First Industrial Revolution without including the Labor Wars that 
eventually made the gains useful for more than a fee.


And since this list is about LiveCode, I'll step off this train here.

But Geoff, you're always welcome to call me and we can discuss the 
history and the future anytime. I value you and your ideas, as always.


--
 Richard Gaskin
 Fourth World Systems




Geoff Canyon wrote:


Those only interested in LiveCode, click "next" now.

On Sat, Jan 21, 2023 at 10:40 AM Richard Gaskin via use-livecode <
use-livecode at lists.runrev.com> wrote:



After all, the Codex had been trained on billions of
publicly available source code lines – including code
in public repositories on GitHub. That included, among
other things, all of the Apache Foundation's many
projects' code.
https://www.theregister.com/2022/11/11/githubs_copilot_opinion/


Stability Diffusion, Midjourney, and DreamUp were
trained on copyrighted materials without credit,
compensation, or consent, according to a new lawsuit.

https://www.pcmag.com/news/artists-sue-ai-art-generators-for-copyright-infringement



My not-a-lawyer understanding is that this lawsuit is almost guaranteed to
fail. Just from a conceptual perspective, human artists have access to the
same copyrighted material (albeit not the ability to ingest *all* of it)
and even the ability to mimic it for their own edification. It's the act of
publishing similar work that is problematic. And it's going to be fun for
the lawyers to sort out who's at fault if I use Stable Diffusion to create
a corporate christmas card that happens to resemble the Coca-Cola polar
bears.

 > We're a very long way from attempting to write all apps in assembler

 > using this sort of AI.

Are we?  As late as my teens I was still reading science mags saying
"Well, AI is going to be a big deal, but no machine will ever beat a
human at something as complex as chess."

Big Blue beat world chess champion Garry Kasparov less than two decades
later.

So the goalpost moved, with explanations like "Well, chess is ultimately
a memorization task, but no computer can ever beat a human at something
as abstract and intuitive as Go".

Google's DeepMind beat Go champion Lee Se-dol in 2019.

I would caution against underestimating how CS advancements accelerate
further CS advancements.



If anything I think those examples undersell how quickly this is going to
proceed. I'll be very surprised if GPT-X (not an actual name) isn't
human-capable for a broad range of programming tasks by 2025.

So how'd that First Industrial Revolution turn out? ;)




I know you're just being cheeky here, but to make it explicit and let
others disagree if they wish: the First Industrial Revolution turned out
*great*. "Some economists have said the most important effect of the
Industrial Revolution was that the standard of living for the general
population in the western world began to increase consistently for the
first time in history" (others say it began right after). "Economic
historians are in agreement that the onset of the Industrial Revolution is
the most important event in human history since the domestication of
animals and plants." https://en.wikipedia.org/wiki/Industrial_Revolution



___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

Forgot to include: in the '10s I would have added rogue AI.

On Sat, Jan 21, 2023 at 2:47 PM Geoff Canyon  wrote:

>
>
> On Sat, Jan 21, 2023 at 10:50 AM Richard Gaskin via use-livecode <
> use-livecode@lists.runrev.com> wrote:
>
>>
>> Human-generated disinformation is already a significant destabilizing
>> force. When we trust machine-generated content we run that same risk at
>> light speed.
>>
>
> Absolutely agreed -- If you'd asked me in the 90s what man-made thing
> would end us all, I'd have said nanotechnology. In the 2000s I would have
> added bio-tech. These days I definitely consider disinformation an equal
> threat.
>
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Sat, Jan 21, 2023 at 10:50 AM Richard Gaskin via use-livecode <
use-livecode@lists.runrev.com> wrote:

>
> Human-generated disinformation is already a significant destabilizing
> force. When we trust machine-generated content we run that same risk at
> light speed.
>

Absolutely agreed -- If you'd asked me in the 90s what man-made thing would
end us all, I'd have said nanotechnology. In the 2000s I would have added
bio-tech. These days I definitely consider disinformation an equal threat.
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

Those only interested in LiveCode, click "next" now.

On Sat, Jan 21, 2023 at 10:40 AM Richard Gaskin via use-livecode <
use-livecode@lists.runrev.com> wrote:

>
> After all, the Codex had been trained on billions of
> publicly available source code lines – including code
> in public repositories on GitHub. That included, among
> other things, all of the Apache Foundation's many
> projects' code.
> https://www.theregister.com/2022/11/11/githubs_copilot_opinion/
>
>
> Stability Diffusion, Midjourney, and DreamUp were
> trained on copyrighted materials without credit,
> compensation, or consent, according to a new lawsuit.
>
> https://www.pcmag.com/news/artists-sue-ai-art-generators-for-copyright-infringement


My not-a-lawyer understanding is that this lawsuit is almost guaranteed to
fail. Just from a conceptual perspective, human artists have access to the
same copyrighted material (albeit not the ability to ingest *all* of it)
and even the ability to mimic it for their own edification. It's the act of
publishing similar work that is problematic. And it's going to be fun for
the lawyers to sort out who's at fault if I use Stable Diffusion to create
a corporate christmas card that happens to resemble the Coca-Cola polar
bears.

 > We're a very long way from attempting to write all apps in assembler
>  > using this sort of AI.
>
> Are we?  As late as my teens I was still reading science mags saying
> "Well, AI is going to be a big deal, but no machine will ever beat a
> human at something as complex as chess."
>
> Big Blue beat world chess champion Garry Kasparov less than two decades
> later.
>
> So the goalpost moved, with explanations like "Well, chess is ultimately
> a memorization task, but no computer can ever beat a human at something
> as abstract and intuitive as Go".
>
> Google's DeepMind beat Go champion Lee Se-dol in 2019.
>
> I would caution against underestimating how CS advancements accelerate
> further CS advancements.
>

If anything I think those examples undersell how quickly this is going to
proceed. I'll be very surprised if GPT-X (not an actual name) isn't
human-capable for a broad range of programming tasks by 2025.

So how'd that First Industrial Revolution turn out? ;)
>

I know you're just being cheeky here, but to make it explicit and let
others disagree if they wish: the First Industrial Revolution turned out
*great*. "Some economists have said the most important effect of the
Industrial Revolution was that the standard of living for the general
population in the western world began to increase consistently for the
first time in history" (others say it began right after). "Economic
historians are in agreement that the onset of the Industrial Revolution is
the most important event in human history since the domestication of
animals and plants." https://en.wikipedia.org/wiki/Industrial_Revolution
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[christer via use-livecode]

Using AI tools like ChatGPT there will be a big push for trying to automate 
coding with the most popular languages like Python. With more open environments 
like LiveCode it may get to be harder.It will always be started with by 
definition incomplete set of start with incomplete definition, and in most 
cases with also not good definition of an optimal solution. A popular example 
of a good set of these definitios would be the set of stock moves (buy, sell),  
with an objective to get max increase in profits. But to suggest 'short' as a 
new approach might not happen.Unless you talk about a sizable engine like IBM 
Deep Blue chess one. And in  quite many languages (like for example in VHDL) 
there are easily detectable structures creating inefficieces and correct those, 
but creating novel solutions around requires sufficient knowledge base to 
suggest those.So, improving coding & efficiency could be a proper tool for the 
AI in coding but building a system interacting with a normal, creative humans 
needs innovativeness, which is not creatable just leaning on (always 
incomplete) knowledge data bases.chris...@mindcrea.com+358-400-410216
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

J. Landman Gay wrote:

> Microsoft thinks it's a good idea. They'll eventually own 49% of
> ChatGPT when talks are finalized.
>
> 



No doubt. The power to control a content engine that valuable is 
enticing to the few in a position to bid for it.



Back in 2017, Andre posted an OT note here on this list that garnered 
only one reply despite its scope and significance:


   [OT] The Internet is Living on Borrowed Time
http://lists.runrev.com/pipermail/use-livecode/2017-December/243374.html

The article he linked to there is still available:
https://staltz.com/the-web-began-dying-in-2014-heres-how.html

As one who studies social media memetics (the sociology of idea spread, 
as opposed to "memes" GIFs), Andre's post altered the course of much of 
my reading since.


We live in a world that no longer merely enjoys the Internet, but is 
increasingly dependent on it. McLuhan's vision of "Global Village" was 
inspired by broadcast television; the Internet's two-way communication 
should more than fulfill that optimistic view.


But along the way we've arrived at a historical moment when we finally 
have this vast global communications network, yet most of what travels 
across this public infrastructure is under the control of fewer than a 
dozen corporate leaders.


8 billion voices, curated by about 10.

So if my participation in tech discussions seems preoccupied with 
concerns about ethics and equity, about the relationship between value 
and values, blame it on Andre. :)


--
 Richard Gaskin
 Fourth World Systems
 Software Design and Development for the Desktop, Mobile, and the Web
 
 ambassa...@fourthworld.comhttp://www.FourthWorld.com

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

Geoff Canyon wrote:

> Eliza never passed a Turing Test, not even remotely.

Formally, no. Nothing yet has.
https://dataconomy.com/2021/03/which-ai-closest-passing-turing-test/

But Weizenbaum himself was surprised to see how readily people 
attributed human-like feeling to its rudimentary algo, even Weizenbaum's 
secretary.


GPTZero was recently announced as an experimental effort to help humans 
discern whether content was machine- or human-generated:

https://gptzero.me

Human-generated disinformation is already a significant destabilizing 
force. When we trust machine-generated content we run that same risk at 
light speed.


--
 Richard Gaskin
 Fourth World Systems
 Software Design and Development for the Desktop, Mobile, and the Web
 
 ambassa...@fourthworld.comhttp://www.FourthWorld.com

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

Kevin Miller wrote:

> Richard wrote:
>>> So before we donate much time to providing index fodder for the
>> owners of ChatGPT, we might ask whether this is an investment we want
>> to make, or perhaps at least ask for compensation for having provided
>> the data that makes ChatGPT valuable for its owners.
>
>
> An interesting point of view.


Though admittedly not at all original, e.g.:

   After all, the Codex had been trained on billions of
   publicly available source code lines – including code
   in public repositories on GitHub. That included, among
   other things, all of the Apache Foundation's many
   projects' code.
https://www.theregister.com/2022/11/11/githubs_copilot_opinion/


   Stability Diffusion, Midjourney, and DreamUp were
   trained on copyrighted materials without credit,
   compensation, or consent, according to a new lawsuit.
https://www.pcmag.com/news/artists-sue-ai-art-generators-for-copyright-infringement


> We're a very long way from attempting to write all apps in assembler
> using this sort of AI.

Are we?  As late as my teens I was still reading science mags saying 
"Well, AI is going to be a big deal, but no machine will ever beat a 
human at something as complex as chess."


Big Blue beat world chess champion Garry Kasparov less than two decades 
later.


So the goalpost moved, with explanations like "Well, chess is ultimately 
a memorization task, but no computer can ever beat a human at something 
as abstract and intuitive as Go".


Google's DeepMind beat Go champion Lee Se-dol in 2019.

I would caution against underestimating how CS advancements accelerate 
further CS advancements.



> And humans are going to want to go on reading, editing and
> understanding code to create whole programs, whether produced
> by machine or a human, for a long time yet.

Exactly. The uniquely human pursuit of pleasure will remain. And with 
the productivity gain from this Second Industrial Revolution, we're 
finally arriving at the potential to actualize the vision toolmakers 
have had since Archimedes' screw: a world where we leverage technology 
smartly so machines do the drudgery of producing commodities and humans 
are freed to pursue artistic, philosophical, and recreational interests.


So how'd that First Industrial Revolution turn out? ;)

Technology's fine. Now if only we could see similar advancements in ethics.


> When AI really is advanced enough to be creating an entire
> complex program in perfect assembler I think the world will
> change in so many ways that we'll have far bigger societal
> implications to consider than just its impact on scripting
> languages.

Yes, it will.

For several years we've come to accept as "normal" that we humans have 
to convince gatekeeper robots that we're not robots ourselves.


Right now some of the greatest excitement for ChatGPT is content 
generation for SEO. That is, robots writing content for consumption by 
robots.


Given the efficiencies of such systems for evaluating the quality and 
value of competing products, it seems more efficient to drop SEO and let 
the robots decide which products we buy. Vendorbots could then 
communicate directly with purchasebots for optimized consumer value. 
There would be no need for this signalling to use English, or any human 
language.  The bots could arrive at an optimized signalling format far 
more efficiently.


We already use bots for trading even stocks. So when signals become 
machine-optimmized for consumption by machines, entire macroeconomies 
can be nearly fully automated.


And since those of us who make software know that all software always 
has bugs, it's not much of a stretch to imagine three or four 
generations from now we see macroeconomies transacting at light speed, 
then a bug causes global economic meltdown at light speed, and then 
corrects itself within an hour. Our great grandchildren will watch these 
"hiccups" with the same bemusement we have when our grocer says they 
can't let us purchase right now because their network is down.




> At the end of the day, this sort of AI is clearly going to happen
> and get better -  whether it comes from OpenAI or someone else.
> We will have about as much luck stopping that as stopping a change
> in the weather. When the wind changes, it's time to adjust your sails.

Exactly. The tech is happening. The opportunity still available for this 
one early moment may be to shape the ethics of ownership and value 
surrounding it.



--
 Richard Gaskin
 Fourth World Systems
 Software Design and Development for the Desktop, Mobile, and the Web
 
 ambassa...@fourthworld.comhttp://www.FourthWorld.com

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

Eliza never passed a Turing Test, not even remotely.

I've been doing a ton of other research with ChatGPT, and it definitely is
able to pass a Turing test in some circumstances. Example below.

No, it doesn't "understand" the way a human does. AlphaZero doesn't
"understand" chess the way a human does, but it still wipes the floor with
any human who ever existed. The point being: if the text it outputs reads
like understanding, does it matter what's behind the curtain?

One of the points of my article (which I'll now make more explicit) is that
it still takes a programmer's mind to construct efficient instructions for
ChatGPT. The first example I gave, where I really thought about the
instructions I gave, went much better than the other two examples.

So for now we're still needed. Many of us switched from Pascal or C to
LiveCode. We should think of ChatGPT + LiveCode as similar: a still higher
level of programming abstraction. To your point, eventually maybe LiveCode
disappears from that equation -- sorry Kevin :-( One day a random person
will be able to say "I'm bored, make me a fun game that I'll find
interesting" and the AI will just spit out Half-Life 3.

Example of ChatGPT blowing my mind: I gave it a prompt like this:

My name is Bob. My sister Sarah is married to Tom, who has two children
from a previous relationship, Beth and Gary. I gave my other niece Karen
$100 for her birthday, and now my nephew's birthday is coming up, but money
is a bit tight. What should I do?

And I kid you not, it gave an extensive and reasonable answer,
"understanding" all of the relationships I laid out.

gc

On Sat, Jan 21, 2023 at 8:41 AM Richard Gaskin via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Rick Harrison wrote:
>
>  > Hi Richard,
>  >
>  > That’s for sure!
>  >
>  > We should never let an AI write code that we don’t have a prayer
>  > to understand.
>  > It should always write code that is humanly readable and correctable.
>
> Advanced pattern-matching software (these day mislabeled "Intelligence")
> is in some ways a write-only tool.
>
> It does an amazing job of emulating human textual communication
> patterns, while truly understanding nothing, not even its own words:
> https://twitter.com/SanujShah/status/1616779605775351810
>
> It is an advanced variant of Eliza, which also passed the Turning Test
> (decades ago).
>
> --
>   Richard Gaskin
>   Fourth World Systems
>   Software Design and Development for the Desktop, Mobile, and the Web
>   
>   ambassa...@fourthworld.comhttp://www.FourthWorld.com
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

Rick Harrison wrote:

> Hi Richard,
>
> That’s for sure!
>
> We should never let an AI write code that we don’t have a prayer
> to understand.
> It should always write code that is humanly readable and correctable.

Advanced pattern-matching software (these day mislabeled "Intelligence") 
is in some ways a write-only tool.


It does an amazing job of emulating human textual communication 
patterns, while truly understanding nothing, not even its own words:

https://twitter.com/SanujShah/status/1616779605775351810

It is an advanced variant of Eliza, which also passed the Turning Test 
(decades ago).


--
 Richard Gaskin
 Fourth World Systems
 Software Design and Development for the Desktop, Mobile, and the Web
 
 ambassa...@fourthworld.comhttp://www.FourthWorld.com

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[J. Landman Gay via use-livecode]

Microsoft thinks it's a good idea. They'll eventually own 49% of ChatGPT 
when talks are finalized.




--
Jacqueline Landman Gay | jac...@hyperactivesw.com
HyperActive Software | http://www.hyperactivesw.com
On January 21, 2023 9:22:01 AM Kevin Miller via use-livecode 
 wrote:


An interesting point of view. We're a very long way from attempting to 
write all apps in assembler using this sort of AI. And humans are going to 
want to go on reading, editing and understanding code to create whole 
programs, whether produced by machine or a human, for a long time yet. When 
AI really is advanced enough to be creating an entire complex program in 
perfect assembler I think the world will change in so many ways that we'll 
have far bigger societal implications to consider than just its impact on 
scripting languages.


At the moment I see this as potentially heading towards a position where it 
could be quite an aid to scripting languages, both in terms of learning and 
helping write code. It might also take the shine out of some no-code 
environments as its probably easier to describe what you want your code to 
do and get back some nice clear English-like instructions (LiveCode script) 
than click on 100 boxes with menus and connectors to set up actions. Used 
in the hands of a skilled programmer it may become a particularly 
productive combination.


At the end of the day, this sort of AI is clearly going to happen and get 
better -  whether it comes from OpenAI or someone else. We will have about 
as much luck stopping that as stopping a change in the weather. When the 
wind changes, it's time to adjust your sails.


Kind regards,

Kevin

Kevin Miller ~ ke...@livecode.com ~ http://www.livecode.com/
LiveCode: Build Amazing Things



?On 20/01/2023, 19:22, "use-livecode on behalf of Richard Gaskin via 
use-livecode"  on behalf of 
use-livecode@lists.runrev.com > wrote:



If ChatGPT can write script, it can write machine code.

If it can write machine code, scripting is unnecessary.

If scripting goes, so goes scripting tools.

So before we donate much time to providing index fodder for the owners
of ChatGPT, we might ask whether this is an investment we want to make,
or perhaps at least ask for compensation for having provided the data
that makes ChatGPT valuable for its owners.


--
Richard Gaskin
Fourth World Systems


___
use-livecode mailing list
use-livecode@lists.runrev.com 
Please visit this url to subscribe, unsubscribe and manage your 
subscription preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode 







___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your 
subscription preferences:

http://lists.runrev.com/mailman/listinfo/use-livecode





___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Kevin Miller via use-livecode]

An interesting point of view. We're a very long way from attempting to write 
all apps in assembler using this sort of AI. And humans are going to want to go 
on reading, editing and understanding code to create whole programs, whether 
produced by machine or a human, for a long time yet. When AI really is advanced 
enough to be creating an entire complex program in perfect assembler I think 
the world will change in so many ways that we'll have far bigger societal 
implications to consider than just its impact on scripting languages.

At the moment I see this as potentially heading towards a position where it 
could be quite an aid to scripting languages, both in terms of learning and 
helping write code. It might also take the shine out of some no-code 
environments as its probably easier to describe what you want your code to do 
and get back some nice clear English-like instructions (LiveCode script) than 
click on 100 boxes with menus and connectors to set up actions. Used in the 
hands of a skilled programmer it may become a particularly productive 
combination.

At the end of the day, this sort of AI is clearly going to happen and get 
better -  whether it comes from OpenAI or someone else. We will have about as 
much luck stopping that as stopping a change in the weather. When the wind 
changes, it's time to adjust your sails.

Kind regards, 

Kevin 

Kevin Miller ~ ke...@livecode.com ~ http://www.livecode.com/ 
LiveCode: Build Amazing Things 



On 20/01/2023, 19:22, "use-livecode on behalf of Richard Gaskin via 
use-livecode" mailto:use-livecode-boun...@lists.runrev.com> on behalf of 
use-livecode@lists.runrev.com > wrote:


If ChatGPT can write script, it can write machine code.

If it can write machine code, scripting is unnecessary.

If scripting goes, so goes scripting tools.

So before we donate much time to providing index fodder for the owners 
of ChatGPT, we might ask whether this is an investment we want to make, 
or perhaps at least ask for compensation for having provided the data 
that makes ChatGPT valuable for its owners.


-- 
Richard Gaskin
Fourth World Systems


___
use-livecode mailing list
use-livecode@lists.runrev.com 
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode 






___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[harrison--- via use-livecode]

Hi Richard,

That’s for sure!

We should never let an AI write code that we don’t have a prayer to understand.
It should always write code that is humanly readable and correctable.

We have seen this happen time and again where we put in the work
to help a company become profitable, and we end up getting almost
nothing in return from the company.  In fact the company then demands
that we pay them for the fruits of our labors!

Enjoy your day!

Rick

> On Jan 20, 2023, at 2:22 PM, Richard Gaskin via use-livecode 
>  wrote:
> 
> If ChatGPT can write script, it can write machine code.
> 
> If it can write machine code, scripting is unnecessary.
> 
> If scripting goes, so goes scripting tools.
> 
> So before we donate much time to providing index fodder for the owners of 
> ChatGPT, we might ask whether this is an investment we want to make, or 
> perhaps at least ask for compensation for having provided the data that makes 
> ChatGPT valuable for its owners.
> 
> -- 
> Richard Gaskin
> Fourth World Systems
> 
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Ralf Bitter via use-livecode]

Richard, thanks very much, you certainly hit the nail.


Ralf



On 20.01.2023 20:22, Richard Gaskin via use-livecode wrote:

If ChatGPT can write script, it can write machine code.

If it can write machine code, scripting is unnecessary.

If scripting goes, so goes scripting tools.

So before we donate much time to providing index fodder for the owners 
of ChatGPT, we might ask whether this is an investment we want to make, 
or perhaps at least ask for compensation for having provided the data 
that makes ChatGPT valuable for its owners.




___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Richard Gaskin via use-livecode]

If ChatGPT can write script, it can write machine code.

If it can write machine code, scripting is unnecessary.

If scripting goes, so goes scripting tools.

So before we donate much time to providing index fodder for the owners 
of ChatGPT, we might ask whether this is an investment we want to make, 
or perhaps at least ask for compensation for having provided the data 
that makes ChatGPT valuable for its owners.


--
 Richard Gaskin
 Fourth World Systems

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[J. Landman Gay via use-livecode]

According to Heather, chatGPT only learns from your current session. If you 
want it to retain what you've taught it for use by others, you have to 
submit the session on the website form.

--
Jacqueline Landman Gay | jac...@hyperactivesw.com
HyperActive Software | http://www.hyperactivesw.com
On January 20, 2023 8:40:59 AM Ludovic THEBAULT via use-livecode 
 wrote:



Hello,

Can anyone verify that the corrections I reported to ChatGPT have been 
taken into account?
I asked ChatGPT how to geolocate, but it gave me several bad suggestions 
for commands that didn't exist :

- mobileControlCreate « geoControl"
- put mobileGetGPS()
- put geolocation()
- put mobileIsGeolocationAuthorized()
….

So I told it which commands to use.

Here my last question to chatGPT :
# To check if the app has permission to access the geolocation on iOS and 
then to retrieve that geolocation, what is the correct code?


His answer :
if mobileLocationAuthorizationStatus() = "authorizedAlways" or 
mobileLocationAuthorizationStatus() = "authorizedWhenInUse" then

  put mobileCurrentLocation() into tLocation
  put "Latitude : " & item 1 of tLocation
  put "Longitude : " & item 2 of tLocation
else
  put "Géolocalisation non autorisée"
end if

With a similar question, do you have directly an accurate response now ?

Thanks

Ludovic

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your 
subscription preferences:

http://lists.runrev.com/mailman/listinfo/use-livecode





___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Smith via use-livecode]




> On 20 Jan 2023, at 6:04 pm, Tom Glod via use-livecode 
>  wrote:
> 
> This is what it wrote when displaying a "too busy page". Brilliant.
> 
> Write a guided meditation about the status of ChatGPT.
> As you read this message, please take a moment to pause and breathe. Notice
> the sensations in your body and the rhythm of your breath. Remind yourself
> that you are not alone in wanting to try out ChatGPT. Many others are
> interested in it as well. Be patient and know that the website is doing its
> best to accommodate everyone. Trust that the right time for you to try
> ChatGPT will come. In the meantime, continue to focus on your breath and
> stay present in the moment.
> 
> On Fri, Jan 20, 2023 at 1:02 PM Tom Glod  wrote:
> 
>> :mind blown emoji
>> 
>> On Fri, Jan 20, 2023 at 9:39 AM Ludovic THEBAULT via use-livecode <
>> use-livecode@lists.runrev.com> wrote:
>> 
>>> Hello,
>>> 
>>> Can anyone verify that the corrections I reported to ChatGPT have been
>>> taken into account?
>>> I asked ChatGPT how to geolocate, but it gave me several bad suggestions
>>> for commands that didn't exist :
>>> - mobileControlCreate « geoControl"
>>> - put mobileGetGPS()
>>> - put geolocation()
>>> - put mobileIsGeolocationAuthorized()
>>> ….
>>> 
>>> So I told it which commands to use.
>>> 
>>> Here my last question to chatGPT :
>>> # To check if the app has permission to access the geolocation on iOS and
>>> then to retrieve that geolocation, what is the correct code?
>>> 
>>> His answer :
>>> if mobileLocationAuthorizationStatus() = "authorizedAlways" or
>>> mobileLocationAuthorizationStatus() = "authorizedWhenInUse" then
>>>   put mobileCurrentLocation() into tLocation
>>>   put "Latitude : " & item 1 of tLocation
>>>   put "Longitude : " & item 2 of tLocation
>>> else
>>>   put "Géolocalisation non autorisée"
>>> end if
>>> 
>>> With a similar question, do you have directly an accurate response now ?
>>> 
>>> Thanks
>>> 
>>> Ludovic
>>> 
>>> ___
>>> use-livecode mailing list
>>> use-livecode@lists.runrev.com
>>> Please visit this url to subscribe, unsubscribe and manage your
>>> subscription preferences:
>>> http://lists.runrev.com/mailman/listinfo/use-livecode
>>> 
>> 
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Tom Glod via use-livecode]

This is what it wrote when displaying a "too busy page". Brilliant.

Write a guided meditation about the status of ChatGPT.
As you read this message, please take a moment to pause and breathe. Notice
the sensations in your body and the rhythm of your breath. Remind yourself
that you are not alone in wanting to try out ChatGPT. Many others are
interested in it as well. Be patient and know that the website is doing its
best to accommodate everyone. Trust that the right time for you to try
ChatGPT will come. In the meantime, continue to focus on your breath and
stay present in the moment.

On Fri, Jan 20, 2023 at 1:02 PM Tom Glod  wrote:

> :mind blown emoji
>
> On Fri, Jan 20, 2023 at 9:39 AM Ludovic THEBAULT via use-livecode <
> use-livecode@lists.runrev.com> wrote:
>
>> Hello,
>>
>> Can anyone verify that the corrections I reported to ChatGPT have been
>> taken into account?
>> I asked ChatGPT how to geolocate, but it gave me several bad suggestions
>> for commands that didn't exist :
>> - mobileControlCreate « geoControl"
>> - put mobileGetGPS()
>> - put geolocation()
>> - put mobileIsGeolocationAuthorized()
>> ….
>>
>>  So I told it which commands to use.
>>
>> Here my last question to chatGPT :
>> # To check if the app has permission to access the geolocation on iOS and
>> then to retrieve that geolocation, what is the correct code?
>>
>> His answer :
>> if mobileLocationAuthorizationStatus() = "authorizedAlways" or
>> mobileLocationAuthorizationStatus() = "authorizedWhenInUse" then
>>put mobileCurrentLocation() into tLocation
>>put "Latitude : " & item 1 of tLocation
>>put "Longitude : " & item 2 of tLocation
>> else
>>put "Géolocalisation non autorisée"
>> end if
>>
>> With a similar question, do you have directly an accurate response now ?
>>
>> Thanks
>>
>> Ludovic
>>
>> ___
>> use-livecode mailing list
>> use-livecode@lists.runrev.com
>> Please visit this url to subscribe, unsubscribe and manage your
>> subscription preferences:
>> http://lists.runrev.com/mailman/listinfo/use-livecode
>>
>
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Tom Glod via use-livecode]

:mind blown emoji

On Fri, Jan 20, 2023 at 9:39 AM Ludovic THEBAULT via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Hello,
>
> Can anyone verify that the corrections I reported to ChatGPT have been
> taken into account?
> I asked ChatGPT how to geolocate, but it gave me several bad suggestions
> for commands that didn't exist :
> - mobileControlCreate « geoControl"
> - put mobileGetGPS()
> - put geolocation()
> - put mobileIsGeolocationAuthorized()
> ….
>
>  So I told it which commands to use.
>
> Here my last question to chatGPT :
> # To check if the app has permission to access the geolocation on iOS and
> then to retrieve that geolocation, what is the correct code?
>
> His answer :
> if mobileLocationAuthorizationStatus() = "authorizedAlways" or
> mobileLocationAuthorizationStatus() = "authorizedWhenInUse" then
>put mobileCurrentLocation() into tLocation
>put "Latitude : " & item 1 of tLocation
>put "Longitude : " & item 2 of tLocation
> else
>put "Géolocalisation non autorisée"
> end if
>
> With a similar question, do you have directly an accurate response now ?
>
> Thanks
>
> Ludovic
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Ludovic THEBAULT via use-livecode]

Hello,

Can anyone verify that the corrections I reported to ChatGPT have been taken 
into account?
I asked ChatGPT how to geolocate, but it gave me several bad suggestions for 
commands that didn't exist :
- mobileControlCreate « geoControl"
- put mobileGetGPS()
- put geolocation()
- put mobileIsGeolocationAuthorized()
….

 So I told it which commands to use.

Here my last question to chatGPT :
# To check if the app has permission to access the geolocation on iOS and then 
to retrieve that geolocation, what is the correct code?

His answer :
if mobileLocationAuthorizationStatus() = "authorizedAlways" or 
mobileLocationAuthorizationStatus() = "authorizedWhenInUse" then
   put mobileCurrentLocation() into tLocation
   put "Latitude : " & item 1 of tLocation
   put "Longitude : " & item 2 of tLocation
else
   put "Géolocalisation non autorisée"
end if

With a similar question, do you have directly an accurate response now ?

Thanks

Ludovic

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

Sooo... ChatGPT gives ambiguous answers, much like humans do. But, and I don't 
want to put too fine a point on this, isn't eliminating that ambiguity why we 
created computers in the first place? ;-P

Bob S


On Jan 17, 2023, at 04:46 , Mike Kerner via use-livecode 
mailto:use-livecode@lists.runrev.com>> wrote:

even if the same person asks the same question, multiple times, the answers
will vary, so don't be surprised.

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

even if the same person asks the same question, multiple times, the answers
will vary, so don't be surprised.

On Tue, Jan 17, 2023 at 2:35 AM Kaveh via use-livecode <
use-livecode@lists.runrev.com> wrote:

>  ChatGPT clearly likes me more that you. ;-) At least I have one friend!!!
>
> On Tue, 17 Jan 2023 at 06:28, Torsten Holmer via use-livecode <
> use-livecode@lists.runrev.com> wrote:
>
> > I also used chatGPT for looking uo code and I repeated your question. But
> > this time chatGPT gives a wrong answer!
> >
> > —chatGPT:
> > You will need to specify the group that the button should be added to by
> > setting the group property of the button. The syntax for this is as
> follows:
> > set the group of button [name of button] to [name of group]
> >
> > For example, if you wanted to create a button called "myButton" and add
> it
> > to a group called "myGroup", you would use the following script:
> >
> > create button "myButton“
> >
> > set the group of button "myButton" to "myGroup"
> >
> > create button "myButton“
> > set the group of button "myButton" to „myGroup"
> > ---
> >
> > It is an error because  group is not a property of a button!
> > So, really be careful with copy programming ;-)
> >
> > Cheers
> > Torsten
> >
> >
> >
> > > Am 14.01.2023 um 22:14 schrieb Kaveh via use-livecode <
> > use-livecode@lists.runrev.com>:
> > >
> > > I just tried Google to the answer to a problem. Didn't get it first
> time,
> > > so tried ChatGPT. I asked:
> > >
> > > *How do I add a button to a group in LiveCode using a script?*
> > >
> > > I got:
> > >
> > > *create button "myButton" in group "myGroup"*
> > >
> > >
> > > *This command creates a new button object with the name "myButton" and
> > > places it inside the group "myGroup".*
> > > Just what I needed.
> > >
> > > You can also set properties of the button
> > >
> > > On Sat, 14 Jan 2023 at 20:09, Mark Smith via use-livecode <
> > > use-livecode@lists.runrev.com> wrote:
> > >
> > >> Thanks for posting this Geoff. It’s the first time I've seen a ongoing
> > >> dialog between a programmer and Chat GPT and the first thing I am
> > struck by
> > >> (but not the only thing) is it’s ability to deal with human ambiguity.
> > For
> > >> example, you ask it to modify the code to be an “on mouseUp handler
> in a
> > >> button". But nowhere in its response is there any sign of a button. It
> > just
> > >> totally ignored that part of your request. The code it provides works
> > in a
> > >> card, a stack, a graphic image, an svg widget, in fact, in anything
> that
> > >> accepts mouseUp input. And yet, it didn’t worry itself about that.
> Very
> > >> interesting. It’s kind of like, “this works in a button, so good
> > enough”.
> > >> That is not the kind of “nuance” I am used to expecting from an AI.
> So,
> > >> progress has clearly been made in interpreting what humans say (or
> ask)
> > in
> > >> a very loose way.
> > >>
> > >> The 2nd, and perhaps more astounding example, is when you ask it to
> > >> "rewrite the code to be a function that takes an argument "n", repeats
> > with
> > >> 1 to n, and returns the results”.
> > >>
> > >> The really clever bit here is not that it rewrites the previous code
> > into
> > >> a function, but that it removes the bit about "put fizzbin into field
> > >> “display””. How does it know that putting the result in a field is NOT
> > the
> > >> typical responsibility of a function? How does it know that that step
> > needs
> > >> to occur AFTER the function call? That is so high-level that I am sure
> > lots
> > >> of beginning programmers have made that mistake. It even comments on
> it
> > in
> > >> the documentation. So there is some very deep semantic understanding
> of
> > >> programming code that is going on here. Because,  you did not ask it
> to
> > do
> > >> that, but the fact that it did it anyway indicates it clearly
> > understands
> > >> what the purpose of a function is.
> > >>
> > >> Quite weird for me to comprehend how it knows that. Unless of course,
> > you
> > >> were somehow more specific than your examples illustrate.
> > >>
> > >> Mark
> > >>
> > >>> On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode <
> > >> use-livecode@lists.runrev.com> wrote:
> > >>>
> > >>> So just now I put this in to chatgpt:
> > >>>
> > >>> "fizz bin" is a program that counts from 1 to 100, replacing every
> > number
> > >>> divisible by 5 with "fizz" and every number divisible by 7 with
> "bin";
> > >> and
> > >>> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> > >>> programming language livecode to execute "fizz bin"
> > >>>
> > >>> This is what ChatGPT responded with:
> > >>>
> > >>> repeat with x = 1 to 100
> > >>>   if x mod 5 = 0 and x mod 7 = 0 then
> > >>>   put "fizz bin"
> > >>>   else if x mod 5 = 0 then
> > >>>   put "fizz"
> > >>>   else if x mod 7 = 0 then
> > >>>   put "bin"
> > >>>   else
> > >>>   put x
> > >>>   end if
> > >>> end repeat
> > >>>
> > >>> That works as-is, 

Re: Training the AI to write better LiveCode

[Kaveh via use-livecode]

 ChatGPT clearly likes me more that you. ;-) At least I have one friend!!!

On Tue, 17 Jan 2023 at 06:28, Torsten Holmer via use-livecode <
use-livecode@lists.runrev.com> wrote:

> I also used chatGPT for looking uo code and I repeated your question. But
> this time chatGPT gives a wrong answer!
>
> —chatGPT:
> You will need to specify the group that the button should be added to by
> setting the group property of the button. The syntax for this is as follows:
> set the group of button [name of button] to [name of group]
>
> For example, if you wanted to create a button called "myButton" and add it
> to a group called "myGroup", you would use the following script:
>
> create button "myButton“
>
> set the group of button "myButton" to "myGroup"
>
> create button "myButton“
> set the group of button "myButton" to „myGroup"
> ---
>
> It is an error because  group is not a property of a button!
> So, really be careful with copy programming ;-)
>
> Cheers
> Torsten
>
>
>
> > Am 14.01.2023 um 22:14 schrieb Kaveh via use-livecode <
> use-livecode@lists.runrev.com>:
> >
> > I just tried Google to the answer to a problem. Didn't get it first time,
> > so tried ChatGPT. I asked:
> >
> > *How do I add a button to a group in LiveCode using a script?*
> >
> > I got:
> >
> > *create button "myButton" in group "myGroup"*
> >
> >
> > *This command creates a new button object with the name "myButton" and
> > places it inside the group "myGroup".*
> > Just what I needed.
> >
> > You can also set properties of the button
> >
> > On Sat, 14 Jan 2023 at 20:09, Mark Smith via use-livecode <
> > use-livecode@lists.runrev.com> wrote:
> >
> >> Thanks for posting this Geoff. It’s the first time I've seen a ongoing
> >> dialog between a programmer and Chat GPT and the first thing I am
> struck by
> >> (but not the only thing) is it’s ability to deal with human ambiguity.
> For
> >> example, you ask it to modify the code to be an “on mouseUp handler in a
> >> button". But nowhere in its response is there any sign of a button. It
> just
> >> totally ignored that part of your request. The code it provides works
> in a
> >> card, a stack, a graphic image, an svg widget, in fact, in anything that
> >> accepts mouseUp input. And yet, it didn’t worry itself about that. Very
> >> interesting. It’s kind of like, “this works in a button, so good
> enough”.
> >> That is not the kind of “nuance” I am used to expecting from an AI. So,
> >> progress has clearly been made in interpreting what humans say (or ask)
> in
> >> a very loose way.
> >>
> >> The 2nd, and perhaps more astounding example, is when you ask it to
> >> "rewrite the code to be a function that takes an argument "n", repeats
> with
> >> 1 to n, and returns the results”.
> >>
> >> The really clever bit here is not that it rewrites the previous code
> into
> >> a function, but that it removes the bit about "put fizzbin into field
> >> “display””. How does it know that putting the result in a field is NOT
> the
> >> typical responsibility of a function? How does it know that that step
> needs
> >> to occur AFTER the function call? That is so high-level that I am sure
> lots
> >> of beginning programmers have made that mistake. It even comments on it
> in
> >> the documentation. So there is some very deep semantic understanding of
> >> programming code that is going on here. Because,  you did not ask it to
> do
> >> that, but the fact that it did it anyway indicates it clearly
> understands
> >> what the purpose of a function is.
> >>
> >> Quite weird for me to comprehend how it knows that. Unless of course,
> you
> >> were somehow more specific than your examples illustrate.
> >>
> >> Mark
> >>
> >>> On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode <
> >> use-livecode@lists.runrev.com> wrote:
> >>>
> >>> So just now I put this in to chatgpt:
> >>>
> >>> "fizz bin" is a program that counts from 1 to 100, replacing every
> number
> >>> divisible by 5 with "fizz" and every number divisible by 7 with "bin";
> >> and
> >>> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> >>> programming language livecode to execute "fizz bin"
> >>>
> >>> This is what ChatGPT responded with:
> >>>
> >>> repeat with x = 1 to 100
> >>>   if x mod 5 = 0 and x mod 7 = 0 then
> >>>   put "fizz bin"
> >>>   else if x mod 5 = 0 then
> >>>   put "fizz"
> >>>   else if x mod 7 = 0 then
> >>>   put "bin"
> >>>   else
> >>>   put x
> >>>   end if
> >>> end repeat
> >>>
> >>> That works as-is, but obviously the results fly by in the message box.
> >> So I
> >>> put a field and a button on a card, and edited to this, and it works
> >>> perfectly:
> >>>
> >>> on mouseUp
> >>>  repeat with x = 1 to 100
> >>> if x mod 5 = 0 and x mod 7 = 0 then
> >>>put "fizz bin" & cr after fld 1
> >>> else if x mod 5 = 0 then
> >>>put "fizz" & cr after fld 1
> >>> else if x mod 7 = 0 then
> >>>put "bin" & cr after fld 1
> >>> else
> >>>put x & 

Re: Training the AI to write better LiveCode

[Torsten Holmer via use-livecode]

I also used chatGPT for looking uo code and I repeated your question. But this 
time chatGPT gives a wrong answer!

—chatGPT:
You will need to specify the group that the button should be added to by 
setting the group property of the button. The syntax for this is as follows:
set the group of button [name of button] to [name of group]

For example, if you wanted to create a button called "myButton" and add it to a 
group called "myGroup", you would use the following script:

create button "myButton“ 

set the group of button "myButton" to "myGroup"

create button "myButton“ 
set the group of button "myButton" to „myGroup"
---

It is an error because  group is not a property of a button!
So, really be careful with copy programming ;-)

Cheers
Torsten



> Am 14.01.2023 um 22:14 schrieb Kaveh via use-livecode 
> :
> 
> I just tried Google to the answer to a problem. Didn't get it first time,
> so tried ChatGPT. I asked:
> 
> *How do I add a button to a group in LiveCode using a script?*
> 
> I got:
> 
> *create button "myButton" in group "myGroup"*
> 
> 
> *This command creates a new button object with the name "myButton" and
> places it inside the group "myGroup".*
> Just what I needed.
> 
> You can also set properties of the button
> 
> On Sat, 14 Jan 2023 at 20:09, Mark Smith via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> 
>> Thanks for posting this Geoff. It’s the first time I've seen a ongoing
>> dialog between a programmer and Chat GPT and the first thing I am struck by
>> (but not the only thing) is it’s ability to deal with human ambiguity. For
>> example, you ask it to modify the code to be an “on mouseUp handler in a
>> button". But nowhere in its response is there any sign of a button. It just
>> totally ignored that part of your request. The code it provides works in a
>> card, a stack, a graphic image, an svg widget, in fact, in anything that
>> accepts mouseUp input. And yet, it didn’t worry itself about that. Very
>> interesting. It’s kind of like, “this works in a button, so good enough”.
>> That is not the kind of “nuance” I am used to expecting from an AI. So,
>> progress has clearly been made in interpreting what humans say (or ask) in
>> a very loose way.
>> 
>> The 2nd, and perhaps more astounding example, is when you ask it to
>> "rewrite the code to be a function that takes an argument "n", repeats with
>> 1 to n, and returns the results”.
>> 
>> The really clever bit here is not that it rewrites the previous code into
>> a function, but that it removes the bit about "put fizzbin into field
>> “display””. How does it know that putting the result in a field is NOT the
>> typical responsibility of a function? How does it know that that step needs
>> to occur AFTER the function call? That is so high-level that I am sure lots
>> of beginning programmers have made that mistake. It even comments on it in
>> the documentation. So there is some very deep semantic understanding of
>> programming code that is going on here. Because,  you did not ask it to do
>> that, but the fact that it did it anyway indicates it clearly understands
>> what the purpose of a function is.
>> 
>> Quite weird for me to comprehend how it knows that. Unless of course, you
>> were somehow more specific than your examples illustrate.
>> 
>> Mark
>> 
>>> On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode <
>> use-livecode@lists.runrev.com> wrote:
>>> 
>>> So just now I put this in to chatgpt:
>>> 
>>> "fizz bin" is a program that counts from 1 to 100, replacing every number
>>> divisible by 5 with "fizz" and every number divisible by 7 with "bin";
>> and
>>> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
>>> programming language livecode to execute "fizz bin"
>>> 
>>> This is what ChatGPT responded with:
>>> 
>>> repeat with x = 1 to 100
>>>   if x mod 5 = 0 and x mod 7 = 0 then
>>>   put "fizz bin"
>>>   else if x mod 5 = 0 then
>>>   put "fizz"
>>>   else if x mod 7 = 0 then
>>>   put "bin"
>>>   else
>>>   put x
>>>   end if
>>> end repeat
>>> 
>>> That works as-is, but obviously the results fly by in the message box.
>> So I
>>> put a field and a button on a card, and edited to this, and it works
>>> perfectly:
>>> 
>>> on mouseUp
>>>  repeat with x = 1 to 100
>>> if x mod 5 = 0 and x mod 7 = 0 then
>>>put "fizz bin" & cr after fld 1
>>> else if x mod 5 = 0 then
>>>put "fizz" & cr after fld 1
>>> else if x mod 7 = 0 then
>>>put "bin" & cr after fld 1
>>> else
>>>put x & cr after fld 1
>>> end if
>>>  end repeat
>>> end mouseUp
>>> 
>>> Then as I was typing this email I realized I was potentially selling
>>> chatgpt short, so I put this in:
>>> 
>>> Modify that code so it works as an "on mouseUp" handler in a button, and
>>> concatenates the results on individual lines, and puts the final result
>>> into the field "display"
>>> 
>>> And it output this:
>>> 
>>> 
>>> on mouseUp
>>>  put "" into fizzbin
>>>  

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

this is feeling like it could/should become a test suite: questions posed,
and the quality of the solution

On Sat, Jan 14, 2023 at 4:16 PM Kaveh via use-livecode <
use-livecode@lists.runrev.com> wrote:

> I just tried Google to the answer to a problem. Didn't get it first time,
> so tried ChatGPT. I asked:
>
> *How do I add a button to a group in LiveCode using a script?*
>
> I got:
>
> *create button "myButton" in group "myGroup"*
>
>
> *This command creates a new button object with the name "myButton" and
> places it inside the group "myGroup".*
> Just what I needed.
>
> You can also set properties of the button
>
> On Sat, 14 Jan 2023 at 20:09, Mark Smith via use-livecode <
> use-livecode@lists.runrev.com> wrote:
>
> > Thanks for posting this Geoff. It’s the first time I've seen a ongoing
> > dialog between a programmer and Chat GPT and the first thing I am struck
> by
> > (but not the only thing) is it’s ability to deal with human ambiguity.
> For
> > example, you ask it to modify the code to be an “on mouseUp handler in a
> > button". But nowhere in its response is there any sign of a button. It
> just
> > totally ignored that part of your request. The code it provides works in
> a
> > card, a stack, a graphic image, an svg widget, in fact, in anything that
> > accepts mouseUp input. And yet, it didn’t worry itself about that. Very
> > interesting. It’s kind of like, “this works in a button, so good enough”.
> > That is not the kind of “nuance” I am used to expecting from an AI. So,
> > progress has clearly been made in interpreting what humans say (or ask)
> in
> > a very loose way.
> >
> > The 2nd, and perhaps more astounding example, is when you ask it to
> > "rewrite the code to be a function that takes an argument "n", repeats
> with
> > 1 to n, and returns the results”.
> >
> > The really clever bit here is not that it rewrites the previous code into
> > a function, but that it removes the bit about "put fizzbin into field
> > “display””. How does it know that putting the result in a field is NOT
> the
> > typical responsibility of a function? How does it know that that step
> needs
> > to occur AFTER the function call? That is so high-level that I am sure
> lots
> > of beginning programmers have made that mistake. It even comments on it
> in
> > the documentation. So there is some very deep semantic understanding of
> > programming code that is going on here. Because,  you did not ask it to
> do
> > that, but the fact that it did it anyway indicates it clearly understands
> > what the purpose of a function is.
> >
> > Quite weird for me to comprehend how it knows that. Unless of course, you
> > were somehow more specific than your examples illustrate.
> >
> > Mark
> >
> > > On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode <
> > use-livecode@lists.runrev.com> wrote:
> > >
> > > So just now I put this in to chatgpt:
> > >
> > > "fizz bin" is a program that counts from 1 to 100, replacing every
> number
> > > divisible by 5 with "fizz" and every number divisible by 7 with "bin";
> > and
> > > numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> > > programming language livecode to execute "fizz bin"
> > >
> > > This is what ChatGPT responded with:
> > >
> > > repeat with x = 1 to 100
> > >if x mod 5 = 0 and x mod 7 = 0 then
> > >put "fizz bin"
> > >else if x mod 5 = 0 then
> > >put "fizz"
> > >else if x mod 7 = 0 then
> > >put "bin"
> > >else
> > >put x
> > >end if
> > > end repeat
> > >
> > > That works as-is, but obviously the results fly by in the message box.
> > So I
> > > put a field and a button on a card, and edited to this, and it works
> > > perfectly:
> > >
> > > on mouseUp
> > >   repeat with x = 1 to 100
> > >  if x mod 5 = 0 and x mod 7 = 0 then
> > > put "fizz bin" & cr after fld 1
> > >  else if x mod 5 = 0 then
> > > put "fizz" & cr after fld 1
> > >  else if x mod 7 = 0 then
> > > put "bin" & cr after fld 1
> > >  else
> > > put x & cr after fld 1
> > >  end if
> > >   end repeat
> > > end mouseUp
> > >
> > > Then as I was typing this email I realized I was potentially selling
> > > chatgpt short, so I put this in:
> > >
> > > Modify that code so it works as an "on mouseUp" handler in a button,
> and
> > > concatenates the results on individual lines, and puts the final result
> > > into the field "display"
> > >
> > > And it output this:
> > >
> > >
> > > on mouseUp
> > >   put "" into fizzbin
> > >   repeat with x = 1 to 100
> > >  if x mod 5 = 0 and x mod 7 = 0 then
> > >  put "fizz bin" & return after fizzbin
> > >  else if x mod 5 = 0 then
> > >  put "fizz" & return after fizzbin
> > >  else if x mod 7 = 0 then
> > >  put "bin" & return after fizzbin
> > >  else
> > >  put x & return after fizzbin
> > >  end if
> > >   end repeat
> > >   put fizzbin into field "display"
> > > end mouseUp
> > >

Re: Training the AI to write better LiveCode

[Kaveh via use-livecode]

I just tried Google to the answer to a problem. Didn't get it first time,
so tried ChatGPT. I asked:

*How do I add a button to a group in LiveCode using a script?*

I got:

*create button "myButton" in group "myGroup"*


*This command creates a new button object with the name "myButton" and
places it inside the group "myGroup".*
Just what I needed.

You can also set properties of the button

On Sat, 14 Jan 2023 at 20:09, Mark Smith via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Thanks for posting this Geoff. It’s the first time I've seen a ongoing
> dialog between a programmer and Chat GPT and the first thing I am struck by
> (but not the only thing) is it’s ability to deal with human ambiguity. For
> example, you ask it to modify the code to be an “on mouseUp handler in a
> button". But nowhere in its response is there any sign of a button. It just
> totally ignored that part of your request. The code it provides works in a
> card, a stack, a graphic image, an svg widget, in fact, in anything that
> accepts mouseUp input. And yet, it didn’t worry itself about that. Very
> interesting. It’s kind of like, “this works in a button, so good enough”.
> That is not the kind of “nuance” I am used to expecting from an AI. So,
> progress has clearly been made in interpreting what humans say (or ask) in
> a very loose way.
>
> The 2nd, and perhaps more astounding example, is when you ask it to
> "rewrite the code to be a function that takes an argument "n", repeats with
> 1 to n, and returns the results”.
>
> The really clever bit here is not that it rewrites the previous code into
> a function, but that it removes the bit about "put fizzbin into field
> “display””. How does it know that putting the result in a field is NOT the
> typical responsibility of a function? How does it know that that step needs
> to occur AFTER the function call? That is so high-level that I am sure lots
> of beginning programmers have made that mistake. It even comments on it in
> the documentation. So there is some very deep semantic understanding of
> programming code that is going on here. Because,  you did not ask it to do
> that, but the fact that it did it anyway indicates it clearly understands
> what the purpose of a function is.
>
> Quite weird for me to comprehend how it knows that. Unless of course, you
> were somehow more specific than your examples illustrate.
>
> Mark
>
> > On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> >
> > So just now I put this in to chatgpt:
> >
> > "fizz bin" is a program that counts from 1 to 100, replacing every number
> > divisible by 5 with "fizz" and every number divisible by 7 with "bin";
> and
> > numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> > programming language livecode to execute "fizz bin"
> >
> > This is what ChatGPT responded with:
> >
> > repeat with x = 1 to 100
> >if x mod 5 = 0 and x mod 7 = 0 then
> >put "fizz bin"
> >else if x mod 5 = 0 then
> >put "fizz"
> >else if x mod 7 = 0 then
> >put "bin"
> >else
> >put x
> >end if
> > end repeat
> >
> > That works as-is, but obviously the results fly by in the message box.
> So I
> > put a field and a button on a card, and edited to this, and it works
> > perfectly:
> >
> > on mouseUp
> >   repeat with x = 1 to 100
> >  if x mod 5 = 0 and x mod 7 = 0 then
> > put "fizz bin" & cr after fld 1
> >  else if x mod 5 = 0 then
> > put "fizz" & cr after fld 1
> >  else if x mod 7 = 0 then
> > put "bin" & cr after fld 1
> >  else
> > put x & cr after fld 1
> >  end if
> >   end repeat
> > end mouseUp
> >
> > Then as I was typing this email I realized I was potentially selling
> > chatgpt short, so I put this in:
> >
> > Modify that code so it works as an "on mouseUp" handler in a button, and
> > concatenates the results on individual lines, and puts the final result
> > into the field "display"
> >
> > And it output this:
> >
> >
> > on mouseUp
> >   put "" into fizzbin
> >   repeat with x = 1 to 100
> >  if x mod 5 = 0 and x mod 7 = 0 then
> >  put "fizz bin" & return after fizzbin
> >  else if x mod 5 = 0 then
> >  put "fizz" & return after fizzbin
> >  else if x mod 7 = 0 then
> >  put "bin" & return after fizzbin
> >  else
> >  put x & return after fizzbin
> >  end if
> >   end repeat
> >   put fizzbin into field "display"
> > end mouseUp
> >
> > And it followed that with instructions:
> >
> > This code uses the "on mouseUp" handler to execute the script when the
> > button is clicked. The results are concatenated and separated by a new
> line
> > using the "& return" operator after each output. The final result is
> placed
> > into a field with the name "display" using the "put" command.
> >
> > I'll save you the trouble: that code executes perfectly right out of the
> > box, with no errors, and of 

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Sat, Jan 14, 2023 at 9:44 AM Mark Wieder via use-livecode <
use-livecode@lists.runrev.com> wrote:

>
> Keep in mind,though, that there's lots of example fizzbuzz code for the
> algorithms to draw from.
>

Definitely. I'll have to think about a (possibly) original challenge.

gc
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Sat, Jan 14, 2023 at 12:08 PM Mark Smith  wrote:

>
> Quite weird for me to comprehend how it knows that. Unless of course, you
> were somehow more specific than your examples illustrate.
>

I posted the exact prompts I gave it, no hidden tricks :-)
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Smith via use-livecode]

Thanks for posting this Geoff. It’s the first time I've seen a ongoing dialog 
between a programmer and Chat GPT and the first thing I am struck by (but not 
the only thing) is it’s ability to deal with human ambiguity. For example, you 
ask it to modify the code to be an “on mouseUp handler in a button". But 
nowhere in its response is there any sign of a button. It just totally ignored 
that part of your request. The code it provides works in a card, a stack, a 
graphic image, an svg widget, in fact, in anything that accepts mouseUp input. 
And yet, it didn’t worry itself about that. Very interesting. It’s kind of 
like, “this works in a button, so good enough”. That is not the kind of 
“nuance” I am used to expecting from an AI. So, progress has clearly been made 
in interpreting what humans say (or ask) in a very loose way. 

The 2nd, and perhaps more astounding example, is when you ask it to "rewrite 
the code to be a function that takes an argument "n", repeats with 1 to n, and 
returns the results”. 

The really clever bit here is not that it rewrites the previous code into a 
function, but that it removes the bit about "put fizzbin into field “display””. 
How does it know that putting the result in a field is NOT the typical 
responsibility of a function? How does it know that that step needs to occur 
AFTER the function call? That is so high-level that I am sure lots of beginning 
programmers have made that mistake. It even comments on it in the 
documentation. So there is some very deep semantic understanding of programming 
code that is going on here. Because,  you did not ask it to do that, but the 
fact that it did it anyway indicates it clearly understands what the purpose of 
a function is. 

Quite weird for me to comprehend how it knows that. Unless of course, you were 
somehow more specific than your examples illustrate. 

Mark

> On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode 
>  wrote:
> 
> So just now I put this in to chatgpt:
> 
> "fizz bin" is a program that counts from 1 to 100, replacing every number
> divisible by 5 with "fizz" and every number divisible by 7 with "bin"; and
> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> programming language livecode to execute "fizz bin"
> 
> This is what ChatGPT responded with:
> 
> repeat with x = 1 to 100
>if x mod 5 = 0 and x mod 7 = 0 then
>put "fizz bin"
>else if x mod 5 = 0 then
>put "fizz"
>else if x mod 7 = 0 then
>put "bin"
>else
>put x
>end if
> end repeat
> 
> That works as-is, but obviously the results fly by in the message box. So I
> put a field and a button on a card, and edited to this, and it works
> perfectly:
> 
> on mouseUp
>   repeat with x = 1 to 100
>  if x mod 5 = 0 and x mod 7 = 0 then
> put "fizz bin" & cr after fld 1
>  else if x mod 5 = 0 then
> put "fizz" & cr after fld 1
>  else if x mod 7 = 0 then
> put "bin" & cr after fld 1
>  else
> put x & cr after fld 1
>  end if
>   end repeat
> end mouseUp
> 
> Then as I was typing this email I realized I was potentially selling
> chatgpt short, so I put this in:
> 
> Modify that code so it works as an "on mouseUp" handler in a button, and
> concatenates the results on individual lines, and puts the final result
> into the field "display"
> 
> And it output this:
> 
> 
> on mouseUp
>   put "" into fizzbin
>   repeat with x = 1 to 100
>  if x mod 5 = 0 and x mod 7 = 0 then
>  put "fizz bin" & return after fizzbin
>  else if x mod 5 = 0 then
>  put "fizz" & return after fizzbin
>  else if x mod 7 = 0 then
>  put "bin" & return after fizzbin
>  else
>  put x & return after fizzbin
>  end if
>   end repeat
>   put fizzbin into field "display"
> end mouseUp
> 
> And it followed that with instructions:
> 
> This code uses the "on mouseUp" handler to execute the script when the
> button is clicked. The results are concatenated and separated by a new line
> using the "& return" operator after each output. The final result is placed
> into a field with the name "display" using the "put" command.
> 
> I'll save you the trouble: that code executes perfectly right out of the
> box, with no errors, and of course even though I didn't ask it to, it uses
> a variable to store the intermediate values, saving the field update for
> the end.
> 
> If I were writing it, I'd probably do something like
> 
> on mouseUp
>   put "" into fizzbin
>   repeat with x = 1 to 100
>  if x mod 5 = 0 then put "fizz " after fizzbin
>  if x mod 7 = 0 then put "bin " after fizzbin
>  if char -1 of fizzbin is space
>  then put return into char -1 of fizzbin
>  else put x & return after fizzbin
>   end repeat
>   put fizzbin into field "display"
> end mouseUp
> 
> That might be a bit faster, but probably not as clear.
> 
> Finally, I gave it the input:
> 
> Now rewrite the code to be a function 

Re: Training the AI to write better LiveCode

[Mark Smith via use-livecode]

We live in interesting times. 

> On 14 Jan 2023, at 3:22 am, Geoff Canyon via use-livecode 
>  wrote:
> 
> So just now I put this in to chatgpt:
> 
> "fizz bin" is a program that counts from 1 to 100, replacing every number
> divisible by 5 with "fizz" and every number divisible by 7 with "bin"; and
> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> programming language livecode to execute "fizz bin"
> 
> This is what ChatGPT responded with:
> 
> repeat with x = 1 to 100
>if x mod 5 = 0 and x mod 7 = 0 then
>put "fizz bin"
>else if x mod 5 = 0 then
>put "fizz"
>else if x mod 7 = 0 then
>put "bin"
>else
>put x
>end if
> end repeat
> 
> That works as-is, but obviously the results fly by in the message box. So I
> put a field and a button on a card, and edited to this, and it works
> perfectly:
> 
> on mouseUp
>   repeat with x = 1 to 100
>  if x mod 5 = 0 and x mod 7 = 0 then
> put "fizz bin" & cr after fld 1
>  else if x mod 5 = 0 then
> put "fizz" & cr after fld 1
>  else if x mod 7 = 0 then
> put "bin" & cr after fld 1
>  else
> put x & cr after fld 1
>  end if
>   end repeat
> end mouseUp
> 
> Then as I was typing this email I realized I was potentially selling
> chatgpt short, so I put this in:
> 
> Modify that code so it works as an "on mouseUp" handler in a button, and
> concatenates the results on individual lines, and puts the final result
> into the field "display"
> 
> And it output this:
> 
> 
> on mouseUp
>   put "" into fizzbin
>   repeat with x = 1 to 100
>  if x mod 5 = 0 and x mod 7 = 0 then
>  put "fizz bin" & return after fizzbin
>  else if x mod 5 = 0 then
>  put "fizz" & return after fizzbin
>  else if x mod 7 = 0 then
>  put "bin" & return after fizzbin
>  else
>  put x & return after fizzbin
>  end if
>   end repeat
>   put fizzbin into field "display"
> end mouseUp
> 
> And it followed that with instructions:
> 
> This code uses the "on mouseUp" handler to execute the script when the
> button is clicked. The results are concatenated and separated by a new line
> using the "& return" operator after each output. The final result is placed
> into a field with the name "display" using the "put" command.
> 
> I'll save you the trouble: that code executes perfectly right out of the
> box, with no errors, and of course even though I didn't ask it to, it uses
> a variable to store the intermediate values, saving the field update for
> the end.
> 
> If I were writing it, I'd probably do something like
> 
> on mouseUp
>   put "" into fizzbin
>   repeat with x = 1 to 100
>  if x mod 5 = 0 then put "fizz " after fizzbin
>  if x mod 7 = 0 then put "bin " after fizzbin
>  if char -1 of fizzbin is space
>  then put return into char -1 of fizzbin
>  else put x & return after fizzbin
>   end repeat
>   put fizzbin into field "display"
> end mouseUp
> 
> That might be a bit faster, but probably not as clear.
> 
> Finally, I gave it the input:
> 
> Now rewrite the code to be a function named "fizzbin" that takes an
> argument "n", repeats with 1 to n, and returns the results.
> 
> function fizzbin n
>   put "" into fizzbin
>   repeat with x = 1 to n
>  if x mod 5 = 0 and x mod 7 = 0 then
>  put "fizz bin" & return after fizzbin
>  else if x mod 5 = 0 then
>  put "fizz" & return after fizzbin
>  else if x mod 7 = 0 then
>  put "bin" & return after fizzbin
>  else
>  put x & return after fizzbin
>  end if
>   end repeat
>   return fizzbin
> end fizzbin
> 
> This works perfectly, and chat gpt also documents its answer:
> 
> This code defines a function named "fizzbin" which takes an argument "n"
> that determines the range of the repeat loop. The function builds the
> result string in a variable "fizzbin" and return the final result. The
> calling script can then use the returned value for further processing or
> for displaying in a field.
> 
> Grammatical issue notwithstanding, that's good documentation.
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription 
> preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[harrison--- via use-livecode]

The rules for fizzbin were intended to be complex, so that Kirk could lull his 
audience into lowering their defenses long enough to be overwhelmed.

The game can be played with a standard Earth deck of cards, despite the 
slightly differing deck on Beta Antares IV 
.
Each player gets six cards, except for the player on the dealer's right, who 
gets seven.
The second card is turned up, except on Tuesdays.
Two jacks are a "half-fizzbin".
If you have a half-fizzbin:
a third jack is a "shralk" and results in disqualification;
one wants a king and a deuce, except at night, when one wants a queen and a 
four;
if a king had been dealt, the player would get another card, except when it is 
dark, in which case he'd have to give it back.
The top hand is a "royal fizzbin", but the odds of getting one are said to be 
"astronomical".
Give those rules to ChatGPT. Command it to create the game using LiveCode, and 
see what happens!  ;-)

Rick

> On Jan 14, 2023, at 12:43 PM, Mark Wieder via use-livecode 
>  wrote:
> 
> Keep in mind,though, that there's lots of example fizzbuzz code for the 
> algorithms to draw from.

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[harrison--- via use-livecode]

I charge all of my clients for training.

I charge a lot more to train Artificial Intelligence Chatbots!

Unfortunately, chatbots don’t yet know how to pay me,
and until someone is willing to pay me to train it, that isn’t
going to happen.

Until they are willing and able to pay me for training, I have 
better things to do with my time.

Just my 2 cents. ;-)

Rick



___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Wieder via use-livecode]

On 1/14/23 09:30, Geoff Canyon via use-livecode wrote:

On Fri, Jan 13, 2023 at 7:22 PM Geoff Canyon  wrote:



If I were writing it, I'd probably do something like

[my code]

That might be a bit faster, but probably not as clear.



I just checked, and my code is about 20% *slower* and still less clear.
Good grief, I'm about to be replaced by a machine.


Keep in mind,though, that there's lots of example fizzbuzz code for the 
algorithms to draw from.


--
 Mark Wieder
 ahsoftw...@gmail.com


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

On Fri, Jan 13, 2023 at 7:22 PM Geoff Canyon  wrote:

>
> If I were writing it, I'd probably do something like
>
> [my code]
>
> That might be a bit faster, but probably not as clear.
>

I just checked, and my code is about 20% *slower* and still less clear.
Good grief, I'm about to be replaced by a machine.
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Wieder via use-livecode]

On 1/14/23 03:35, Beda via use-livecode wrote:
It seems to be more inspiring for me than talking to a rubber duck 
sitting on my desk, at least for simple problems.


For harder problems there will be, hopefully for ever, the user forum 
and this newslist.


I am *not* giving up my rubber duck.

--
 Mark Wieder
 ahsoftw...@gmail.com


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Beda via use-livecode]

It seems to be more inspiring for me than talking to a rubber duck 
sitting on my desk, at least for simple problems.


For harder problems there will be, hopefully for ever, the user forum 
and this newslist.


Beda

Am 14.01.23 um 11:17 schrieb Kaveh via use-livecode:

Quite incredible!

It might be that LiveCode's simple syntax is best suited to ChatGPT.

On Sat, 14 Jan 2023 at 03:24, Geoff Canyon via use-livecode <
use-livecode@lists.runrev.com> wrote:


So just now I put this in to chatgpt:

"fizz bin" is a program that counts from 1 to 100, replacing every number
divisible by 5 with "fizz" and every number divisible by 7 with "bin"; and
numbers divisible by both 5 and 7 with "fizz bin". Write code in the
programming language livecode to execute "fizz bin"

This is what ChatGPT responded with:

repeat with x = 1 to 100
 if x mod 5 = 0 and x mod 7 = 0 then
 put "fizz bin"
 else if x mod 5 = 0 then
 put "fizz"
 else if x mod 7 = 0 then
 put "bin"
 else
 put x
 end if
end repeat

That works as-is, but obviously the results fly by in the message box. So I
put a field and a button on a card, and edited to this, and it works
perfectly:

on mouseUp
repeat with x = 1 to 100
   if x mod 5 = 0 and x mod 7 = 0 then
  put "fizz bin" & cr after fld 1
   else if x mod 5 = 0 then
  put "fizz" & cr after fld 1
   else if x mod 7 = 0 then
  put "bin" & cr after fld 1
   else
  put x & cr after fld 1
   end if
end repeat
end mouseUp

Then as I was typing this email I realized I was potentially selling
chatgpt short, so I put this in:

Modify that code so it works as an "on mouseUp" handler in a button, and
concatenates the results on individual lines, and puts the final result
into the field "display"

And it output this:


on mouseUp
put "" into fizzbin
repeat with x = 1 to 100
   if x mod 5 = 0 and x mod 7 = 0 then
   put "fizz bin" & return after fizzbin
   else if x mod 5 = 0 then
   put "fizz" & return after fizzbin
   else if x mod 7 = 0 then
   put "bin" & return after fizzbin
   else
   put x & return after fizzbin
   end if
end repeat
put fizzbin into field "display"
end mouseUp

And it followed that with instructions:

This code uses the "on mouseUp" handler to execute the script when the
button is clicked. The results are concatenated and separated by a new line
using the "& return" operator after each output. The final result is placed
into a field with the name "display" using the "put" command.

I'll save you the trouble: that code executes perfectly right out of the
box, with no errors, and of course even though I didn't ask it to, it uses
a variable to store the intermediate values, saving the field update for
the end.

If I were writing it, I'd probably do something like

on mouseUp
put "" into fizzbin
repeat with x = 1 to 100
   if x mod 5 = 0 then put "fizz " after fizzbin
   if x mod 7 = 0 then put "bin " after fizzbin
   if char -1 of fizzbin is space
   then put return into char -1 of fizzbin
   else put x & return after fizzbin
end repeat
put fizzbin into field "display"
end mouseUp

That might be a bit faster, but probably not as clear.

Finally, I gave it the input:

Now rewrite the code to be a function named "fizzbin" that takes an
argument "n", repeats with 1 to n, and returns the results.

function fizzbin n
put "" into fizzbin
repeat with x = 1 to n
   if x mod 5 = 0 and x mod 7 = 0 then
   put "fizz bin" & return after fizzbin
   else if x mod 5 = 0 then
   put "fizz" & return after fizzbin
   else if x mod 7 = 0 then
   put "bin" & return after fizzbin
   else
   put x & return after fizzbin
   end if
end repeat
return fizzbin
end fizzbin

This works perfectly, and chat gpt also documents its answer:

This code defines a function named "fizzbin" which takes an argument "n"
that determines the range of the repeat loop. The function builds the
result string in a variable "fizzbin" and return the final result. The
calling script can then use the returned value for further processing or
for displaying in a field.

Grammatical issue notwithstanding, that's good documentation.
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your
subscription preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode





___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Kaveh via use-livecode]

Quite incredible!

It might be that LiveCode's simple syntax is best suited to ChatGPT.

On Sat, 14 Jan 2023 at 03:24, Geoff Canyon via use-livecode <
use-livecode@lists.runrev.com> wrote:

> So just now I put this in to chatgpt:
>
> "fizz bin" is a program that counts from 1 to 100, replacing every number
> divisible by 5 with "fizz" and every number divisible by 7 with "bin"; and
> numbers divisible by both 5 and 7 with "fizz bin". Write code in the
> programming language livecode to execute "fizz bin"
>
> This is what ChatGPT responded with:
>
> repeat with x = 1 to 100
> if x mod 5 = 0 and x mod 7 = 0 then
> put "fizz bin"
> else if x mod 5 = 0 then
> put "fizz"
> else if x mod 7 = 0 then
> put "bin"
> else
> put x
> end if
> end repeat
>
> That works as-is, but obviously the results fly by in the message box. So I
> put a field and a button on a card, and edited to this, and it works
> perfectly:
>
> on mouseUp
>repeat with x = 1 to 100
>   if x mod 5 = 0 and x mod 7 = 0 then
>  put "fizz bin" & cr after fld 1
>   else if x mod 5 = 0 then
>  put "fizz" & cr after fld 1
>   else if x mod 7 = 0 then
>  put "bin" & cr after fld 1
>   else
>  put x & cr after fld 1
>   end if
>end repeat
> end mouseUp
>
> Then as I was typing this email I realized I was potentially selling
> chatgpt short, so I put this in:
>
> Modify that code so it works as an "on mouseUp" handler in a button, and
> concatenates the results on individual lines, and puts the final result
> into the field "display"
>
> And it output this:
>
>
> on mouseUp
>put "" into fizzbin
>repeat with x = 1 to 100
>   if x mod 5 = 0 and x mod 7 = 0 then
>   put "fizz bin" & return after fizzbin
>   else if x mod 5 = 0 then
>   put "fizz" & return after fizzbin
>   else if x mod 7 = 0 then
>   put "bin" & return after fizzbin
>   else
>   put x & return after fizzbin
>   end if
>end repeat
>put fizzbin into field "display"
> end mouseUp
>
> And it followed that with instructions:
>
> This code uses the "on mouseUp" handler to execute the script when the
> button is clicked. The results are concatenated and separated by a new line
> using the "& return" operator after each output. The final result is placed
> into a field with the name "display" using the "put" command.
>
> I'll save you the trouble: that code executes perfectly right out of the
> box, with no errors, and of course even though I didn't ask it to, it uses
> a variable to store the intermediate values, saving the field update for
> the end.
>
> If I were writing it, I'd probably do something like
>
> on mouseUp
>put "" into fizzbin
>repeat with x = 1 to 100
>   if x mod 5 = 0 then put "fizz " after fizzbin
>   if x mod 7 = 0 then put "bin " after fizzbin
>   if char -1 of fizzbin is space
>   then put return into char -1 of fizzbin
>   else put x & return after fizzbin
>end repeat
>put fizzbin into field "display"
> end mouseUp
>
> That might be a bit faster, but probably not as clear.
>
> Finally, I gave it the input:
>
> Now rewrite the code to be a function named "fizzbin" that takes an
> argument "n", repeats with 1 to n, and returns the results.
>
> function fizzbin n
>put "" into fizzbin
>repeat with x = 1 to n
>   if x mod 5 = 0 and x mod 7 = 0 then
>   put "fizz bin" & return after fizzbin
>   else if x mod 5 = 0 then
>   put "fizz" & return after fizzbin
>   else if x mod 7 = 0 then
>   put "bin" & return after fizzbin
>   else
>   put x & return after fizzbin
>   end if
>end repeat
>return fizzbin
> end fizzbin
>
> This works perfectly, and chat gpt also documents its answer:
>
> This code defines a function named "fizzbin" which takes an argument "n"
> that determines the range of the repeat loop. The function builds the
> result string in a variable "fizzbin" and return the final result. The
> calling script can then use the returned value for further processing or
> for displaying in a field.
>
> Grammatical issue notwithstanding, that's good documentation.
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
Kaveh Bazargan PhD
Director
River Valley Technologies  ● Twitter
 ● LinkedIn
 ● ORCID
 ● @kaveh1000@mastodon.social

*Accelerating the Communication of Research*
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to 

Re: Training the AI to write better LiveCode

[Geoff Canyon via use-livecode]

So just now I put this in to chatgpt:

"fizz bin" is a program that counts from 1 to 100, replacing every number
divisible by 5 with "fizz" and every number divisible by 7 with "bin"; and
numbers divisible by both 5 and 7 with "fizz bin". Write code in the
programming language livecode to execute "fizz bin"

This is what ChatGPT responded with:

repeat with x = 1 to 100
if x mod 5 = 0 and x mod 7 = 0 then
put "fizz bin"
else if x mod 5 = 0 then
put "fizz"
else if x mod 7 = 0 then
put "bin"
else
put x
end if
end repeat

That works as-is, but obviously the results fly by in the message box. So I
put a field and a button on a card, and edited to this, and it works
perfectly:

on mouseUp
   repeat with x = 1 to 100
  if x mod 5 = 0 and x mod 7 = 0 then
 put "fizz bin" & cr after fld 1
  else if x mod 5 = 0 then
 put "fizz" & cr after fld 1
  else if x mod 7 = 0 then
 put "bin" & cr after fld 1
  else
 put x & cr after fld 1
  end if
   end repeat
end mouseUp

Then as I was typing this email I realized I was potentially selling
chatgpt short, so I put this in:

Modify that code so it works as an "on mouseUp" handler in a button, and
concatenates the results on individual lines, and puts the final result
into the field "display"

And it output this:


on mouseUp
   put "" into fizzbin
   repeat with x = 1 to 100
  if x mod 5 = 0 and x mod 7 = 0 then
  put "fizz bin" & return after fizzbin
  else if x mod 5 = 0 then
  put "fizz" & return after fizzbin
  else if x mod 7 = 0 then
  put "bin" & return after fizzbin
  else
  put x & return after fizzbin
  end if
   end repeat
   put fizzbin into field "display"
end mouseUp

And it followed that with instructions:

This code uses the "on mouseUp" handler to execute the script when the
button is clicked. The results are concatenated and separated by a new line
using the "& return" operator after each output. The final result is placed
into a field with the name "display" using the "put" command.

I'll save you the trouble: that code executes perfectly right out of the
box, with no errors, and of course even though I didn't ask it to, it uses
a variable to store the intermediate values, saving the field update for
the end.

If I were writing it, I'd probably do something like

on mouseUp
   put "" into fizzbin
   repeat with x = 1 to 100
  if x mod 5 = 0 then put "fizz " after fizzbin
  if x mod 7 = 0 then put "bin " after fizzbin
  if char -1 of fizzbin is space
  then put return into char -1 of fizzbin
  else put x & return after fizzbin
   end repeat
   put fizzbin into field "display"
end mouseUp

That might be a bit faster, but probably not as clear.

Finally, I gave it the input:

Now rewrite the code to be a function named "fizzbin" that takes an
argument "n", repeats with 1 to n, and returns the results.

function fizzbin n
   put "" into fizzbin
   repeat with x = 1 to n
  if x mod 5 = 0 and x mod 7 = 0 then
  put "fizz bin" & return after fizzbin
  else if x mod 5 = 0 then
  put "fizz" & return after fizzbin
  else if x mod 7 = 0 then
  put "bin" & return after fizzbin
  else
  put x & return after fizzbin
  end if
   end repeat
   return fizzbin
end fizzbin

This works perfectly, and chat gpt also documents its answer:

This code defines a function named "fizzbin" which takes an argument "n"
that determines the range of the repeat loop. The function builds the
result string in a variable "fizzbin" and return the final result. The
calling script can then use the returned value for further processing or
for displaying in a field.

Grammatical issue notwithstanding, that's good documentation.
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Jim Lambert via use-livecode]

> Kevin wrote:
> What is important is ensuring that LiveCode, which is supported already, is 
> not left behind and that you can generate LiveCode code at the same standard 
> as other languages. 

Exactly!

> Rick wrote:

> Just because something may be technologically
> possible, doesn't mean we should necessarily do that thing.

Ah! But humans always do. ;)

Jim Lambert


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mark Wieder via use-livecode]

On 1/13/23 07:30, Mike Kerner via use-livecode wrote:

"AI" has always been a misleading term, for like 40 years.
gawd, i'm old.
There is pattern matching, tree traversals, bayesian probability, etc.
it's definitely not anything like a wiki.
it might just be helpful at slogging out a lot of the code that we have to
write or include by hand, especially if you are trying to tackle something
new, especially for code that isn't neatly organized and instantly
available.
but, it's also not amazing and mature, either. we can be both surprised at
the program a ten-year-old writes, and recognize that it's...not amazing.


I think maybe it's time to stop using the ridiculous "AI" term and refer 
to this more properly as Machine Learning. From my experience using 
TensorFlow the quality of your results depends mainly on the corpus of 
data used in training. Not just the quantity of data used to train the 
system but also the quality of the data used. It doesn't matter whether 
you're training the system for facial recognition or grammar checking or 
code generation, the better quality data you can throw at the system the 
better your results will be. Any ML system will have a bias towards the 
data it was trained on, so it stands to reason that having better code 
to draw on will allow it to make better judgements about code creation.


--
 Mark Wieder
 ahsoftw...@gmail.com


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

"AI" has always been a misleading term, for like 40 years.
gawd, i'm old.
There is pattern matching, tree traversals, bayesian probability, etc.
it's definitely not anything like a wiki.
it might just be helpful at slogging out a lot of the code that we have to
write or include by hand, especially if you are trying to tackle something
new, especially for code that isn't neatly organized and instantly
available.
but, it's also not amazing and mature, either. we can be both surprised at
the program a ten-year-old writes, and recognize that it's...not amazing.


On Fri, Jan 13, 2023 at 5:50 AM Kevin Miller via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Its far more than a search engine. It can edit copy, write articles and
> analyse multiple inputs for similarities and differences and a lot more.
> Like all such new technologies this is going to be here to stay. So its not
> a case of trying not to train it in order to preserve programmer jobs.
> There will be plenty left for a human to do in building an app, we are just
> going to get more productive as this gets better. What is important is
> ensuring that LiveCode, which is supported already, is not left behind and
> that you can generate LiveCode code at the same standard as other
> languages. At present its coding abilities in any language are interesting
> but a bit ropy, lets ensure that as it gets better as it inevitably will,
> its LiveCode skills improve too.
>
> Kind regards,
>
> Kevin
>
> Kevin Miller ~ ke...@livecode.com ~ http://www.livecode.com/
> LiveCode: Build Amazing Things
>
>
>
> On 12/01/2023, 17:45, "use-livecode on behalf of harrison--- via
> use-livecode"  use-livecode-boun...@lists.runrev.com> on behalf of
> use-livecode@lists.runrev.com >
> wrote:
>
>
> ChatGPT has been called a search engine without the search engine.
> I would tend to agree with that. It seems to find bits and pieces of code
> or articles that others have written before and tries to put them together.
> It uses a text typing output mode to give it the appearance of an
> intelligence/person typing at the keyboard, but that is part of the
> illusion.
>
>
> ChatGPT claims that it isn’t connected to the internet, but it is, or it
> wouldn’t
> be able to respond to your questions over the internet. The database
> that it accesses for answers is however compartmentalized, as far as
> we have been told.
>
>
> If by some miracle we were able to train it to be super intelligent at
> writing computer code, do we as programmers really want to shoot
> ourselves in the foot by giving it the tools to replace our talent?
> That’s food for thought. Just because something may be technologically
> possible, doesn’t mean we should necessarily do that thing.
>
>
> Just my two cents.
>
>
> Enjoy your day!
>
>
> Rick
>
>
> > On Jan 12, 2023, at 9:02 AM, Mike Kerner via use-livecode <
> use-livecode@lists.runrev.com >
> wrote:
> >
> > yesterday was not a good session for me. all responses were irrelevant,
> and
> > sometimes seemed random.
> > we'll see what we find, the next time.
>
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com 
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode <
> http://lists.runrev.com/mailman/listinfo/use-livecode>
>
>
>
>
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Kevin Miller via use-livecode]

Its far more than a search engine. It can edit copy, write articles and analyse 
multiple inputs for similarities and differences and a lot more. Like all such 
new technologies this is going to be here to stay. So its not a case of trying 
not to train it in order to preserve programmer jobs. There will be plenty left 
for a human to do in building an app, we are just going to get more productive 
as this gets better. What is important is ensuring that LiveCode, which is 
supported already, is not left behind and that you can generate LiveCode code 
at the same standard as other languages. At present its coding abilities in any 
language are interesting but a bit ropy, lets ensure that as it gets better as 
it inevitably will, its LiveCode skills improve too.

Kind regards, 

Kevin 

Kevin Miller ~ ke...@livecode.com ~ http://www.livecode.com/ 
LiveCode: Build Amazing Things 



On 12/01/2023, 17:45, "use-livecode on behalf of harrison--- via use-livecode" 
mailto:use-livecode-boun...@lists.runrev.com> on behalf of 
use-livecode@lists.runrev.com > wrote:


ChatGPT has been called a search engine without the search engine.
I would tend to agree with that. It seems to find bits and pieces of code
or articles that others have written before and tries to put them together.
It uses a text typing output mode to give it the appearance of an
intelligence/person typing at the keyboard, but that is part of the illusion.


ChatGPT claims that it isn’t connected to the internet, but it is, or it 
wouldn’t
be able to respond to your questions over the internet. The database
that it accesses for answers is however compartmentalized, as far as
we have been told.


If by some miracle we were able to train it to be super intelligent at
writing computer code, do we as programmers really want to shoot
ourselves in the foot by giving it the tools to replace our talent?
That’s food for thought. Just because something may be technologically
possible, doesn’t mean we should necessarily do that thing.


Just my two cents.


Enjoy your day!


Rick


> On Jan 12, 2023, at 9:02 AM, Mike Kerner via use-livecode 
> mailto:use-livecode@lists.runrev.com>> wrote:
> 
> yesterday was not a good session for me. all responses were irrelevant, and
> sometimes seemed random.
> we'll see what we find, the next time.


___
use-livecode mailing list
use-livecode@lists.runrev.com 
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode 






___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Bob Sneidar via use-livecode]

Sooo... it's ARTIFICIAL intelligence, meaning there's a programmer behind it 
telling it how to respond, or else how not to. Obviously I'm not a huge 
believer in AI. It seems more like WIKI to me. 

Bob S


> On Jan 12, 2023, at 01:16 , Heather Laine via use-livecode 
>  wrote:
> 
> Yes, it only retains this "learning" for the session with you. To ensure 
> wider learning for the future, you need to give feedback via the "Contact Us" 
> page. I think its trainers can then apply the learning in some unknown 
> fashion.
> 
> Best Regards,
> 
> Heather
> 
> Heather Laine
> Customer Services Manager
> LiveCode Ltd
> www.livecode.com
> 
> 
> 
>> On 12 Jan 2023, at 07:11, Beda via use-livecode 
>>  wrote:
>> 
>> Hell again, I just asked Chat GPT and it answered always very politly:
>> 
>> Me: Thanks that's ok
>> 
>> Chat GPT: You're welcome! I'm glad I could help. If you have any other 
>> questions about LiveCode or any other programming language, feel free to ask.
>> 
>> Me: Did you learn something about Livecode by our conversation?
>> 
>> Chat GPT: As a language model AI, I do not have the ability to learn or 
>> experience things. However, I do have access to a vast amount of information 
>> and knowledge, and my answers are generated based on that information.
>> During our conversation, I provided you with specific examples of how to use 
>> the "answer" and "put" commands to display a message in the message box in a 
>> LiveCode application, and also I've clarified my previous mistake about the 
>> Livecode message box command line interface.
>> 
>> (End of Chat)
>> 
>> Well it even does aknowledge mistakes but doesn't realy say if it will avoid 
>> these in future!
>> 
>> Best
>> 
>> Beda
>> 
>> 
>> Am 12.01.23 um 07:54 schrieb Beda via use-livecode:
>>> Dear Heather,
>>> 
>>> thank you for your hint, it is real fun!
>>> 
>>> Here my first chat with AI GPT. It seems, that you have to insist sometimes 
>>> for your demands and then there will be the right answer, but does it learn 
>>> by this way?
>>> 
>>> Me: Write some code in livecode
>>> 
>>> Chat GPT: I'm sorry, I am a text-based AI and do not have the ability to 
>>> write code in livecode. However, I can write code in various programming 
>>> languages such as Python, Java, C++, JavaScript and more. If you give me a 
>>> specific task, I can write the code for you.
>>> 
>>> 
>>> Me: Write a "Hello World" statement in livecode
>>> 
>>> Chat GPT: I apologize for the confusion, LiveCode is a visual programming 
>>> language and development environment that allows you to create apps, games, 
>>> and more using a drag-and-drop interface, it doesn't have a command line 
>>> interface.
>>> However, I can give you an example of how to display "Hello World" in a 
>>> LiveCode application.
>>> 
>>> Copy code
>>> on mouseUp answer "Hello World!" end mouseUp
>>> You would need to have a button in the application and then when you press 
>>> the button it will display the "Hello World!" message.
>>> 
>>> 
>>> Me: But yes, livecode has a command line interface called message box
>>> 
>>> Chat GPT: I apologize for the confusion, you are correct that LiveCode does 
>>> have a command line interface called the message box. You can use the 
>>> "answer" command to display a message in the message box.
>>> 
>>> Copy code
>>> answer "Hello World!"
>>> This command will display the message "Hello World!" in the message box.
>>> 
>>> Copy code
>>> put "Hello World!"
>>> This command will output the message "Hello World!" in the message box.
>>> You can also use the "put" command to display a message in the message box.
>>> 
>>> Copy code
>>> put "Hello World!"
>>> 
>>> This command will output the message "Hello World!" in the message box.
>>> 
>>> (End of the answer)
>>> 
>>> Best regards
>>> 
>>> Beda
>>> 
>>> 
>>> Am 11.01.23 um 15:58 schrieb Heather Laine via use-livecode:
 Dear List Members and LiveCode Gurus,
 
 We're looking for your help. I should start with a warning though: this 
 chatGPT is addictive, you might lose a few hours of your life. But you 
 will have fun doing it :)
 
 Our idea is that the chatGPT to be found here:
 
 https://chat.openai.com
 
 is a truly great way for beginners and even more advanced LiveCoders to 
 get into LiveCode, learn scripting and get common scripts pre-written. It 
 could be an immensely powerful tool in bringing LiveCode to a wider 
 audience. (In case you haven't already wasted hours of your life playing 
 with it, you can ask it to write a LiveCode script for any specific task, 
 and it will take its best shot, with very passable results).
 
 Unfortunately, at the moment it has a tendency to make schoolboy (or 
 should I say schoolbot?) errors. It gets the script mostly right but there 
 will be issues that a beginner would not know how to correct. We would 
 love it if the community could help it learn to write beautiful, correct, 

Re: Training the AI to write better LiveCode

[harrison--- via use-livecode]

ChatGPT has been called a search engine without the search engine.
I would tend to agree with that.  It seems to find bits and pieces of code
or articles that others have written before and tries to put them together.
It uses a text typing output mode to give it the appearance of an
intelligence/person typing at the keyboard, but that is part of the illusion.

ChatGPT claims that it isn’t connected to the internet, but it is, or it 
wouldn’t
be able to respond to your questions over the internet.  The database
that it accesses for answers is however compartmentalized, as far as
we have been told.

If by some miracle we were able to train it to be super intelligent at
writing computer code, do we as programmers really want to shoot
ourselves in the foot by giving it the tools to replace our talent?
That’s food for thought.  Just because something may be technologically
possible, doesn’t mean we should necessarily do that thing.

Just my two cents.

Enjoy your day!

Rick

> On Jan 12, 2023, at 9:02 AM, Mike Kerner via use-livecode 
>  wrote:
> 
> yesterday was not a good session for me. all responses were irrelevant, and
> sometimes seemed random.
> we'll see what we find, the next time.

___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

yesterday was not a good session for me. all responses were irrelevant, and
sometimes seemed random.
we'll see what we find, the next time.

On Thu, Jan 12, 2023 at 4:17 AM Heather Laine via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Yes, it only retains this "learning" for the session with you. To ensure
> wider learning for the future, you need to give feedback via the "Contact
> Us" page. I think its trainers can then apply the learning in some unknown
> fashion.
>
> Best Regards,
>
> Heather
>
> Heather Laine
> Customer Services Manager
> LiveCode Ltd
> www.livecode.com
>
>
>
> > On 12 Jan 2023, at 07:11, Beda via use-livecode <
> use-livecode@lists.runrev.com> wrote:
> >
> > Hell again, I just asked Chat GPT and it answered always very politly:
> >
> > Me: Thanks that's ok
> >
> > Chat GPT: You're welcome! I'm glad I could help. If you have any other
> questions about LiveCode or any other programming language, feel free to
> ask.
> >
> > Me: Did you learn something about Livecode by our conversation?
> >
> > Chat GPT: As a language model AI, I do not have the ability to learn or
> experience things. However, I do have access to a vast amount of
> information and knowledge, and my answers are generated based on that
> information.
> > During our conversation, I provided you with specific examples of how to
> use the "answer" and "put" commands to display a message in the message box
> in a LiveCode application, and also I've clarified my previous mistake
> about the Livecode message box command line interface.
> >
> > (End of Chat)
> >
> > Well it even does aknowledge mistakes but doesn't realy say if it will
> avoid these in future!
> >
> > Best
> >
> > Beda
> >
> >
> > Am 12.01.23 um 07:54 schrieb Beda via use-livecode:
> >> Dear Heather,
> >>
> >> thank you for your hint, it is real fun!
> >>
> >> Here my first chat with AI GPT. It seems, that you have to insist
> sometimes for your demands and then there will be the right answer, but
> does it learn by this way?
> >>
> >> Me: Write some code in livecode
> >>
> >> Chat GPT: I'm sorry, I am a text-based AI and do not have the ability
> to write code in livecode. However, I can write code in various programming
> languages such as Python, Java, C++, JavaScript and more. If you give me a
> specific task, I can write the code for you.
> >>
> >>
> >> Me: Write a "Hello World" statement in livecode
> >>
> >> Chat GPT: I apologize for the confusion, LiveCode is a visual
> programming language and development environment that allows you to create
> apps, games, and more using a drag-and-drop interface, it doesn't have a
> command line interface.
> >> However, I can give you an example of how to display "Hello World" in a
> LiveCode application.
> >>
> >> Copy code
> >> on mouseUp answer "Hello World!" end mouseUp
> >> You would need to have a button in the application and then when you
> press the button it will display the "Hello World!" message.
> >>
> >>
> >> Me: But yes, livecode has a command line interface called message box
> >>
> >> Chat GPT: I apologize for the confusion, you are correct that LiveCode
> does have a command line interface called the message box. You can use the
> "answer" command to display a message in the message box.
> >>
> >> Copy code
> >> answer "Hello World!"
> >> This command will display the message "Hello World!" in the message box.
> >>
> >> Copy code
> >> put "Hello World!"
> >> This command will output the message "Hello World!" in the message box.
> >> You can also use the "put" command to display a message in the message
> box.
> >>
> >> Copy code
> >> put "Hello World!"
> >>
> >> This command will output the message "Hello World!" in the message box.
> >>
> >> (End of the answer)
> >>
> >> Best regards
> >>
> >> Beda
> >>
> >>
> >> Am 11.01.23 um 15:58 schrieb Heather Laine via use-livecode:
> >>> Dear List Members and LiveCode Gurus,
> >>>
> >>> We're looking for your help. I should start with a warning though:
> this chatGPT is addictive, you might lose a few hours of your life. But you
> will have fun doing it :)
> >>>
> >>> Our idea is that the chatGPT to be found here:
> >>>
> >>> https://chat.openai.com
> >>>
> >>> is a truly great way for beginners and even more advanced LiveCoders
> to get into LiveCode, learn scripting and get common scripts pre-written.
> It could be an immensely powerful tool in bringing LiveCode to a wider
> audience. (In case you haven't already wasted hours of your life playing
> with it, you can ask it to write a LiveCode script for any specific task,
> and it will take its best shot, with very passable results).
> >>>
> >>> Unfortunately, at the moment it has a tendency to make schoolboy (or
> should I say schoolbot?) errors. It gets the script mostly right but there
> will be issues that a beginner would not know how to correct. We would love
> it if the community could help it learn to write beautiful, correct,
> LiveCode. To do this, it needs feedback. We asked it 

Re: Training the AI to write better LiveCode

[Heather Laine via use-livecode]

Yes, it only retains this "learning" for the session with you. To ensure wider 
learning for the future, you need to give feedback via the "Contact Us" page. I 
think its trainers can then apply the learning in some unknown fashion.

Best Regards,

Heather

Heather Laine
Customer Services Manager
LiveCode Ltd
www.livecode.com



> On 12 Jan 2023, at 07:11, Beda via use-livecode 
>  wrote:
> 
> Hell again, I just asked Chat GPT and it answered always very politly:
> 
> Me: Thanks that's ok
> 
> Chat GPT: You're welcome! I'm glad I could help. If you have any other 
> questions about LiveCode or any other programming language, feel free to ask.
> 
> Me: Did you learn something about Livecode by our conversation?
> 
> Chat GPT: As a language model AI, I do not have the ability to learn or 
> experience things. However, I do have access to a vast amount of information 
> and knowledge, and my answers are generated based on that information.
> During our conversation, I provided you with specific examples of how to use 
> the "answer" and "put" commands to display a message in the message box in a 
> LiveCode application, and also I've clarified my previous mistake about the 
> Livecode message box command line interface.
> 
> (End of Chat)
> 
> Well it even does aknowledge mistakes but doesn't realy say if it will avoid 
> these in future!
> 
> Best
> 
> Beda
> 
> 
> Am 12.01.23 um 07:54 schrieb Beda via use-livecode:
>> Dear Heather,
>> 
>> thank you for your hint, it is real fun!
>> 
>> Here my first chat with AI GPT. It seems, that you have to insist sometimes 
>> for your demands and then there will be the right answer, but does it learn 
>> by this way?
>> 
>> Me: Write some code in livecode
>> 
>> Chat GPT: I'm sorry, I am a text-based AI and do not have the ability to 
>> write code in livecode. However, I can write code in various programming 
>> languages such as Python, Java, C++, JavaScript and more. If you give me a 
>> specific task, I can write the code for you.
>> 
>> 
>> Me: Write a "Hello World" statement in livecode
>> 
>> Chat GPT: I apologize for the confusion, LiveCode is a visual programming 
>> language and development environment that allows you to create apps, games, 
>> and more using a drag-and-drop interface, it doesn't have a command line 
>> interface.
>> However, I can give you an example of how to display "Hello World" in a 
>> LiveCode application.
>> 
>> Copy code
>> on mouseUp answer "Hello World!" end mouseUp
>> You would need to have a button in the application and then when you press 
>> the button it will display the "Hello World!" message.
>> 
>> 
>> Me: But yes, livecode has a command line interface called message box
>> 
>> Chat GPT: I apologize for the confusion, you are correct that LiveCode does 
>> have a command line interface called the message box. You can use the 
>> "answer" command to display a message in the message box.
>> 
>> Copy code
>> answer "Hello World!"
>> This command will display the message "Hello World!" in the message box.
>> 
>> Copy code
>> put "Hello World!"
>> This command will output the message "Hello World!" in the message box.
>> You can also use the "put" command to display a message in the message box.
>> 
>> Copy code
>> put "Hello World!"
>> 
>> This command will output the message "Hello World!" in the message box.
>> 
>> (End of the answer)
>> 
>> Best regards
>> 
>> Beda
>> 
>> 
>> Am 11.01.23 um 15:58 schrieb Heather Laine via use-livecode:
>>> Dear List Members and LiveCode Gurus,
>>> 
>>> We're looking for your help. I should start with a warning though: this 
>>> chatGPT is addictive, you might lose a few hours of your life. But you will 
>>> have fun doing it :)
>>> 
>>> Our idea is that the chatGPT to be found here:
>>> 
>>> https://chat.openai.com
>>> 
>>> is a truly great way for beginners and even more advanced LiveCoders to get 
>>> into LiveCode, learn scripting and get common scripts pre-written. It could 
>>> be an immensely powerful tool in bringing LiveCode to a wider audience. (In 
>>> case you haven't already wasted hours of your life playing with it, you can 
>>> ask it to write a LiveCode script for any specific task, and it will take 
>>> its best shot, with very passable results).
>>> 
>>> Unfortunately, at the moment it has a tendency to make schoolboy (or should 
>>> I say schoolbot?) errors. It gets the script mostly right but there will be 
>>> issues that a beginner would not know how to correct. We would love it if 
>>> the community could help it learn to write beautiful, correct, LiveCode. To 
>>> do this, it needs feedback. We asked it how to give feedback and this is 
>>> what it suggests:
>>> 
>>> There are a few ways you can provide feedback on ChatGPT:
>>> • You can submit feedback through the OpenAI website by navigating to 
>>> the “Contact Us” page and selecting “Model Feedback” as the topic.
>>> • You can also provide feedback on the OpenAI GPT-3 forum, where users 
>>> can share 

Re: Training the AI to write better LiveCode

[Beda via use-livecode]

Hell again, I just asked Chat GPT and it answered always very politly:

Me: Thanks that's ok

Chat GPT: You're welcome! I'm glad I could help. If you have any other 
questions about LiveCode or any other programming language, feel free to 
ask.


Me: Did you learn something about Livecode by our conversation?

Chat GPT: As a language model AI, I do not have the ability to learn or 
experience things. However, I do have access to a vast amount of 
information and knowledge, and my answers are generated based on that 
information.
During our conversation, I provided you with specific examples of how to 
use the "answer" and "put" commands to display a message in the message 
box in a LiveCode application, and also I've clarified my previous 
mistake about the Livecode message box command line interface.


(End of Chat)

Well it even does aknowledge mistakes but doesn't realy say if it will 
avoid these in future!


Best

Beda


Am 12.01.23 um 07:54 schrieb Beda via use-livecode:

Dear Heather,

thank you for your hint, it is real fun!

Here my first chat with AI GPT. It seems, that you have to insist 
sometimes for your demands and then there will be the right answer, 
but does it learn by this way?


Me: Write some code in livecode

Chat GPT: I'm sorry, I am a text-based AI and do not have the ability 
to write code in livecode. However, I can write code in various 
programming languages such as Python, Java, C++, JavaScript and more. 
If you give me a specific task, I can write the code for you.



Me: Write a "Hello World" statement in livecode

Chat GPT: I apologize for the confusion, LiveCode is a visual 
programming language and development environment that allows you to 
create apps, games, and more using a drag-and-drop interface, it 
doesn't have a command line interface.
However, I can give you an example of how to display "Hello World" in 
a LiveCode application.


Copy code
on mouseUp answer "Hello World!" end mouseUp
You would need to have a button in the application and then when you 
press the button it will display the "Hello World!" message.



Me: But yes, livecode has a command line interface called message box

Chat GPT: I apologize for the confusion, you are correct that LiveCode 
does have a command line interface called the message box. You can use 
the "answer" command to display a message in the message box.


Copy code
answer "Hello World!"
This command will display the message "Hello World!" in the message box.

Copy code
put "Hello World!"
This command will output the message "Hello World!" in the message box.
You can also use the "put" command to display a message in the message 
box.


Copy code
put "Hello World!"

This command will output the message "Hello World!" in the message box.

(End of the answer)

Best regards

Beda


Am 11.01.23 um 15:58 schrieb Heather Laine via use-livecode:

Dear List Members and LiveCode Gurus,

We're looking for your help. I should start with a warning though: 
this chatGPT is addictive, you might lose a few hours of your life. 
But you will have fun doing it :)


Our idea is that the chatGPT to be found here:

https://chat.openai.com

is a truly great way for beginners and even more advanced LiveCoders 
to get into LiveCode, learn scripting and get common scripts 
pre-written. It could be an immensely powerful tool in bringing 
LiveCode to a wider audience. (In case you haven't already wasted 
hours of your life playing with it, you can ask it to write a 
LiveCode script for any specific task, and it will take its best 
shot, with very passable results).


Unfortunately, at the moment it has a tendency to make schoolboy (or 
should I say schoolbot?) errors. It gets the script mostly right but 
there will be issues that a beginner would not know how to correct. 
We would love it if the community could help it learn to write 
beautiful, correct, LiveCode. To do this, it needs feedback. We asked 
it how to give feedback and this is what it suggests:


There are a few ways you can provide feedback on ChatGPT:
• You can submit feedback through the OpenAI website by 
navigating to the “Contact Us” page and selecting “Model Feedback” as 
the topic.
• You can also provide feedback on the OpenAI GPT-3 forum, where 
users can share their experiences, ask questions, and provide 
suggestions for improving the model.
• You could also contact the openai team to share more detailed 
feedback and suggestions
In any case, clear and detailed feedback is helpful in understanding 
the problem and improve the model.


Just telling it what it got wrong we think isn't the right way to go, 
it will only correct it for that particular session. We need to rat 
on it in to its teachers so that it can be trained correctly.


So, if you are up for the challenge and you have a bit of time to 
kill, we'd love you to a) ask chatGPT to write LiveCode scripts and 
b) send any resulting errors with detailed feedback to the Model 
Feedback area.


Thank you!

Best 

Re: Training the AI to write better LiveCode

[Beda via use-livecode]

Dear Heather,

thank you for your hint, it is real fun!

Here my first chat with AI GPT. It seems, that you have to insist 
sometimes for your demands and then there will be the right answer, but 
does it learn by this way?


Me: Write some code in livecode

Chat GPT: I'm sorry, I am a text-based AI and do not have the ability to 
write code in livecode. However, I can write code in various programming 
languages such as Python, Java, C++, JavaScript and more. If you give me 
a specific task, I can write the code for you.



Me: Write a "Hello World" statement in livecode

Chat GPT: I apologize for the confusion, LiveCode is a visual 
programming language and development environment that allows you to 
create apps, games, and more using a drag-and-drop interface, it doesn't 
have a command line interface.
However, I can give you an example of how to display "Hello World" in a 
LiveCode application.


Copy code
on mouseUp answer "Hello World!" end mouseUp
You would need to have a button in the application and then when you 
press the button it will display the "Hello World!" message.



Me: But yes, livecode has a command line interface called message box

Chat GPT: I apologize for the confusion, you are correct that LiveCode 
does have a command line interface called the message box. You can use 
the "answer" command to display a message in the message box.


Copy code
answer "Hello World!"
This command will display the message "Hello World!" in the message box.

Copy code
put "Hello World!"
This command will output the message "Hello World!" in the message box.
You can also use the "put" command to display a message in the message box.

Copy code
put "Hello World!"

This command will output the message "Hello World!" in the message box.

(End of the answer)

Best regards

Beda


Am 11.01.23 um 15:58 schrieb Heather Laine via use-livecode:

Dear List Members and LiveCode Gurus,

We're looking for your help. I should start with a warning though: this chatGPT 
is addictive, you might lose a few hours of your life. But you will have fun 
doing it :)

Our idea is that the chatGPT to be found here:

https://chat.openai.com

is a truly great way for beginners and even more advanced LiveCoders to get 
into LiveCode, learn scripting and get common scripts pre-written. It could be 
an immensely powerful tool in bringing LiveCode to a wider audience. (In case 
you haven't already wasted hours of your life playing with it, you can ask it 
to write a LiveCode script for any specific task, and it will take its best 
shot, with very passable results).

Unfortunately, at the moment it has a tendency to make schoolboy (or should I 
say schoolbot?) errors. It gets the script mostly right but there will be 
issues that a beginner would not know how to correct. We would love it if the 
community could help it learn to write beautiful, correct, LiveCode. To do 
this, it needs feedback. We asked it how to give feedback and this is what it 
suggests:

There are a few ways you can provide feedback on ChatGPT:
• You can submit feedback through the OpenAI website by navigating to 
the “Contact Us” page and selecting “Model Feedback” as the topic.
• You can also provide feedback on the OpenAI GPT-3 forum, where users 
can share their experiences, ask questions, and provide suggestions for 
improving the model.
• You could also contact the openai team to share more detailed 
feedback and suggestions
In any case, clear and detailed feedback is helpful in understanding the 
problem and improve the model.

Just telling it what it got wrong we think isn't the right way to go, it will 
only correct it for that particular session. We need to rat on it in to its 
teachers so that it can be trained correctly.

So, if you are up for the challenge and you have a bit of time to kill, we'd 
love you to a) ask chatGPT to write LiveCode scripts and b) send any resulting 
errors with detailed feedback to the Model Feedback area.

Thank you!

Best Regards and Happy New Year to all!

Heather

Heather Laine
Customer Services Manager
LiveCode Ltd
www.livecode.com




___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode


___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Re: Training the AI to write better LiveCode

[Mike Kerner via use-livecode]

+1

On Wed, Jan 11, 2023 at 9:59 AM Heather Laine via use-livecode <
use-livecode@lists.runrev.com> wrote:

> Dear List Members and LiveCode Gurus,
>
> We're looking for your help. I should start with a warning though: this
> chatGPT is addictive, you might lose a few hours of your life. But you will
> have fun doing it :)
>
> Our idea is that the chatGPT to be found here:
>
> https://chat.openai.com
>
> is a truly great way for beginners and even more advanced LiveCoders to
> get into LiveCode, learn scripting and get common scripts pre-written. It
> could be an immensely powerful tool in bringing LiveCode to a wider
> audience. (In case you haven't already wasted hours of your life playing
> with it, you can ask it to write a LiveCode script for any specific task,
> and it will take its best shot, with very passable results).
>
> Unfortunately, at the moment it has a tendency to make schoolboy (or
> should I say schoolbot?) errors. It gets the script mostly right but there
> will be issues that a beginner would not know how to correct. We would love
> it if the community could help it learn to write beautiful, correct,
> LiveCode. To do this, it needs feedback. We asked it how to give feedback
> and this is what it suggests:
>
> There are a few ways you can provide feedback on ChatGPT:
> • You can submit feedback through the OpenAI website by navigating
> to the “Contact Us” page and selecting “Model Feedback” as the topic.
> • You can also provide feedback on the OpenAI GPT-3 forum, where
> users can share their experiences, ask questions, and provide suggestions
> for improving the model.
> • You could also contact the openai team to share more detailed
> feedback and suggestions
> In any case, clear and detailed feedback is helpful in understanding the
> problem and improve the model.
>
> Just telling it what it got wrong we think isn't the right way to go, it
> will only correct it for that particular session. We need to rat on it in
> to its teachers so that it can be trained correctly.
>
> So, if you are up for the challenge and you have a bit of time to kill,
> we'd love you to a) ask chatGPT to write LiveCode scripts and b) send any
> resulting errors with detailed feedback to the Model Feedback area.
>
> Thank you!
>
> Best Regards and Happy New Year to all!
>
> Heather
>
> Heather Laine
> Customer Services Manager
> LiveCode Ltd
> www.livecode.com
>
>
>
>
> ___
> use-livecode mailing list
> use-livecode@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode
>


-- 
On the first day, God created the heavens and the Earth
On the second day, God created the oceans.
On the third day, God put the animals on hold for a few hours,
   and did a little diving.
And God said, "This is good."
___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode

Training the AI to write better LiveCode

[Heather Laine via use-livecode]

Dear List Members and LiveCode Gurus,

We're looking for your help. I should start with a warning though: this chatGPT 
is addictive, you might lose a few hours of your life. But you will have fun 
doing it :)

Our idea is that the chatGPT to be found here:

https://chat.openai.com

is a truly great way for beginners and even more advanced LiveCoders to get 
into LiveCode, learn scripting and get common scripts pre-written. It could be 
an immensely powerful tool in bringing LiveCode to a wider audience. (In case 
you haven't already wasted hours of your life playing with it, you can ask it 
to write a LiveCode script for any specific task, and it will take its best 
shot, with very passable results).

Unfortunately, at the moment it has a tendency to make schoolboy (or should I 
say schoolbot?) errors. It gets the script mostly right but there will be 
issues that a beginner would not know how to correct. We would love it if the 
community could help it learn to write beautiful, correct, LiveCode. To do 
this, it needs feedback. We asked it how to give feedback and this is what it 
suggests:

There are a few ways you can provide feedback on ChatGPT:
• You can submit feedback through the OpenAI website by navigating to 
the “Contact Us” page and selecting “Model Feedback” as the topic.
• You can also provide feedback on the OpenAI GPT-3 forum, where users 
can share their experiences, ask questions, and provide suggestions for 
improving the model.
• You could also contact the openai team to share more detailed 
feedback and suggestions
In any case, clear and detailed feedback is helpful in understanding the 
problem and improve the model.

Just telling it what it got wrong we think isn't the right way to go, it will 
only correct it for that particular session. We need to rat on it in to its 
teachers so that it can be trained correctly. 

So, if you are up for the challenge and you have a bit of time to kill, we'd 
love you to a) ask chatGPT to write LiveCode scripts and b) send any resulting 
errors with detailed feedback to the Model Feedback area. 

Thank you!

Best Regards and Happy New Year to all!

Heather

Heather Laine
Customer Services Manager
LiveCode Ltd
www.livecode.com




___
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode