Generative AI is like a genie that gives you what you asked for but not what 
you want.

I asked ChatGPT: "Please write a Linux C subprogram to return the CPU time used 
by the current process". It gave me an answer. I also googled, and found a 
website that discussed "all the ways to get CPU time on Linux", with examples.

I tested the solution on z/OS, but always got zeros back.

It turns out I should have asked ChatGPT for the CPU time *in microseconds*, 
i.e. with a high-precision time service. The answers from ChatGPT and the 
website only had a 10th of a second precision. My tests didn't use more than .1 
CPU seconds, so didn't work.

This requires calls to a completely different function* than the answers I got.

* need to use clock_gettime() or getrusage(), not times().

-----Original Message-----
From: IBM Mainframe Discussion List <[email protected]> On Behalf Of 
Lionel B Dyck
Sent: Monday, January 13, 2025 10:50 AM
To: [email protected]
Subject: Re: AI makes stuff up

And the AI will conflate information from multiple sources - for
example it may take information from Linux and apply it to an answer
for z/OS.

Trust but verify is always a wise move.

On Mon, Jan 13, 2025 at 10:48 AM Bob Bridges
<[email protected]> wrote:
>
> I don't see anything wrong with asking ChatGPT and checking out the options 
> it offers, if you feel you have the time.  But if it doesn't work, I'd just 
> shrug and forget it.  Those AI engines make stuff up sometimes.  I gather 
> their target is to write something that is grammatical and often even 
> idiomatic, but there's no ability to discern when TRUTH is required.
>
> ---
> Bob Bridges, [email protected], cell 336 382-7313
>
> /* I would as soon write free verse as play tennis with the net down.  
> -Robert Frost */
>
> -----Original Message-----
> From: IBM Mainframe Discussion List <[email protected]> On Behalf Of 
> Hobart Spitz
> Sent: Monday, January 13, 2025 11:24
>
> Update from ChatGPT:  For ISPF edit macros, use the *IRXJSTUB* stub. It is 
> designed for REXX programs that interact with ISPF.
>
> Seems like it would be worth a try.
>
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO IBM-MAIN



--
Lionel B. Dyck <><
Website:https://github.com/lbdyck

"Worry more about your character than your reputation.  Character is
what you are, reputation merely what others think you are." - John
Wooden

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN



----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to