I've had mixed results with AI, specifically Grok. It was dead on about MIH settings and virtual tape libraries. It wasn't close about other things, except that it was a general pointer into where to look.
Considering the sheer volume of information out there these days about z/OS, even hallucinating commands isn't a total loss if it points you to look in the right area. On Sun, May 25, 2025 at 7:35 PM Steve Beaver < [email protected]> wrote: > If AI can build a knowledge base of what we do as a group. It would be > valuable > > Sent from my iPhone > > No one said I could type with one thumb > > > On May 25, 2025, at 19:33, roscoe5 < > [email protected]> wrote: > > > > Yes, I too have found different AIs to be helpful at times and also > fabricating MVS commands that don’t exist. And yes, when I push back enough > it may admit it had extrapolated from other commands. > > > > My question is, given that there is a lot of AI power and success in > some areas, why not in z/OS, etc.? > > > > What is IBM doing in this space? > > I’ve heard a little about Watsonx. Surely, if IBM wanted to, they could > make a powerful and useful instance around mainframe products. > > > > I don’t see it replacing IBM Support, but it could ingest manuals and > other knowledge, making it easy to find the syntax for that obscure command > with examples. > > > > Giving advice on how to implement and tune things for a given shop would > be limited. IBM Support treads softly in some of these areas. > > > > But if a Junior SysProg could find an answer to specific question > without having to spend hours in manuals, that should help with the > knowledge retirement issues. > > > > Bob > > Sent from [Proton Mail](https://proton.me/mail/home) for iOS > > > >> On Sun, May 25, 2025 at 8:04 PM, Seymour J Metz <[[email protected] > ](mailto:On Sun, May 25, 2025 at 8:04 PM, Seymour J Metz <<a href=)> > wrote: > >> > >> Don't judge what AI will eventually be able to do by current artificial > stupidity engines. > >> > >> -- > >> Shmuel (Seymour J.) Metz > >> http://mason.gmu.edu/~smetz3 > >> עַם יִשְׂרָאֵל חַי > >> נֵ֣צַח יִשְׂרָאֵ֔ל לֹ֥א יְשַׁקֵּ֖ר > >> > >> ________________________________________ > >> From: IBM Mainframe Discussion List <[email protected]> on > behalf of Andrew Rowley <[email protected]> > >> Sent: Sunday, May 25, 2025 7:51 PM > >> To: [email protected] <[email protected]> > >> Subject: Re: RTFM > >> > >> External Message: Use Caution > >> > >>> On 26/05/2025 9:15 am, Lennie Bradshaw wrote: > >>> Wrong answers can be challenged in AI. You can ask AI where the > references come from. You can suggest alternatives and get a changed > response. This isn't just a better google. This is useful to me. Work with > the AI and together you can work out things that are difficult on your own. > >> I'm not saying it can't be useful. But it produces wrong answers with > >> equal confidence as correct answers. If you know enough to challenge the > >> wrong answers that's OK. But it's used in lots of situations where > >> people won't recognize the wrong answers. > >> > >> The success of AI is based on situations where people want convincing > >> answers quickly and cheaply but don't care whether they are correct e.g. > >> customer support. I don't think that's going to have good results long > >> term... > >> > >> -- > >> Andrew Rowley > >> Black Hill Software > >> > >> ---------------------------------------------------------------------- > >> 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 > > > > ---------------------------------------------------------------------- > > 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 > -- Jay Maynard ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
