I guess if we consider “classic search” to be unreliable this will one day not be any worse. And might have some utility of its own.
>From my POV I’m going to concentrate on making the infrastructure workable. >And let others do the “believing”, at least for now. So stuff around AIU >Instrumentation and Spyre planning, perhaps. Cheers, Martin From: IBM Mainframe Discussion List <[email protected]> on behalf of roscoe5 <[email protected]> Date: Monday, 26 May 2025 at 18:32 To: [email protected] <[email protected]> Subject: [EXTERNAL] Re: RTFM Yes, for those things that I know well, I don’t ask AI. And for the deep and individual questions such as what are the implications of setting a parameter to some value, AI may/not point out something, but I need to do my own work. However, there are many things I don’t keep in my head that are documented. I sometimes find AI faster and better than multiple IBM manuals. Sent from [Proton Mail](https://proton.me/mail/home) for iOS On Mon, May 26, 2025 at 11:17 AM, Lennie Bradshaw <[[email protected]](mailto:On Mon, May 26, 2025 at 11:17 AM, Lennie Bradshaw <<a href=)> wrote: > Martin, > I agree that accuracy is low. However, the ability to reason with AI is good. > It can be persuaded it is wrong and go and re-examine the question and its > sources. It can state what its sources are, and can produce other > suggestions. It's a bit like have a colleague to discuss the subject with; > except this colleague has all the references to hand immediately. > > Lennie > > -----Original Message----- > From: IBM Mainframe Discussion List <[email protected]> On Behalf Of > Martin Packer > Sent: 26 May 2025 09:49 > To: [email protected] > Subject: Re: RTFM > > Periodically I test Generative AI with stuff I darned well know about. I'm > still unimpressed. > > Oh well, accuracy seems to be deprecated in favour of expedience. > > Cheers, Martin > > From: IBM Mainframe Discussion List <[email protected]> on behalf of > Andrew Rowley <[email protected]> > Date: Sunday, 25 May 2025 at 23:56 > To: [email protected] <[email protected]> > Subject: [EXTERNAL] Re: RTFM > On 26/05/2025 3:31 am, Lennie Bradshaw wrote: >> I have found AI to be really useful in negotiating the wealth of IBM >> documentation. I can ask AI a question that is pretty detailed and get a >> meaningful response. It is not always right, but I can then discuss it and >> refine it with the AI and find the correct answer. > > My attempts to use AI for mainframe related stuff have yielded 100% wrong > answers. Some have been quite convincing and taken some research to find it's > wrong, but still 100% wrong. > > Non-mainframe stuff I would say the rate is about 50%. > > If there's a lot of examples of something out there it maybe generates a good > result. But there's no actual intelligence in AI. What worries me most is > what happens when knowledge changes, e.g. if a new feature is developed that > makes old information obsolete. How does that get incorporated in the AI > model? Where does the training data come from, and how does the model verify > the reliability? > > -- > Andrew Rowley > Black Hill Software > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, send email to > [email protected] with the message: INFO IBM-MAIN > > Unless otherwise stated above: > > IBM United Kingdom Limited > Registered in England and Wales with number 741598 Registered office: > Building C, IBM Hursley Office, Hursley Park Road, Winchester, Hampshire SO21 > 2JN > > ---------------------------------------------------------------------- > 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 Unless otherwise stated above: IBM United Kingdom Limited Registered in England and Wales with number 741598 Registered office: Building C, IBM Hursley Office, Hursley Park Road, Winchester, Hampshire SO21 2JN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
