LLMs need to be treated as what they are. A 12 year old with internet connection and very fast reading/writing skills.
The first Google hit is the truth and is used. Use it as this. It is immensely helpful in stupid repetitive tasks. And you have to check after it like it is out to hurt you, because the errors it makes are human. I really don't get the issue here. We won't use an LLM as the lead. But as a tool to ease some work and speed up tasks, why not? Copyright is not an issue with short code samples. Regards Julian Am 21. März 2026 11:53:40 MEZ schrieb Bertho Stultiens <[email protected]>: >On 3/20/26 11:44 PM, Steffen Möller via Emc-developers wrote: >> Hm. I failed. I truly wanted to find some bits that we agree that >> those LLMs can help us with. >Why is this a failure? Your views and priorities in using LLMs are apparently >different than what some other people (including me) think. It is fine to have >different views. > >In the end it is a risk assessment whether to do one or the other thing. And >there we have not yet reached consensus and seen the end of discussion. > >That said, what you do /privately/ is up to you. Nobody is telling you not to >use these tools. But there is a clear separation between what you do privately >and what you contribute publicly. > >As long as there is no consensus and rule outline how to handle LLM stuff, we >all should refrain from contributing it publicly. Again, what you do privately >is up to you, but you are responsible for keeping a clear separation when you >want to contribute to the public repo(s). > > >> ... I am still quite confident that those LLMs can help making each >> of us better - yes, everyone! >That is opinion, not fact. And going by the current research on the issue, you >may actually be wrong for many use cases. However, you are welcome to do what >you want to do, /privately/. > >But, once you know what is in the magic box, the magic spontaneously >disappears. > > >> Something I yet have no answer for is that we lack the capacity to >> review everything in the level of detail that a potential killing >> machine like a CNC-whatever deserves. I mean - we cannot just let >> some robot program the code that controls robots, right? >So, why use it? If you cannot review it properly, then it is, by definition, >not safe to use. Please re-read Ken Thompsen's "Reflections on Trusting Trust" >why this is such a big issue (Trojans come in many forms, but all give the >same result). > > >> Still, and if it is only for educating ourselves on the code base of >> LinuxCNC, I consider contributions "mostly harmless" >For your /private/ use, again, go for it if you want to use it. > >But I strongly disagree that this is "mostly harmless". It is harmless at >best, a potential Trojan horse if you are just lucky and very deadly if you >get unlucky. > >To use another HHGTTG quote; You are about to be fed to the ><strike>LLM</strike> Ravenous Bugblatter Beast of Traal... >Consider how lucky you are that life has been good to you so far. >Alternatively, if life hasn't been good to you so far, which, given your >current circumstances, seems more likely, consider how lucky you are that it >won't be troubling you much longer. > > >-- >Greetings Bertho > >(disclaimers are disclaimed) > > > >_______________________________________________ >Emc-developers mailing list >[email protected] >https://lists.sourceforge.net/lists/listinfo/emc-developers -- Mit freundlichen Grüßen aus Pinneberg Julian Wingert Subtilitas Consulting Holstenstr. 25 25421 Pinneberg Phone: 0170/4516094 FAX: 03212-1479681 Mail: [email protected] USt-IdNr.: DE272503212 _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
