ThomasWaldmann changed the topic of #moin to: MoinMoin Wiki - release: 1.9.11, beta: 2.0.0b2 - subscribe to https://moinmo.in/SecurityFixes - Just ask/tell AND WAIT. Log: https://libera.irclog.whitequark.org/moin
pabs3 has quit [Read error: Connection reset by peer]
pabs3 has joined #moin
RogerHaase has joined #moin
<RogerHaase> ThomasWaldmann: yes
<ThomasWaldmann> RogerHaase: tried Junie already?
ThomasWaldmann changed the topic of #moin to: MoinMoin Wiki - release: 1.9.11, beta: 2.0.0b3 - subscribe to https://moinmo.in/SecurityFixes - Just ask/tell AND WAIT. Log: https://libera.irclog.whitequark.org/moin
UlrichB has joined #moin
UlrichB has quit [Quit: Client closed]
<RogerHaase> ThomasWaldmann: no, I haven't tried Junie; wikk add to my list of todos
<ThomasWaldmann> it can be quite helpful. not always, but surprisingly often. can definitely type faster than me.
<ThomasWaldmann> has well earned its 20 bucks in the last weeks.
<ThomasWaldmann> btw, don't use it to split a monster test module into a package with separate test files. although i told it not to modify the tests, it did so and due to lacking diff tooling that supports moved blocks, that was a pain to fix.
<ThomasWaldmann> https://github.com/borgbackup/borg/pull/8863 most of PR from Junie. last commit from me, fixing what it modified.
<ThomasWaldmann> https://github.com/borgbackup/borg/pull/8882 95% of first commit by Junie, translating C code to Cython.
<RogerHaase> ThomasWaldmann: my experience with AI is mostly negative. If I could turn it off in my favorate browser, I would.
<ThomasWaldmann> I am also critical of it. It can tell you lies while sounding very convincing. And after you tell it that its output was not true, it will tell you a different story. Repeatedly... (seen with chatgpt).
<ThomasWaldmann> that's why i was surprised that junie works quite well.
<ThomasWaldmann> it sure needs supervision and direction, but can do complex tasks.
<RogerHaase> in the 80's IBM had an "AI" system for laying out stations on a factory floor. I was assisting a process engineer over a period of several months while he tweaked the program's inputs. He finally got the "AI" system to output a solution that matched what some other process engineers had achieved manually. We called it a success. End of problem.
<ThomasWaldmann> guess it depends now on whether it only learned the solution to that input or whether it now can create reasonable layouts in general, for different inputs.
<ThomasWaldmann> when I studied CS, they tried to use lisp and prolog to process language and knowledge. my diploma thesis was about a neuronal network implementation in hardware.
<ThomasWaldmann> the problem was just that the hardware sucked (a small FPGA with way to little transistors) and was nowhere powerful enough to do anything really useful.
RogerHaase has quit [Quit: Leaving.]
russell-- has quit [Ping timeout: 248 seconds]
russell-- has joined #moin
pabs3 has quit [Ping timeout: 248 seconds]