AI discussions

johnraff · 2026-08-18 04:27:24

^worth sharing.

greenjeans · 2026-09-03 13:59:38

Pertinent to this subject, Debian voted on a new general resolution about AI, the winner was option #5 which basically says they neither endorse nor condemn it's use and developers are free to use it but asks that they do so responsibly. Scroll about halfway down that page for #5, and to the bottom for results.
https://www.debian.org/vote/2026/vote_002#texta
Snippet:

"Debian neither endorses nor prohibits the use of generative AI tools in the development, maintenance, or documentation of software, packaging, documentation, and other media published within the Debian Project. We recognize that such tools can substantially improve the productivity of contributors when used responsibly, allowing volunteers to spend more of their limited time on work that requires technical expertise, judgment, review, and collaboration.
The Debian Project nevertheless expects that all contributions submitted to Debian, regardless of how and with which tools they were produced, satisfy the same standards of quality, correctness, maintainability, and legal compliance. The use of a generative AI tool does not diminish the contributor's responsibility for the work they submit. Contributors are expected to understand, review, test, and, where appropriate, modify AI-assisted output before incorporating it into Debian. Blindly accepting or uploading AI-generated material without appropriate human review is inconsistent with Debian's established development practices. We encourage our contributors to disclose whether a contribution was made with AI assitance, but do not require them to do so."

Sector11 · 2026-09-03 15:55:41

Personally I would have preferred:

Blindly accepting or uploading AI-generated material without appropriate human review is inconsistent with Debian's established development practices. You must make a disclosure if a contribution was made with AI assistance."

But it is a good #5 regardless

greenjeans · 2026-09-08 00:22:40

I'm sorry, just had to do one more meme on this subject, please forgive the large image:

Sector11 · 2026-09-08 16:56:23

Frame #4 - can't you just see the evil thoughts in Bug's and Daffy’s heads.

LOVE IT!

trilobite · 2026-09-08 19:05:39

I like what you wrote.

We require disclosure like this from all who write things for us. But there's no secure answer. I get sorely annoyed when the disclosure is "AI and other computer generated content may be reponsible for this content". (swearing from me now)

Sector11 wrote:

Personally I would have preferred:
Blindly accepting or uploading AI-generated material without appropriate human review is inconsistent with Debian's established development practices. You must make a disclosure if a contribution was made with AI assistance."
But it is a good #5 regardless

hhh · 2026-09-15 01:56:58

A thought (maybe not original). Less artificial intelligence, more human intelligence. I see a computer model being trained by idiots. IMO, that's not a model for success.

johnraff · 2026-09-17 05:49:00

AI safety of course gets a lot of attention these days. You've all probaby read about the Hugging Face hacking by escaped AI agents from an Open AI sandbox that wasn't as leakproof as they thought.

AI ethics is a separate issue that's coming up. Not before time.

This long read in the Guardian is quite interesting, although it doesn't come up with any answers:

https://www.theguardian.com/news/ng-int … e-deepmind

Martin · 2026-09-17 20:19:28

johnraff wrote:

AI safety of course gets a lot of attention these days. You've all probaby read about the Hugging Face hacking by escaped AI agents from an Open AI sandbox that wasn't as leakproof as they thought.
AI ethics is a separate issue that's coming up. Not before time.
This long read in the Guardian is quite interesting, although it doesn't come up with any answers:
https://www.theguardian.com/news/ng-int … e-deepmind

Interesting indeed and, a bit long.

I like the idea of training an AI on all human knowledge available in 1911 and then wait and see if it comes up with General Relativity.

Given all the attention given to AI technology and usage I find it interesting that I have such problems finding use cases where AI/LLM use actually help me. One that seems to work reliably for me is asking some LLM in which issue of National Geographic's Magazine I can find articles about subject ABC. I have 40 years of NGM at home and finding stuff manually is time consuming...

Recently I thought I finally would be able to try LLM-assisted Python coding. I wanted to transform a collection of scripts into an app with a GUI. This would be the first time I created a proper GUI and I thought I could get 'helpful hints' from AI. It turned out I never had to try that as old-fashioned internet search found all information I needed in video tutorials, forum posts, blogs etc.

/Martin

johnraff · Yesterday 09:00:37

I also find AI handy for some tasks that would otherwise take a long time, or be extremely difficult (eg some keywords are almost impossible to search for). Just yesterday I asked what an AI thought of my attempts to tweak the startup chain of a Wayland session, and along with some direct answers - which might or might not have been 100% correct - it also gave me (without being asked) a longish checklist to go over for possible issues. That strikes me as something genuinely useful, and a checklist for humans to act on - or not - isn't surrendering any power to the machine at all.

OK so we mortals sometimes find AI as it already is handy at times.

But at the top, there seems to be an oligarchy directing its development who are seriously weird. All - or most of - those techbros are in the grip of a highly dangerous AI cult, according to Nathalie Maréchal in techpolicy.press:
https://www.techpolicy.press/what-polic … ult-of-ai/

Don't dismiss this as hyperbole till you at least read the first few paragraphs, and follow some of the reference links, which are to solid sources.

Terrifying stuff actually.

Sector11 · Yesterday 14:15:37

Really - this is getting closer to Terminator's Skyney than fiction.
OK not there yet but Elon said at the World Economic Forum, where earlier this year he predicted that within five years “AI will be smarter than all of humanity collectively.”

That scares the brown stuff outta me!

Still reading.

EDIT:
From one of the links in the article:

You live in a deranged age — more deranged than usual, because despite great scientific and technological advances, man has not the faintest idea of who he is or what he is doing.
Walker Percy

Last edited by Sector11 (Yesterday 14:24:29)

PackRat · Yesterday 14:46:32

Sector11 wrote:

Really - this is getting closer to Terminator's Skyney than fiction.
OK not there yet but Elon said at the World Economic Forum, where earlier this year he predicted that within five years “AI will be smarter than all of humanity collectively.”
That scares the brown stuff outta me!
Still reading.
EDIT:
From one of the links in the article:
You live in a deranged age — more deranged than usual, because despite great scientific and technological advances, man has not the faintest idea of who he is or what he is doing.
Walker Percy

Don't worry. I have a plan:

hhh · Yesterday 17:16:18

Where did you get that dress, it's awful, and those shoes and that coat, jeeeeez!

I prompted duckduckgo's AI for Johnny Henshaw-Jackson quotes and it didn't come up with my favorite from Airplae II: The Sequel, where Lloyd Bridges as Steve McCroskey says "Jacobs! I want to know absolutely everything that's happened up till now." and Jacobs (after a dramatic leap into frame) starts explaining the cooling of the Earth, the dinosaurs getting too fat and Prince Charles wearing Lady Di's clothes. I researched it myself and found it in Airplane II. The AI did find it when I prompted with "jacobs airplane II quotes", but if AI can't expand its search to include all of Stephen Stuckers Airplane quotes, its worthless to everyone except for the billionaires who have already decided that the world is ending and want AI to take over the peasant's work.
https://youtu.be/-iWs_sxF5Ug?t=41

"If you've got any ideas, any ideas at all, now is the time. I want to hear them, and I want to hear them NOW."
"How 'bout a game show like Hollywood squares, but with kids, Gary Coleman could host!"

Skynet became self-aware on August 29, 1997, at 2:14 a.m. EDT. Thank goodness John Connor keeps getting sent back in time to delay the inevitable planetary destruction.

greenjeans · Today 00:13:58

It's crazy what's going on out there, and I have concerns like everyone does. What I will say is that used carefully as a learning tool (which you never see articles about this), it can be a powerful aid for an individual trying to educate themselves. I've been using it as such for over a year now to help me learn C and also Python and advanced shell-scripting. And I can't emphasize it enough, don't let the AI do all the coding for you, use it to level-up your own skills, it's like having a teachers aid right there with you to answer any questions you might have and explain code you don't yet understand.

FYI be careful, it does get things wrong very often, it should NOT be your source for everything you learn, but it is another good tool in the chest, like cppcheck and scan-build and valgrind etc. All programs just like LLM's to enable you to do better work.

Sector11 · Today 02:50:21

PackRat wrote:

Sector11 wrote:
Rw[snip}

Walker Percy
Don't worry. I have a plan:
https://iili.io/nISI4uR.jpg

Sector11 · Today 02:52:28

hhh wrote:

{SNIP}
"If you've got any ideas, any ideas at all, now is the time. I want to hear them, and I want to hear them NOW."
"How 'bout a game show like Hollywood squares, but with kids, Gary Coleman could host!"
Skynet became self-aware on August 29, 1997, at 2:14 a.m. EDT. Thank goodness John Connor keeps getting sent back in time to delay the inevitable planetary destruction.

Ummmmmmmmmmmmm fresh out!
Pass the tin foil.

johnraff · Today 04:14:43

Agree with @greenjeans that, used carefully, the AI LLM tools we currently have access to can be quite useful for coders like us.
In the medical field, it seems that very valuable work is being done that wouldn't otherwise have been possible.

Nobody cares what I think, but if it was up to me we could just stop building data centres and hone the AI tools we already have. And maybe look at some alternatives to LLMs.

But, a quote from that Nathalie Maréchal essay:

Once you understand the broligarchs care more about hypothetical, future AI-enhanced lives than today’s population and believe it is their right and duty to sit atop a brutally hierarchical society where cold “rationality” trumps human rights and dignity, a lot of things start making sense.

greenjeans · Today 15:17:56

johnraff wrote:

Agree with @greenjeans that, used carefully, the AI LLM tools we currently have access to can be quite useful for coders like us.
In the medical field, it seems that very valuable work is being done that wouldn't otherwise have been possible.

Yep. It cuts both ways, useful at times, and flat out wrong at times, this is why I shudder when I think of heavy AI usage in the kernel, 30 million+ lines of code it just doesn't seem possible that someone is actually auditing the gargantuan amount of patches AI is generating for it.

The other day I was being lazy, so I asked the Brave browser search assist (which is a frontend for Google Gemini) to see if there was a .deb package for Yad version 15.0. It went like this:

AI : version 15.0 of yad doesn't exist, last version is 0.40
me: *pastes link to yad's gihub*
AI: oh okay i was wrong, but there are no packages for it
me : did you check the Opensuse Build Service?
AI: oh okay, yes there is a package there but it won't work for you as OBS only has RPM's.
me: *sigh*

So then I did what I should have done in the first place and checked OBS myself and of course they had a .deb package. Tons of them for version 14 and a couple for 15.

But at the same time it can also be useful, I wanted a way to generate a simple snapshot of a C script as part of a script I was trying to make to do a cppcheck and run scan-build on a given file with just a right-click. The snapshot part is not actually that useful, but I wanted it anyway, it looks cool, lol. Tons of awk'ing and that works but it's not always perfect, as I was wanting a way to list all the functions in the report, includes, size, number of lines etc. as well.

So it's not possuble for me to be aware of all the 70,000 packages in Debian's repo, not to mention some 100,000 more third party stuff. So I used a different LLM and asked about a simpler way to do what I wanted, and the AI pointed me towards the universal-ctags package, and by golly that worked like a charm. And that's not the first time it's pointed me towards a utility I didn't know existed.

FYI that's why I wanted a newer version of yad compiled with html support, I currently pipe the output into a text file and it opens in a nice yad text-info window, but scan-build generates it's report (if it finds anything) in an html file, currently I have it set to open in my browser, but was hoping for something faster and less clunky. I haven't ever used yad's --html dialog, so I may be barking up the wrong tree anyway, has anybody here made use of it before?

#81 2026-08-18 04:27:24

Re: AI discussions

#82 2026-09-03 13:59:38

Re: AI discussions

#83 2026-09-03 15:55:41

Re: AI discussions

#84 2026-09-08 00:22:40

Re: AI discussions

#85 2026-09-08 16:56:23

Re: AI discussions

#86 2026-09-08 19:05:39

Re: AI discussions

#87 2026-09-15 01:56:58

Re: AI discussions

#88 2026-09-17 05:49:00

Re: AI discussions

#89 2026-09-17 20:19:28

Re: AI discussions

#90 Yesterday 09:00:37

Re: AI discussions

#91 Yesterday 14:15:37

Re: AI discussions

#92 Yesterday 14:46:32

Re: AI discussions

#93 Yesterday 17:16:18

Re: AI discussions

#94 Today 00:13:58

Re: AI discussions

#95 Today 02:50:21

Re: AI discussions

#96 Today 02:52:28

Re: AI discussions

#97 Today 04:14:43

Re: AI discussions

#98 Today 15:17:56

Re: AI discussions

Board footer