Re: Things learned from the expeirment (Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-16 Thread Joerg Jaspert
On 17414 March 1977, Mo Zhou wrote: 1. Let LLM answer the NM templates (maybe with debian policy or debian developer reference in context) and see the percentage of questions that can be answered correctly. Even if I don't do it, maybe new DD applicants will. And those who actually do this sho

Re: Things learned from the expeirment (Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-15 Thread Charles Plessy
Hi Mo, thanks again for your posts, I was just thinking that the debian-mentors list could be a good target for summarisation too: it is high traffic, email subject lines are focused on what to upload, but discussions are focused on problem-solving, thus some intersting tips & trick will be easy

Things learned from the expeirment (Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-15 Thread Mo Zhou
Hi folks, On 11/9/24 01:26, DebGPT wrote: This is an experiment, by letting LLM go through all 369 emails from debian-devel on Oct. I received lots of feedbacks from the experiments, from positive ones to negative ones. It wasn't discouraging to see negative feedbacks since that is usually wha

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-15 Thread Steve McIntyre
lu...@debian.org wrote: > >While hallucinating too much, LLMs can still correctly >teach me how to use urwid (which I could never understand by going >through their tutorial many times...). >That leads to the `debgpt config` TUI configuration wizard. > >Any suggestion on a place where I can safely

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-14 Thread Sean Whitton
Hello, On Sun 10 Nov 2024 at 08:24am GMT, Holger Levsen wrote: > On Sun, Nov 10, 2024 at 08:48:21AM +0900, Charles Plessy wrote: >> Our mailing lists were a ground-breaking technological avance in the >> past that would open Debian to the whole World, but now are they not >> working exactly again

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-10 Thread Tiago Bortoletto Vaz
Hi, On Sun, Nov 10, 2024 at 08:24:19AM GMT, Holger Levsen wrote: > On Sun, Nov 10, 2024 at 08:48:21AM +0900, Charles Plessy wrote: > > Our mailing lists were a ground-breaking technological avance in the > > past that would open Debian to the whole World, but now are they not > > working exactly a

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-10 Thread Holger Levsen
On Sun, Nov 10, 2024 at 08:48:21AM +0900, Charles Plessy wrote: > Our mailing lists were a ground-breaking technological avance in the > past that would open Debian to the whole World, but now are they not > working exactly against that? first: citation needed. second: summaries written by applie

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Mo Zhou
Hi Charles, On 11/9/24 15:48, Charles Plessy wrote: Thanks a lot Mo for this exciting experiment! And having two ex-DPLs pressing the big red stop button is not necessarly a bad sign in an ageing project. Often you will see ideas rejected in a very dismissive if not insulting way (for example

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Charles Plessy
Thanks a lot Mo for this exciting experiment! And having two ex-DPLs pressing the big red stop button is not necessarly a bad sign in an ageing project. Often you will see ideas rejected in a very dismissive if not insulting way (for example source-only uploads or HTTPS URLs in /etc/apt/sources.l

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread G. Branden Robinson
At 2024-11-09T14:46:45-0800, Mo Zhou wrote: > The tone can change: http://paste.debian.net/1335055/ I recognize a different style there... > LLMs are being improved rapidly over time. ...but I'm not sure I would call the new example an improvement. With all those exclamation marks it seems more

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Mo Zhou
The tone can change: http://paste.debian.net/1335055/ LLMs are being improved rapidly over time. I guess it's due to some potential safety issues so that LLM uses a dull corporate tone by default. Those models should have been trained on different tones, as long as we instruct it to use them. On

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread G. Branden Robinson
At 2024-11-09T21:44:40+, Steve McIntyre wrote: > Please, no further. We don't need hallucinated summaries on our > lists. If you want to publish them, publish them somewhere separately > IMHO. Oh, good--since it's not a CoC violation to express an unflattering opinion of this experiment, did a

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Steve McIntyre
lu...@debian.org wrote: >I just realized that the news report could be more useful if it cites >the information source. Here we go: > >debgpt -Hx ldo:debian-devel/2024/10 -a 'write a news report based on the >provided information. Cover as many topics as possible. You may expand a >little bit on

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Serafeim (Serafi) Zanikolas
On Sat Nov 9, 2024 at 10:00 PM CET, Mo Zhou wrote: > I just realized that the news report could be more useful if it cites > the information source. Here we go: thanks! the links are really key, given the known hallucination issues of LLMs signature.asc Description: PGP signature

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Mo Zhou
I just realized that the news report could be more useful if it cites the information source. Here we go: debgpt -Hx ldo:debian-devel/2024/10 -a 'write a news report based on the provided information. Cover as many topics as possible. You may expand a little bit on important matter. include lin

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Mo Zhou
The LLM I used to produce that exact news report was gpt-4o-mini, from openai. ChatGPT is the name of openai's LLM web interface and its underlying LLM model name could change. It took roughly 3 minutes to perform the bulk API calls. That said, I basically implemented support for all commonly see

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Jeremy Stanley
On 2024-11-09 14:19:53 +0100 (+0100), PICCA Frederic-Emmanuel wrote: > is it via ChatGPT or an llm self hosted ? [...] It's DebGPT: https://salsa.debian.org/deeplearning-team/debgpt -- Jeremy Stanley signature.asc Description: PGP signature

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread PICCA Frederic-Emmanuel
is it via ChatGPT or an llm self hosted ? Can we imagine having a Debian hosted computer with and AMD GPU dedicated to this use case ? Se should provide these summaries letter for most of our mailing list :) cheers Fred - Le 9 Nov 24, à 14:09, Hector Oron zu...@debian.org a écrit : > Hel

Re: Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread Hector Oron
Hello Lumin, El sáb, 9 nov 2024 a las 10:27, DebGPT () escribió: > > This is an experiment, by letting LLM go through all 369 emails from > debian-devel on Oct. The command for producing the news report > is included below. Use debgpt's git HEAD if you want to try. First time I see this kind of e

Debian Monthly [debian-devel]: AI News Report 2024/10

2024-11-09 Thread DebGPT
This is an experiment, by letting LLM go through all 369 emails from debian-devel on Oct. The command for producing the news report is included below. Use debgpt's git HEAD if you want to try. THE FOLLOWING CONTAINS AI-G