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
source-only uploads or HTTPS URLs in /etc/apt/sources.list), and a
couple of years later they are mainstream!
Those new technologies are sure to make more impact in the future. We
need to find a proper place where people are comforatable with LLM
noises. Currently it seems to me posting AI-generated news on the main
mailing lists leads to a "pollution effect" -- LLM will grab its report
from the last month If I don't mention it...
What do people do when there is a long thread on debian-devel,
debian-project, debian-private or the like? Well, what I do is that I
check the first ~6 messages and then cherry-pick 3 or 4 answers deeper
in the thread from
I do more or less the same, or even worse -- keep them marked as unread
and see my unread mail in inbox go to 10k+. Those LLM tools, while
noisy, can still provide some help for me to keep me updated a little
bit -- to know what happened, what are being discussed and what is the
conclusion.
The LLM reports are not quite accurate. But I still see value there --
as long as we ask it to provide URL citations to the information source,
I can still catch up something with minimum read.
I see a big transformative potential for our future discussions: even if
a crowd is shouting circular arguments around, we can use AI to reassure
participants that original point of views can have good chances to be
part of a summary. Taking the effort to contribute is rewarded. This
can change Debian considerably. So please, more DebGPT summaries !
DebGPT is now more about a general terminal LLM tool which I develop
and use very frequently. Summarizing mailing list is just one of the
not yet announced new features. Efforts in exploring the way to make
LLM useful will continue.
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 direct those generated
contents, with the audience being comfortable with LLM noise? As noted
by the ex-DPLs, it should happen at a dedicated experiment ground.
