Re: Debian Monthly : AI News Report 2024/10

2024-11-10 Thread Richard Lewis
Mo Zhou  writes:

> 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.

I think it's slightly misdiagnosed here. to me, it comes accross as
"tedious" to read because the tone is more like a university essay with
everything being overly emphatic adjectives and adverbs everywhere.

(everything is a "flurry of activity" or "gaining traction" or
"bolstering" or "a bid to enhance" ... and everything is a "significant
devlopment".. it's all just too much.)

It's perhaps the training data that causes this -- this is how a lot of
people-who-dont-write write because schools and universities teach them
to write essays but not read/edit -- and the aim ends up being to
convince the reader that they are giving an opinion, which is not what
you want here.

So it might help to tell it to be more factual and neutral in
tone. Maybe you could even try telling it to adopt the style used on the
debian lists themselves? (lwn.net is also a good reference point)



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 against that?
>  
> first: citation needed.
> second: summaries written by applied statistics systems will not help.
> (also citation needed I guess.) 

I'm glad I know Holger in person, so I can smile instead of feeling furious
about this kind of input :-)

> -> please post these summaries to a dedicated applied statistics summary
> mailinglist, but please dont spam the original lists with this bot content.

Hmm... An opt-in llm-generated-monthly-digest mailinglist/feed/whatever... Not
really a fan of this kind of content (yet?), but I believe that something like
this could be seen as a fair balance between avoiding annoyance and providing a
space for experimentation on the subject within Debian.

> btw, the signature below was choosen by an 'artificial intelligence'
> system called fortune. it's so amazing and wise!!1

See how this whole thing has its fun side. Let's just relax and play,
responsibly(-ish). That said, please no LLM feeds in planet.d.o :p

+thanks lumin for all the effort and thought you have put into this.

Bests,

-- 
Tiago


signature.asc
Description: PGP signature


Re: Debian Monthly : AI News Report 2024/10

2024-11-10 Thread G. Branden Robinson
At 2024-11-10T11:21:43+, Richard Lewis wrote:
> > 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.
> 
> I think it's slightly misdiagnosed here. to me, it comes accross as
> "tedious" to read because the tone is more like a university essay
> with everything being overly emphatic adjectives and adverbs
> everywhere.

This is a better description than mine.  I carelessly attributed the
impact the generated content had on me as a reader to properties of the
prose style, which wasn't quite right.  I'm so accustomed to that style
being simultaneously purple and lackadaisical that I expressed the
boredom it aroused in me through long, miserable experience with
corporate communications that do not reward the effort of reading.

> (everything is a "flurry of activity" or "gaining traction" or
> "bolstering" or "a bid to enhance" ... and everything is a "significant
> devlopment".. it's all just too much.)

Yes.  Such phrases rapidly weary the reader when the factual payoff for
what they promise is almost always meager, and thus has a perversely
opposite effect from what was apparently intended.  The style touts a
firm's business doings as matters of high adventure when they better
resemble a long session in a sensory deprivation tank.  Most comm majors
seem to think they're writing for the J. Peterman catalog.

As is often the case in matters of style, William Strunk and George
Orwell knew whereof they spoke.

Regards,
Branden


signature.asc
Description: PGP signature


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 applied statistics systems will not help.
(also citation needed I guess.) 

-> please post these summaries to a dedicated applied statistics summary
mailinglist, but please dont spam the original lists with this bot content.

btw, the signature below was choosen by an 'artificial intelligence'
system called fortune. it's so amazing and wise!!1


-- 
cheers,
Holger

 ⢀⣴⠾⠻⢶⣦⠀
 ⣾⠁⢠⠒⠀⣿⡁  holger@(debian|reproducible-builds|layer-acht).org
 ⢿⡄⠘⠷⠚⠋⠀  OpenPGP: B8BF54137B09D35CF026FE9D 091AB856069AAA1C
 ⠈⠳⣄

Smart things make us dumb.


signature.asc
Description: PGP signature


Bug#1087287: ITP: golang-github-google-flatbuffers -- FlatBuffers: Memory Efficient Serialization Library

2024-11-10 Thread Andreas Henriksson
Package: wnpp
Severity: wishlist
Owner: Andreas Henriksson 

* Package name: golang-github-google-flatbuffers
  Version : 24.3.25-1
  Upstream Author : Google
* URL : https://github.com/google/flatbuffers
* License : Apache-2.0
  Programming Lang: Go
  Description : FlatBuffers: Memory Efficient Serialization Library

 FlatBuffers is a cross platform serialization library architected
 for maximum memory efficiency. It allows you to directly access
 serialized data without parsing/unpacking it first, while still having
 great forwards/backwards compatibility.
 .
 This is a dependency needed for updating badger to a newer upstream release.