On Sat, Mar 18, 2023 at 3:49 PM Erik Moeller <eloquence(a)gmail.com> wrote:
>
> ...With image-generating models like Stable Diffusion, it's been found
> that the models sometimes generate output nearly indistinguishable
> from source material [1]. I don't know if similar studies have been
> undertaken for text-generating models yet.
They have, and LLMs absolutely do encode a verbatim copy of their
training data, which can be produced intact with little effort. See
https://arxiv.org/pdf/2205.10770.pdf -- in particular the first
paragraph of the Background and Related Work section on page 2, where
document extraction is considered an "attack" against such systems,
which to me implies that the researchers fully realize they are
involved with copyright issues on an enormous scale. Please see also
https://bair.berkeley.edu/blog/2020/12/20/lmmem/
On Sat, Mar 18, 2023 at 9:17 PM Steven Walling <steven.walling(a)gmail.com> wrote:
>
> The whole thing is definitely a hot mess. If the remixing/transformation by the model is a derivative work, it means OpenAI is potentially violating the ShareAlike requirement by not distributing the text output as CC....
The Foundation needs to get on top of this, by making a public request
to all of the LLM providers which use Wikipedia as training data,
asking that they acknowledge attribution of any output which may have
depended on CC-BY-SA content, licence model productions as CC-BY-SA,
and most importantly, disclaim any notion of accuracy or fidelity to
the training data. This needs to be done soon. So many people are
preparing to turn the reins of their editorial control over to these
new LLMs which they don't understand, and the problems at
CNET[https://gizmodo.com/cnet-ai-chatgpt-news-robot-1849996151], let
alone Tyler Cowen's blog, have already felt the pain but sadly decided
to hastily try to cover it up. The overarching risk here is akin to
"citogenesis" but much more pernicious.
On Sun, Mar 19, 2023 at 1:20 AM Kimmo Virtanen
<kimmo.virtanen(a)wikimedia.fi> wrote:
>
>> Or, maybe just require an open disclosure of where the bot pulled from and how much, instead of having it be a black box? "Text in this response derived from: 17% Wikipedia article 'Example', 12% Wikipedia article 'SomeOtherThing', 10%...".
>
> Current (ie. ChatGPT) systems doesn't work that way, as the source of information is lost in the process when the information is encoded into the model....
In fact, they do work that way, but it takes some effort to elucidate
the source of any given output. Anyone discussing these issues needs
to become familiar with ROME:
https://twitter.com/mengk20/status/1588581237345595394 Please see also
https://www.youtube.com/watch?v=_NMQyOu2HTo
With luck we will all have the chance to discuss these issues in
detail on the March 23 Zoom discussion of large language models for
Wikimedia projects:
https://meta.wikimedia.org/wiki/Wikimedia_Foundation_Annual_Plan/2023-2024/…
--LW
Hello friends
Short version : We need to find solutions to avoid so many africans
being globally IP blocked due to our No Open Proxies policy.
*https://meta.wikimedia.org/wiki/No_open_proxies/Unfair_blocking*
Long version :
I'd like to raise attention on an issue, which has been getting worse in
the past couple of weeks/months.
Increasing number of editors getting blocked due to the No Open Proxies
policy [1]
In particular africans.
In February 2004, the decision was made to block open proxies on Meta
and all other Wikimedia projects.
According to theno open proxiespolicy : Publicly available proxies
(including paid proxies) may be blocked for any period at any time.
While this may affect legitimate users, they are not the intended
targets and may freely use proxies until those are blocked [...]
Non-static IP addresses or hosts that are otherwise not permanent
proxies should typically be blocked for a shorter period of time, as it
is likely the IP address will eventually be transferred or dynamically
reassigned, or the open proxy closed. Once closed, the IP address should
be unblocked.
According to the policy page, « the Editors can be permitted to edit by
way of an open proxy with the IP block exempt flag. This is granted on
local projects by administrators and globally by stewards. »
I repeat -----> ... legitimate users... may freely use proxies until
those are blocked. the Editors can be permitted to edit by way of an
open proxy with the IP block exempt flag <------ it is not illegal to
edit using an open proxy
Most editors though... have no idea whatsoever what an open proxy is.
They do not understand well what to do when they are blocked.
In the past few weeks, the number of African editors reporting being
blocked due to open proxy has been VERY significantly increasing.
New editors just as old timers.
Unexperienced editors but also staff members, president of usergroups,
organizers of edit-a-thons and various wikimedia initiatives.
At home, but also during events organized with usergroup members or
trainees, during edit-a-thons, photo uploads sessions etc.
It is NOT the occasional highly unlikely situation. This has become a
regular occurence.
There are cases and complains every week. Not one complaint per week.
Several complaints per week.
*This is irritating. This is offending. This is stressful. This is
disrupting activities organized in _good faith_ by _good people_,
activities set-up with _our donors funds. _**And the disruption**is
primarlly taking place in a geographical region supposingly to be
nurtured (per our strategy for diversity, equity, inclusion blahblahblah). *
The open proxy policy page suggests that, should a person be unfairly
blocked, it is recommended
* * to privately email stewards(_AT_)wikimedia.org.
* * or alternatively, to post arequest (if able to edit, if the editor
doesn't mind sharing their IP for global blocks or their reasons to
desire privacy (for Tor usage)).
* * the current message displayed to the blocked editor also suggest
contacting User:Tks4Fish. This editor is involved in vandalism
fighting and is probably the user blocking open proxies IPs the
most. See log
So...
Option 1: contacting stewards : it seems that they are not answering. Or
not quickly. Or requesting lengthy justifications before adding people
to IP block exemption list.
Option 2: posting a request for unblock on meta. For those who want to
look at the process, I suggest looking at it [3] and think hard about
how a new editor would feel. This is simply incredibly complicated
Option 3 : user:TksFish answers... sometimes...
As a consequence, most editors concerned with those global blocks...
stay blocked several days.
We do not know know why the situation has rapidly got worse recently.
But it got worse. And the reports are spilling all over.
We started collecting negative experiences on this page [4].
Please note that people who added their names here are not random
newbies. They are known and respected members of our community, often
leaders of activities and/or representant of their usergroups, who are
confronted to this situation on a REGULAR basis.
I do not know how this can be fixed. Should we slow down open proxy
blocking ? Should we add a mecanism and process for an easier and
quicker IP block exemption process post-blocking ? Should we improve a
process for our editors to pre-emptively be added to this IP block
exemption list ? Or what ? I do not know what's the strategy to fix
that. But there is a problem. Who should that problem be addressed to ?
Who has solutions ?
Flo
[1] https://meta.wikimedia.org/wiki/No_open_proxies
[2] https://meta.wikimedia.org/wiki/Special:Log/Tks4Fish
[3]
https://meta.wikimedia.org/wiki/Steward_requests/Global_permissions#Request…
*[4] https://meta.wikimedia.org/wiki/No_open_proxies/Unfair_blocking*
Hi Folks,
If your language on this list still shows 0% or less than 100% translated,
we appreciate if you take time to translate to your language urgently as
central notice will only run until end of *March 28*. Just click on your
language and edit accordingly. It will be short language translations only.
https://meta.wikimedia.org/wiki/Special:MessageGroupStats?group=Centralnoti…
Kind regards,
Butch Bustria
Hello everyone,
[A short version of this message *translated *into additional languages is
available on Meta-wiki.
<https://meta.wikimedia.org/wiki/Special:MyLanguage/Universal_Code_of_Conduc…>
]
As some of you may know, the Wikimedia Foundation Board of Trustees met
earlier this month in New York. A full report is coming up shortly, but in
the meantime we wanted to share that the Board has resolved to ratify the
Universal Code of Conduct Enforcement Guidelines (EG)
<https://foundation.wikimedia.org/wiki/Resolution:Approval_of_Universal_Code…>
.
To recall, last year on March 2022, the community votes on the EG showed a
slim majority agreeing that the EG was ready to be implemented, with 58.6%
voting in support of the EG. The comments added to the votes revealed
concerns regarding four areas, and the Board requested Foundation staff to
address these concerns.
After the work done by the UCoC Revisions committee and the ongoing input
from the community facilitated by the support of Wikimedia Foundation
staff, confidence in the EG increased, with 76% of voters supporting the
EG in January 2023. Vote statistics are available on Meta-wiki
<https://meta.wikimedia.org/wiki/Universal_Code_of_Conduct/Revised_enforceme…>.
This increase in confidence is remarkable and reflects the thoughtfulness
and care of the global Wikimedia community on this important topic. The
Wikimedia Foundation Board of Trustees has voted to ratify the Enforcement
Guidelines. You can read more about the process behind the 2023 Enforcement
Guidelines review on Diff
<https://diff.wikimedia.org/2023/03/21/the-enforcement-guidelines-received-s…>
.
We are now one step closer to having a fully functioning Universal Code of
Conduct, but the journey will never be truly over. The Universal Code of
Conduct and its Enforcement Guidelines will need regular community review
and improvement as our movement grows and as the world changes around us. As
called upon in our collective Movement Strategy, we will need to iterate
and adapt as we learn, and we will evolve our best practices for how we
work in our shared spaces. For now, there are some next steps to take with
the important recommendations provided by the Enforcement Guidelines. More
details will come in the next few weeks about timelines.
Sending a special {{Thank you}} and {{WikiLove}}) to all the volunteers and
staff members who joined forces to make this a reality, and hope that you
join us in recognizing this significant milestone in creating a safer, more
welcoming space, for all participants and contributors to our projects .
Best,
Shani, on behalf of the Wikimedia Foundation Board of Trustees.
Shani Evenstein Sigalov
<https://wikimediafoundation.org/profile/shani-evenstein-sigalov/>
Vice Chair, Board of Trustees
Chair, Community Affair Committee
Wikimedia Foundation <https://wikimediafoundation.org/>
You can find this message translated into additional languages on Meta-wiki.
<https://meta.wikimedia.org/wiki/Special:MyLanguage/Wikimedia_Foundation_Leg…>
More languages
<https://meta.wikimedia.org/wiki/Special:MyLanguage/Wikimedia_Foundation_Leg…>
• Please help translate to your language
<https://meta.wikimedia.org/w/index.php?title=Special:Translate&group=page-W…>
Hi everyone! The Ombuds commission (OC) and the Case Review Committee (CRC)
are looking for members. People are encouraged to nominate themselves or
encourage others they feel would contribute to these groups to do so. There
is more information below about the opportunity and the skills that are
needed.
About the Ombuds commission
The Ombuds commission (OC) works on all Wikimedia projects to investigate
complaints about violations of the privacy policy, especially in use of
CheckUser <https://meta.wikimedia.org/wiki/CheckUser_policy> and Oversight
<https://meta.wikimedia.org/wiki/Oversight_policy> (also known as
Suppression) tools. The Commission mediates between the parties of the
investigation and, when violations of the policies are identified, advises
the Wikimedia Foundation on best handling. They may also assist the General
Counsel, the Chief Executive Officer, or the Board of Trustees of the
Foundation in these investigations when legally necessary. For more on the
OC's duties and roles, see Ombuds commission on Meta-Wiki
<https://meta.wikimedia.org/wiki/Ombuds_commission>.
Volunteers serving in this role should be experienced Wikimedians, active
on any project, who have previously used the CheckUser/Oversight tools OR
who have the technical ability to understand these tools and the
willingness to learn them. They must be able to communicate in English, the
common language of the commission. They are expected to be able to engage
neutrally in investigating these concerns and to know when to recuse when
other roles and relationships may cause conflict. Commissioners will
serve two-year
terms (note that this is different from past years, when the terms have
been for one year).
About the Case Review Committee
The Case Review Committee (CRC) reviews appeals of eligible Trust & Safety
office actions. The CRC is a critical layer of oversight to ensure that
Wikimedia Foundation office actions are fair and unbiased. They also make
sure the Wikimedia Foundation doesn’t overstep established practices or
boundaries. For more about the role, see Case Review Committee on Meta-Wiki
<https://meta.wikimedia.org/wiki/Case_Review_Committee>.
We are looking for current or former functionaries and experienced
volunteers with an interest in joining this group. Applicants must be
fluent in English (additional languages are a strong plus) and willing to
abide by the terms of the Committee charte
<https://meta.wikimedia.org/wiki/Trust_and_Safety/Case_Review_Committee/Char…>r.
If the work resonates and you qualify, please apply. Committee members will
serve two-year terms (note that this is different from past years, when the
terms have been for one year).
Applying to join either of these groups
Members are required to sign the Confidentiality agreement for nonpublic
information
<https://meta.wikimedia.org/wiki/Confidentiality_agreement_for_nonpublic_inf…>
and must be willing to comply with the appropriate Wikimedia Foundation
board policies (such as the access to non-public information policy
<https://meta.wikimedia.org/wiki/Access_to_nonpublic_information_policy>
and the Foundation privacy policy
<https://foundation.wikimedia.org/wiki/Privacy_policy>). These positions
require a high degree of discretion and trust. Members must also be over 18
years of age.
If you are interested in serving in either capacity listed above, please
write in English to the Trust and Safety team at ca[image: @]wikimedia[image:
.]org (to apply to the OC) or to the Legal Team at legal[image:
@]wikimedia[image:
.]org (to apply to the CRC) with information about:
-
Your primary projects
-
Languages you speak/write
-
Any experience you have serving on committees, whether movement or
non-movement
-
Your thoughts on what you could bring to the OC or CRC if appointed
-
Any experience you have with the Checkuser or Oversight tools (OC only)
-
Any other information you think is relevant
There will be two conversation hours to answer any questions that potential
applicants may have:
-
17 October 2022, 03:00 UTC (other timezones
<https://iw.toolforge.org/zonestamp/1665975648>) (Zoom meeting link)
(add to calendar)
-
16 November 2022, 18:00 UTC (other timezones
<https://iw.toolforge.org/zonestamp/1668621642>) (Zoom meeting link)
(add to calendar)
The deadline for applications is 31 December 2022 in any timezone.
Please feel free to pass this invitation along to any users who you think
may be qualified and interested. Thank you!
On behalf of the Committee Support team,
Karen
--
Karen Brown
Trust & Safety Specialist
Wikimedia Foundation
kbrown(a)wikimedia.org
My Fellow Wikimedians,
There are only 11 days left to submit your program idea
<https://wikimania.wikimedia.org/wiki/2023:Program/Submissions> for
Wikimania 2023. Whether in person in Singapore or online wherever you are,
pre-recorded or live, an interactive workshop, panel discussion, lecture,
lightning talk or an awesome poster, submissions close *Tuesday, March 28.*
Take a look at more than 90 submissions (
https://wikimania.wikimedia.org/wiki/Category:Wikimania_2023_Program_submis…)
already received, perhaps you’ll discover an opportunity to collaborate
with a fellow Wikimedian. If you have any questions, we’d be happy to
answer them at our upcoming conversation hour this *Sunday March 19* at
00:00 <https://iw.toolforge.org/zonestamp/1679184000> and 14:00
<https://iw.toolforge.org/zonestamp/1679234400> UTC. You can also reach out
to us on the help page <https://wikimania.wikimedia.org/wiki/2023:Help_desk>,
our official email address *wikimania(a)wikimedia.org
<wikimania(a)wikimedia.org>* or on Telegram (https://t.me/wikimaniachat) .
All the information you need is available on wiki
<https://wikimania.wikimedia.org/wiki/2023:Program/FAQ>. Making a
submission is easy and we encourage each of you to create one … if you
haven’t already.
The theme for this year's Wikimania is *Diversity, Collaboration, Future*,
and there are 11 tracks
<https://wikimania.wikimedia.org/wiki/2023:Program/Submissions#Tracks> to
choose from: familiar ones like Community Initiatives, Governance, GLAM, or
Technology; and new ones like Open Data and Wild Ideas.
Looking forward to reading yours. To submit a program proposal, go to *this
link.* <https://wikimania.wikimedia.org/wiki/2023:Program/Submissions>
--
Kind regards,
*Butch Bustria*
On behalf of the ESEAP Wikimania 2023 Core Organizing Team and the
Programming Subcommittee
-----------------------------------------------------------------------------------------------------------------------------
The information contained in this message is privileged and intended only
for the recipients named. If the reader is not a representative of the
intended recipient, any review, dissemination or copying of this message or
the information it contains is prohibited. If you have received this
message in error, please immediately notify the sender, and delete the
original message and attachments.
Please consider the environment before printing this email.
Read in عربي, bahasa Indonesia, français, español, Kiswahili, and português on
Diff <https://diff.wikimedia.org/2023/03/17/celebrating-penny-richards/>.
[1]
Wikimedia is a living, digital repository of information, media, and
knowledge, and we wouldn’t be here without the contributions of thousands
of volunteers, countless Wikimedians, like you, sharing their sum of human
knowledge for more than twenty years.
We are excited to celebrate one particularly exceptional volunteer this
month, Penny Richards. Finding joy in what she contributes to Wikimedia and
how she accomplishes that is fundamental for her, it’s what keeps her
going. Visit WikiCelebrate
<https://meta.wikimedia.org/wiki/Communications/WikiCelebrate> to learn
more about Penny and congratulate her. [2]
Penny has a passion for learning about people and leafing through old
newspapers, yearbooks and journals. No surprise that she contributes a lot
of biographies, especially to her favourite project WikiProject Women in
Red. She loves that she can make connections with folks around the world
and join efforts with a large global project like Wikipedia all the while
from home and in her pajamas.
Each month we will celebrate a different Wikimedian, acknowledging the
great people that have contributed so much to bringing us to where we are
today, and continue to do so. We warmly invite you to write about the
people celebrated each month. If you know them, share some wiki love. If
there’s an outstanding Wikimedian that you think should be celebrated,
recommend
them <https://wikimediafoundation.limesurvey.net/WikiCelebrate>. [3]
Happy celebrating,
Natalia and Mehrdad
[1] Diff post:
https://diff.wikimedia.org/2023/03/17/celebrating-penny-richards/
[2] Introducing WikiCelebrate:
https://meta.wikimedia.org/wiki/Communications/WikiCelebrate
[3] Nomination box: https://wikimediafoundation.limesurvey.net/WikiCelebrate
--
*Natalia Szafran-Kozakowska* (she/her)
Senior Global Movement Communications Specialist (European Region)
Wikimedia Foundation <https://wikimediafoundation.org/>