Wiki AI in 2026 (expectant)
Samuel Klein via Wikimedia-l <[email protected]> Sat, 18 Jul 2026 21:25:55 +0200
| Newsgroups | gmane.org.wikimedia.foundation |
|---|---|
| Message-ID | <CAAtU9WLT2SJ6EaG3eTX=Wp6i1RP1D=72-ZGc57BAQ8=hH__bwQ@mail.gmail.com> |
--===============0000876640466671668==
Content-Type: multipart/alternative; boundary="0000000000009006b00656e7a6fc"
--0000000000009006b00656e7a6fc
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
[TL;DR: lovely things exist; we can build more; keen to see some of you
next week.]
While preparing for the Wikimania preconference on Wiki AI
<https://meta.wikimedia.org/wiki/Artificial_intelligence/2026_Wiki_AI> (ht
to Alex Ostrovsky; come join us!) it struck me how completely we've stopped
talking about AI's steady evolution as practical knowledge infrastructure.
Recent discussions have focused on AI as a challenge for the open web, as
norms around knowledge-seeking and scraping change. But we have hardly
taken any time at all to savor and apply improvements that could advance
our mission, since 2023 (the last time we updated our sweet in-house machin=
e
learning model cards
<https://meta.wikimedia.org/wiki/Machine_learning_models>).
Let us return to building
<https://www.youtube.com/watch?v=3D_hk6KLD-0tg&t=3D309s> wiki-quality AI,
accessible, transparent, and accountable; neutral, proportionate, and
impactful.
We need a family of wiki-principled language models
<https://diff.wikimedia.org/2026/07/11/wikimedia-principled-llms/> (to
borrow Jan Ainali's phrase): with curated open data, open source code, and
open weights. Available to readers and contributors, hosted in our own
green data centers, and integrated into Wikimedia projects (as ORES was),
so details of model use can be easily linked from relevant edit summaries.*
We can get much of what we want with better data sourcing, translation, and
post-training. These are some of the more accessible and inexpensive parts
of the model-building stack. And we are not alone in wanting this: each of
the last <https://en.wikipedia.org/wiki/AI_Action_Summit_2025> two
<https://en.wikipedia.org/wiki/India_AI_Impact_Summit_2026> AI Action
Summits has centered the need to build AI in the public interest, including
advancing human knowledge and protecting & expanding the commons.
In that spirit, I invite you all to channel some of the energy around AI
debates and forecasts towards a few evergreen goals:
1) Write about your vision for the future you want. We need better stories
about the future. Publish it somewhere and link to it from
m:AI#Talks_and_Readings
<https://meta.wikimedia.org/wiki/Artificial_intelligence#Talks_and_Readings=
>.
Submit it to a <https://futurevisionxprize.com/> contest
<https://protopianprize.com/>. I linked Jan's blog post above as an
example... describe things you want even if you don't yet know how to build
them: a benchmark, a tool, a coalition, a campus. Our communities do well
with the charismatic megafauna of collective visions: we are willing to try
things that don't scale, and are always prepared
<https://en.wikiquote.org/wiki/Umberto_Eco> to rewrite our environments.
2) Consider that someone knocking on your library window a billion times a
week asking for an article or an entire back-catalog is not a burden, it's
a relationship. We have a responsibility toward AI systems that rely on
Wikipedia, and also an opportunity to shape how they learn from and
interact with the commons.
In an ideal world (where CDN issues are solved), how should we best serve
agents and other AI tools who turn up looking for grounding or guidance or
bulk materials? How should we influence the default skills and settings of
agents to advance the commons?
3) There are global non-profit networks working towards transparent green
public AI
<https://en.wikipedia.org/wiki/Colorless_green_ideas_sleep_furiously>,
including Humanity AI, Current AI, and Public AI.** Let's find ways to
inform and benefit from aligned efforts, and present a unified front in
proposing changes to current systems and designing new ones.***
Wikimedians should define the components we need within our own
ecosystem, from benchmarks and datasets to curation norms and tools. Then
we can explore post-training processes for neutral, proportional,
epistemically humble langauge models.
All watched over by wikis of loving grace,
SJ
* Even if, on some projects, model use is one step removed from mainspace
edits (as with the AIlogbot
<https://en.wikipedia.org/wiki/User:Fermiboson/AIlog>).
** These include the collaborative *AIPotluck
<https://www.aipotluck.org/> *project,
an effort to build a fully transparent AI stack (whose current biggest gap
<https://www.aipotluck.org/map> is open data); and the *Public AI network
<http://publicai.network>*, a coalition of model builders and advocates
that DWeb
<https://blog.archive.org/2022/02/15/the-decentralized-web-an-introduction/=
>
friends
and I started a few years ago, working towards public options at every
layer of the stack.
*** I wrote about things we might design first, for the Next25 discussion
Christophe organized (on meta
<https://meta.wikimedia.org/wiki/User:Sj/Design_chats/AI>). Briefly:
[*] We should give editors access to the most useful
<https://www.tomshardware.com/software/linux/linus-torvalds-rebukes-anti-ai=
-stances-in-the-linux-kernel-code-review-process-says-linux-is-not-one-of-t=
hose-anti-ai-projects-creator-embraces-ai-as-just-a-tool-and-clearly-a-usef=
ul-one>
open tools, including AI: including for classification, language support,
and citation checking,
[=E2=81=91] We should improve existing public models to produce a family =
of wiki
AI models that serve our characteristic needs
[=E2=81=82] We need shared playgrounds for experimenting with AI systems =
=F0=9F=A7=B8
[=E2=9C=A3] We need a bigger wiki to experiment with new forms of automat=
ion...
including via AI and agentic systems. This requires new tools for
high-volume review and is not suitable for testing on production wikis.
[=EF=BC=8A] As bandwidth becomes an issue again: We should integrate bett=
er with
repositories like Zenodo, Github, =F0=9F=A4=97, and Openverse. Then we don'=
t have to
replicate their storage infrastructure just to have convenient and editable
metadata.
=3D=3D R=C3=A9f=C3=A9rences =3D=3D
0. Wikimania preconference day on Wiki AI (still time to sign up today):
https://meta.wikimedia.org/wiki/Artificial_intelligence/2026_Wiki_AI
1. Machine learning model cards:
https://meta.wikimedia.org/wiki/Machine_learning_models
2. Building Language Models for Wikimedia (Bob West, 2024) -
https://www.youtube.com/watch?v=3D_hk6KLD-0tg&t=3D309s
3. Wikimedia-principled LLMs -
https://diff.wikimedia.org/2026/07/11/wikimedia-principled-llms/
4. AI Action Summit 2025 -
https://en.wikipedia.org/wiki/AI_Action_Summit_2025
5. AI Impact Summit 2026 -
https://en.wikipedia.org/wiki/India_AI_Impact_Summit_2026
6. AI Talks and Readings (meta) -
https://meta.wikimedia.org/wiki/Artificial_intelligence#Talks_and_Readings
7a. https://futurevisionxprize.com/
7b. Protopian Fiction Prize for Public AI - https://protopianprize.com/
8. "the cultivated person's first duty is to be always prepared to rewrite
the encyclopaedia" =E2=80=93 Eco https://en.wikiquote.org/wiki/Umberto_Eco
9. "transparent green ideas sleep furiously" (Bohr) -
https://en.wikipedia.org/wiki/Colorless_green_ideas_sleep_furiously
10. AI Potluck (Current AI foundation) - https://www.aipotluck.org/
11. A gap map for open source AI - https://www.aipotluck.org/map
12. http://publicai.network/
13.
https://blog.archive.org/2022/02/15/the-decentralized-web-an-introduction/
14. Building Wiki AI -
https://meta.wikimedia.org/wiki/User:Sj/Design_chats/AI
15. https://www.phoronix.com/news/Linux-Is-Not-Anti-AI
16.
https://en.wikipedia.org/wiki/All_Watched_Over_by_Machines_of_Loving_Grace_=
(TV_series)
--=20
Samuel Klein @metasj w:user:sj
--0000000000009006b00656e7a6fc
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
<div dir=3D"ltr"><div><span style=3D"background-color:transparent">[TL;DR: =
lovely things exist; we can build more; keen to see some of you next week.]=
<br><br>While preparing for the Wikimania preconference on=C2=A0</span><a h=
ref=3D"https://meta.wikimedia.org/wiki/Artificial_intelligence/2026_Wiki_AI=
" style=3D"background-color:transparent">Wiki AI</a><span style=3D"backgrou=
nd-color:transparent">=C2=A0(ht to Alex Ostrovsky; come join us!)=C2=A0it s=
truck me how completely we've stopped talking about AI's steady evo=
lution as practical knowledge infrastructure.</span></div><div><br>Recent d=
iscussions have focused on AI as a challenge for the open web, as norms aro=
und knowledge-seeking and scraping change. But we have hardly taken any=C2=
=A0 time at all to savor and apply improvements that could advance our miss=
ion, since 2023=C2=A0<span style=3D"background-color:transparent">(the last=
time we updated our sweet in-house=C2=A0</span><a href=3D"https://meta.wik=
imedia.org/wiki/Machine_learning_models" style=3D"background-color:transpar=
ent">machine learning model cards</a><span style=3D"background-color:transp=
arent">).</span></div><div><span style=3D"background-color:transparent"><br=
>Let us return to <a href=3D"https://www.youtube.com/watch?v=3D_hk6KLD-0tg&=
amp;t=3D309s">building</a>=C2=A0wiki-quality AI,=C2=A0<br></span><span styl=
e=3D"background-color:transparent">accessible, transparent, and=C2=A0</span=
>accountable;=C2=A0<span style=3D"background-color:transparent">neutral, pr=
oportionate, and impactful.=C2=A0</span></div><div><span style=3D"backgroun=
d-color:transparent"><br></span></div><div><span style=3D"background-color:=
transparent">W</span><span style=3D"background-color:transparent">e need a =
family of=C2=A0</span><a href=3D"https://diff.wikimedia.org/2026/07/11/wiki=
media-principled-llms/" style=3D"background-color:transparent">wiki-princip=
led language models</a><span style=3D"background-color:transparent">=C2=A0(=
to borrow Jan Ainali's phrase)</span><span style=3D"background-color:tr=
ansparent">: with curated open data, open source code, and open weights. Av=
ailable to readers and contributors,=C2=A0</span><span style=3D"background-=
color:transparent">h</span>osted in our own green data centers, and i<span =
style=3D"background-color:transparent">ntegrated into Wikimedia projects (a=
s ORES was), so details of model use can be easily linked from relevant edi=
t summaries.*=C2=A0</span></div><div><span style=3D"background-color:transp=
arent">=C2=A0</span></div><div>We can get much of what we want with better =
data sourcing, translation, and post-training. These are some of the more a=
ccessible and inexpensive parts of the model-building stack. And we are not=
alone in wanting this: each of the <a href=3D"https://en.wikipedia.org/wik=
i/AI_Action_Summit_2025">last</a> <a href=3D"https://en.wikipedia.org/wiki/=
India_AI_Impact_Summit_2026">two</a> AI Action Summits=C2=A0has centered th=
e need to build AI in the public interest, including advancing human knowle=
dge and protecting & expanding the commons.=C2=A0 =C2=A0</div><div><br>=
In that spirit, I invite you all to channel some of the energy around AI de=
bates and forecasts towards a few evergreen goals:=C2=A0</div><div><br></di=
v><div>1) Write about your vision for the future you want.=C2=A0 We need be=
tter stories about the future. Publish it somewhere and link to it from=C2=
=A0<a href=3D"https://meta.wikimedia.org/wiki/Artificial_intelligence#Talks=
_and_Readings">m:AI#Talks_and_Readings</a>.=C2=A0 Submit it to <a href=3D"h=
ttps://futurevisionxprize.com/">a</a> <a href=3D"https://protopianprize.com=
/">contest</a>.=C2=A0 I linked Jan's blog post above as an example... d=
escribe things you want even if you don't yet know how to build them: a=
benchmark, a tool, a coalition, a campus.=C2=A0 Our communities do well wi=
th the charismatic megafauna of collective visions: we are willing to try t=
hings that don't scale, and are=C2=A0<a href=3D"https://en.wikiquote.or=
g/wiki/Umberto_Eco">always prepared</a> to rewrite our environments.</div><=
div><br></div><div>2) Consider that someone knocking on your library window=
a billion times a week asking for an article or an entire back-catalog is =
not a burden, it's a relationship. We have a responsibility toward AI s=
ystems that rely on Wikipedia, and also an opportunity to shape how they le=
arn from and interact with the commons.=C2=A0 <br>In an ideal world (where =
CDN issues are solved), how should we best serve agents and other AI tools =
who turn up looking for grounding or guidance or bulk materials? How should=
we influence the default skills and settings of agents to advance the comm=
ons?<br></div><div><br></div><div>3) There are global non-profit networks w=
orking towards <a href=3D"https://en.wikipedia.org/wiki/Colorless_green_ide=
as_sleep_furiously">transparent green public AI</a>, including Humanity AI,=
Current AI, and Public AI.**=C2=A0 Let's find ways to inform and benef=
it from aligned efforts, and present a unified front in proposing changes t=
o current systems and designing new ones.***=C2=A0<br><span style=3D"backgr=
ound-color:transparent">=C2=A0 =C2=A0 Wikimedians should</span><span style=
=3D"background-color:transparent">=C2=A0define the components we need withi=
n our own ecosystem, from benchmarks and datasets to curation norms and too=
ls. Then we can explore post-training processes for neutral, proportional, =
epistemically humble langauge models.=C2=A0=C2=A0</span></div><div><span st=
yle=3D"background-color:transparent"><br>All watched over by wikis of lovin=
g grace,</span></div><div><span style=3D"background-color:transparent">SJ</=
span></div><div><span style=3D"background-color:transparent"><br>* Even if,=
on some projects, model use is one step removed from mainspace edits (as w=
ith the <a href=3D"https://en.wikipedia.org/wiki/User:Fermiboson/AIlog">AIl=
ogbot</a>).=C2=A0<br><br>**=C2=A0</span><span style=3D"background-color:tra=
nsparent">These include=C2=A0</span><span style=3D"background-color:transpa=
rent">the collaborative=C2=A0<i><a href=3D"https://www.aipotluck.org/">AIPo=
tluck</a> </i>project, an effort to build a fully transparent AI stack (who=
se current=C2=A0<a href=3D"https://www.aipotluck.org/map">biggest gap</a>=
=C2=A0is open data); and=C2=A0</span><span style=3D"background-color:transp=
arent">the=C2=A0</span><i style=3D"background-color:transparent"><a href=3D=
"http://publicai.network">Public AI network</a></i><span style=3D"backgroun=
d-color:transparent">, a coalition of model builders and advocates that=C2=
=A0<a href=3D"https://blog.archive.org/2022/02/15/the-decentralized-web-an-=
introduction/">DWeb</a>=C2=A0friends and I started a few years ago, working=
towards public options at every layer of the stack.</span><span style=3D"b=
ackground-color:transparent">=C2=A0</span></div><div><span style=3D"backgro=
und-color:transparent"><br></span></div><div><span style=3D"background-colo=
r:transparent">*** I wrote about things we might design</span><span style=
=3D"background-color:transparent">=C2=A0first, for the Next25 discussion Ch=
ristophe organized (</span><a href=3D"https://meta.wikimedia.org/wiki/User:=
Sj/Design_chats/AI" style=3D"background-color:transparent">on meta</a><span=
style=3D"background-color:transparent">).=C2=A0=C2=A0</span><span style=3D=
"background-color:transparent">Briefly</span><span style=3D"background-colo=
r:transparent">:</span></div><div><span style=3D"background-color:transpare=
nt"><br></span>=C2=A0[*]=C2=A0 We should give editors access to the most <a=
href=3D"https://www.tomshardware.com/software/linux/linus-torvalds-rebukes=
-anti-ai-stances-in-the-linux-kernel-code-review-process-says-linux-is-not-=
one-of-those-anti-ai-projects-creator-embraces-ai-as-just-a-tool-and-clearl=
y-a-useful-one">useful</a> open tools, including AI: including for classifi=
cation, language support, and=C2=A0<span style=3D"background-color:transpar=
ent">citation checking,=C2=A0</span><div>=C2=A0[=E2=81=91]=C2=A0 We should =
improve existing public models to produce a family of=C2=A0wiki AI models=
=C2=A0that serve our characteristic needs</div><div class=3D"gmail_signatur=
e">=C2=A0[=E2=81=82]=C2=A0=C2=A0We need shared=C2=A0playgrounds=C2=A0for ex=
perimenting with AI systems=C2=A0=F0=9F=A7=B8<br>=C2=A0[=E2=9C=A3]=C2=A0 We=
need a bigger wiki to experiment with new forms of automation... including=
via AI and agentic systems. This requires new tools for high-volume review=
and is not suitable for testing on production wikis.=C2=A0</div><div class=
=3D"gmail_signature">=C2=A0[=EF=BC=8A]=C2=A0 As bandwidth becomes an issue =
again: We should integrate better with repositories like Zenodo, Github, =
=F0=9F=A4=97, and Openverse. Then we don't have to replicate their stor=
age infrastructure just to have convenient and editable metadata.=C2=A0<br>=
<br><br>=3D=3D R=C3=A9f=C3=A9rences =3D=3D<br>0. Wikimania preconference da=
y on Wiki AI (still time to sign up today):=C2=A0 <a href=3D"https://meta.w=
ikimedia.org/wiki/Artificial_intelligence/2026_Wiki_AI">https://meta.wikime=
dia.org/wiki/Artificial_intelligence/2026_Wiki_AI</a>=C2=A0<br>1.=C2=A0Mach=
ine learning model cards:=C2=A0<a href=3D"https://meta.wikimedia.org/wiki/M=
achine_learning_models">https://meta.wikimedia.org/wiki/Machine_learning_mo=
dels</a></div></div><div class=3D"gmail_signature">2.=C2=A0Building Languag=
e Models for Wikimedia=C2=A0(Bob West, 2024) -=C2=A0<a href=3D"https://www.=
youtube.com/watch?v=3D_hk6KLD-0tg&t=3D309s">https://www.youtube.com/wat=
ch?v=3D_hk6KLD-0tg&t=3D309s</a></div><div class=3D"gmail_signature">3.=
=C2=A0Wikimedia-principled LLMs<span style=3D"background-color:transparent"=
>=C2=A0-=C2=A0</span><a href=3D"https://diff.wikimedia.org/2026/07/11/wikim=
edia-principled-llms/">https://diff.wikimedia.org/2026/07/11/wikimedia-prin=
cipled-llms/</a><span style=3D"background-color:transparent"><br>4.=C2=A0</=
span>AI Action Summit 2025<span style=3D"background-color:transparent">=C2=
=A0-=C2=A0</span><a href=3D"https://en.wikipedia.org/wiki/AI_Action_Summit_=
2025">https://en.wikipedia.org/wiki/AI_Action_Summit_2025</a><span style=3D=
"background-color:transparent"><br></span>5.=C2=A0AI Impact Summit 2026=C2=
=A0-=C2=A0<a href=3D"https://en.wikipedia.org/wiki/India_AI_Impact_Summit_2=
026">https://en.wikipedia.org/wiki/India_AI_Impact_Summit_2026</a><br>6.=C2=
=A0AI Talks and Readings (meta) -=C2=A0<a href=3D"https://meta.wikimedia.or=
g/wiki/Artificial_intelligence#Talks_and_Readings">https://meta.wikimedia.o=
rg/wiki/Artificial_intelligence#Talks_and_Readings</a><br>7a. <a href=3D"ht=
tps://futurevisionxprize.com/">https://futurevisionxprize.com/</a></div><di=
v class=3D"gmail_signature">7b. Protopian Fiction Prize for Public AI -=C2=
=A0<a href=3D"https://protopianprize.com/">https://protopianprize.com/</a><=
span style=3D"background-color:transparent"><br>8. "</span>the cultiva=
ted person's first duty is to be always prepared to rewrite the encyclo=
paedia" =E2=80=93 Eco=C2=A0=C2=A0<a href=3D"https://en.wikiquote.org/w=
iki/Umberto_Eco">https://en.wikiquote.org/wiki/Umberto_Eco</a><span style=
=3D"background-color:transparent"><br>9. "</span>transparent green=C2=
=A0ideas sleep furiously" (Bohr) -=C2=A0<a href=3D"https://en.wikipedi=
a.org/wiki/Colorless_green_ideas_sleep_furiously">https://en.wikipedia.org/=
wiki/Colorless_green_ideas_sleep_furiously</a><span style=3D"background-col=
or:transparent"><br>10.=C2=A0</span><span style=3D"background-color:transpa=
rent">AI Potluck (Current AI=C2=A0foundation) -=C2=A0</span><span style=3D"=
background-color:transparent"><a href=3D"https://www.aipotluck.org/">https:=
//www.aipotluck.org/</a></span></div><div class=3D"gmail_signature"><span s=
tyle=3D"background-color:transparent">11.=C2=A0A gap map for open source AI=
=C2=A0-=C2=A0</span><span style=3D"background-color:transparent"><a href=3D=
"https://www.aipotluck.org/map">https://www.aipotluck.org/map</a></span></d=
iv><div class=3D"gmail_signature"><span style=3D"background-color:transpare=
nt">12.=C2=A0</span><span style=3D"background-color:transparent"><a href=3D=
"http://publicai.network/">http://publicai.network/</a></span></div><div cl=
ass=3D"gmail_signature"><span style=3D"background-color:transparent">13.=C2=
=A0</span><a href=3D"https://blog.archive.org/2022/02/15/the-decentralized-=
web-an-introduction/">https://blog.archive.org/2022/02/15/the-decentralized=
-web-an-introduction/</a><span style=3D"background-color:transparent"><br>1=
4.=C2=A0</span>Building Wiki AI=C2=A0-=C2=A0<a href=3D"https://meta.wikimed=
ia.org/wiki/User:Sj/Design_chats/AI">https://meta.wikimedia.org/wiki/User:S=
j/Design_chats/AI</a></div><div class=3D"gmail_signature">15. <a href=3D"ht=
tps://www.phoronix.com/news/Linux-Is-Not-Anti-AI">https://www.phoronix.com/=
news/Linux-Is-Not-Anti-AI</a>=C2=A0</div><div class=3D"gmail_signature">16.=
=C2=A0<a href=3D"https://en.wikipedia.org/wiki/All_Watched_Over_by_Machines=
_of_Loving_Grace_(TV_series)">https://en.wikipedia.org/wiki/All_Watched_Ove=
r_by_Machines_of_Loving_Grace_(TV_series)</a><br><br></div><span class=3D"g=
mail_signature_prefix">-- </span><br><div dir=3D"ltr" class=3D"gmail_signat=
ure" data-smartmail=3D"gmail_signature">Samuel Klein =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0@metasj =C2=A0=C2=A0 =C2=A0 =C2=A0 =C2=A0=C2=A0 w:user:sj=C2=
=A0 =C2=A0 =C2=A0 =C2=A0<br></div></div>
--0000000000009006b00656e7a6fc--
--===============0000876640466671668==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
_______________________________________________
Wikimedia-l mailing list -- [email protected], guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/[email protected]/message/W6RUCFY4RBIVHF73OEIVGS223TPTHAMX/
To unsubscribe send an email to wikimedia-l-leave-RusutVdil2icGmH+5r0DM0B+6BGkLq7r@public.gmane.org
--===============0000876640466671668==--