2024:Program/Wikimedia and Public AI: a tale of two cultural technologies
Session title: Wikimedia and Public AI: a tale of two cultural technologies
- Session type: Lecture
- Track: Technology
- Language: en
What does the use of Wikimedia projects for AI training imply? How can our projects serve as a model for public AI systems, platforms, and communities? Is AI more than the latest way wiki content is being reused, or does it present a challenge that calls for the exploration of new governance methods for the production, distribution, and usage of the knowledge commons?
This session will address these questions and introduce the Public AI Network, an international coalition dedicated to advancing public-interest and publicly-owned AI around the world.
Description
[edit | edit source]Public AI Whitepaper: https://drive.google.com/file/d/1BPonjBZllV9O5OTLYPQdz3KL1a7kjtWN/view
What does the use of Wikimedia projects for AI training imply? In what ways can our projects serve as a model for public AI systems, platforms, and communities? Is AI more this simply the latest way in which Wikimedia is being reused, in line with the vision of free knowledge established two decades ago? Does AI development present a challenge that necessitates the exploration of new governance methods for the production, distribution, and usage of the knowledge commons?
We will address these questions and introduce the Public AI Network, an international coalition dedicated to advancing public-interest and publicly-owned AI around the world, and drawing on the work the Open Future Foundation has done on the commons-based governance of AI datasets. This work assumes that various collections forming the knowledge commons - especially Wikimedia, are increasingly being conceptualized and treated as AI training datasets. And that this is a new way of thinking about the knowledge commons that may require new frontiers of development, curation and governance from their stewards.
It is AI developers who largely define how the commons is conceptualized and used as AI datasets. And to date, the wiki community has not extended to include development many AI systems of its own.
We need to establish rules for using knowledge and cultural commons for AI training, and to embody them in the tools and systems that we build, train, and use.
To aid this effort, Open Future has recently undertaken two activities:
- Developed a six-principle framework for commons-based dataset governance, which can guide organizations in devising new ways of managing resources that are either purposely made available as training datasets, or are being used as such.
- Organized an online Alignment Assembly on AI and the commons to better comprehend the perspectives of creators, commons stewards, and activists. We utilized the pol.is platform to identify aligned and divisive issues.
Wikimedia projects exemplify new governance mechanisms for the commons. Gaps still exist, and additional mechanisms could enhance the governance and production of new Wikimedia resources, considering the emerging technologies and their associated challenges.
Wikimedia is not solely a source of data and human curation for AI training. It is a unique "cultural technology" that is collectively built, whose processes and participants can inform the development of new cultural technologies such as the ecosystem of AI built on collective knowledge.
Session recording: https://www.youtube.com/watch?v=XOeZ8sHimN0&list=PLhV3K_DS5YfJdC5P86rsDsUtxEow0gDnR&t=4706
- How does your session relate to the event theme, Collaboration of the Open?
AI development, including approaches based on the idea of the commons and open-source development, are collaborative in nature. Understanding how openness plays out in this space, and setting new rules for using free knowledge in AI development, requires development between the various parties involved: those working with content, data, code, and models.
- What is the experience level needed for the audience for your session?
Some experience will be needed
- Etherpad link
https://etherpad.wikimedia.org/p/WM2024_Day4_Ochrid_-_Room_9
Resources
[edit | edit source]Speakers
[edit | edit source]- Alek Tarkowski
- Alek Tarkowski is the Director of Strategy at Open Future. He is also a member of the Board of Directors of Creative Commons. He has 20 years of experience with public interest advocacy and movement building. He is a sociologist by training and holds a Ph.D. in sociology from the Polish Academy of Science. Previously, he was part of the Wikimedia Movement Strategy process as a member of the Partnerships working group.
- Sam Klein
- Ideographer, generist, cryptanalyst. Stable diffusor.