Session
Organizer 1: Civil Society, Eastern European Group
Organizer 2: Civil Society, Western European and Others Group (WEOG)
Organizer 2: Civil Society, Western European and Others Group (WEOG)
Speaker 1: Fabro Steibel, Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 2: Lea Gimpel, Private Sector, Western European and Others Group (WEOG)
Speaker 3: Deshni Govender, Intergovernmental Organization, Intergovernmental Organization
Speaker 2: Lea Gimpel, Private Sector, Western European and Others Group (WEOG)
Speaker 3: Deshni Govender, Intergovernmental Organization, Intergovernmental Organization
Format
Roundtable
Duration (minutes): 90
Format description: The roundtable format will allow us to organize an inclusive conversation that will highlight the perspectives of various participants, other than the speakers. We aim for this session to be inclusive, as we would like to show the diversity of approaches to democratization of AI, and the various contexts in which they are replied. The extended duration of the session will let us explore in more depth a topic that is cross-cutting across several policy debates: on compute, data and model governance.
Duration (minutes): 90
Format description: The roundtable format will allow us to organize an inclusive conversation that will highlight the perspectives of various participants, other than the speakers. We aim for this session to be inclusive, as we would like to show the diversity of approaches to democratization of AI, and the various contexts in which they are replied. The extended duration of the session will let us explore in more depth a topic that is cross-cutting across several policy debates: on compute, data and model governance.
Policy Question(s)
1. To what extent can sharing of resources necessary for AI development and deployment, including open-source models and datasets, help develop public AI solutions? What additional measures, beyond ensuring openness, are needed to achieve democratization of AI?
2. What specific governance mechanisms should be introduced to ensure democratization of AI technologies? How do we balance access and openness with control and protection from value extraction?
3. What is the role of private companies, public administrations and international organizations in democratizing AI through public AI programs?
What will participants gain from attending this session? Thanks to the workshop, participants will gain an understanding of the various policies for public AI, aimed at democratizing AI and curbing concentrations of power in AI. The workshop will clarify how public AI policies combine measures related to computing power, data and model development, and include governance mechanisms that help democratize AI. The session will also highlight the global dimension of AI development by presenting viewpoints on democratizing AI from various regions. Finally, specific initiatives that bring these policies to life will be presented.
Description:
In recent years, AI systems became increasingly connected with the internet, which enabled their rapid diffusion. Simultaneously, AI development is characterized by extreme market concentration, with a few companies controlling layers of the AI technology stack, resulting in unchecked power hegemonies. This concentration of power limits innovation, creates environmental challenges and exacerbates global inequalities. Progress in AI capabilities therefore offers potential for solving global problems, but at the same time uneven allocation of AI’s benefits worsens inequalities and translates into social challenges. To combat this trend, sustainable AI innovation and development can be ensured by adopting, at global scale, public AI policies. These include measures that democratize AI governance, level the playing field and make AI technologies more accessible, and ensure that their deployment respects human rights and democratic values. The aim of the workshop will be to explore key elements of proposals for public AI policies, drawing on a range of proposals, including those presented in policy briefs on democratizing AI published last year by both T7 and T20 groups. The workshop will address four pillars of sustainable and responsible AI innovation: * Open-source AI ecosystems: Supporting innovation through open models that enable broader participation in AI development, particularly from underrepresented communities and global majority countries; * Data commons for AI training: Creating AI training datasets that ensure transparency and enable access to data, while protecting rights and ensuring sustainability of the commons; * Public compute capacity: Building infrastructures that provide compute capacity necessary to support research into more sustainable, public-interest AI solutions * Reduce global inequalities: Ensuring that the capacity to deploy and benefit from AI technologies is equally distributed, also through international AI governance. The session will engage diverse stakeholders to define public AI policies and measures that ensure democratization of AI technologies.
In recent years, AI systems became increasingly connected with the internet, which enabled their rapid diffusion. Simultaneously, AI development is characterized by extreme market concentration, with a few companies controlling layers of the AI technology stack, resulting in unchecked power hegemonies. This concentration of power limits innovation, creates environmental challenges and exacerbates global inequalities. Progress in AI capabilities therefore offers potential for solving global problems, but at the same time uneven allocation of AI’s benefits worsens inequalities and translates into social challenges. To combat this trend, sustainable AI innovation and development can be ensured by adopting, at global scale, public AI policies. These include measures that democratize AI governance, level the playing field and make AI technologies more accessible, and ensure that their deployment respects human rights and democratic values. The aim of the workshop will be to explore key elements of proposals for public AI policies, drawing on a range of proposals, including those presented in policy briefs on democratizing AI published last year by both T7 and T20 groups. The workshop will address four pillars of sustainable and responsible AI innovation: * Open-source AI ecosystems: Supporting innovation through open models that enable broader participation in AI development, particularly from underrepresented communities and global majority countries; * Data commons for AI training: Creating AI training datasets that ensure transparency and enable access to data, while protecting rights and ensuring sustainability of the commons; * Public compute capacity: Building infrastructures that provide compute capacity necessary to support research into more sustainable, public-interest AI solutions * Reduce global inequalities: Ensuring that the capacity to deploy and benefit from AI technologies is equally distributed, also through international AI governance. The session will engage diverse stakeholders to define public AI policies and measures that ensure democratization of AI technologies.
Expected Outcomes
The goal of the workshop is to raise awareness about public AI policies as measures that can serve to democratize AI development. In particular, the session will help to:
* build a multistakeholder coalition of actors supporting public AI policies across the world;
* define actionable strategies for governments to secure public AI goals;
* demonstrate steps that enable equitable AI resource sharing between Global Minority and Majority countries;
* demonstrate relevance of AI governance and AI democratization for internet governance, taking into account the importance of emergent AI technologies.
Hybrid Format: * We aim for at least one of the speakers to be remote, to ensure hybrid format of conversation between the speakers.
* The online moderator will manage the online chat, which will serve as "back channel" conversation. The inputs from this channel will be highlighted by the moderator, including questions from remote participants.
* The question and answers part will be intentionally moderated to include, in changing order, input from on-site and online participants.