Conference: ReDDDoT Phase 1: Workshop Towards the promise of open source AI models - A workshop to co-create a vision for responsibility and corresponding research roadmap

Generative Artificial Intelligence (AI) technologies will have profound impacts on various domains from science to the economy, education, the environment and more. The future of generative AI technology is under intense debate and a central topic is related to the openness and open sourcing of foundation models. Open source models offer immense promise towards societal impacts as compared to closed models across key areas of research, innovation, transparency, equity, and more.

In particular, open foundation models promise to ‘democratize’ access to AI by enabling people to examine, reuse and build on top of these powerful systems. On the flip side, bad actors may more efficiently and effectively create and deploy AI building upon open models in ways that are harmful to individuals, communities, and/or societies. While open source models can create greater opportunities to ‘democratize’ AI development, for both open and closed models, key decisions such as data the models learn from, transparency provided, and other considerations are currently undemocratic.

They are informed by the values and market priorities of their largely for-profit driven creators and managers. To truly democratize AI, greater efforts are needed to integrate broader perspectives and voices in the design and development of large open source models.

The goal of the project is to conduct a workshop to integrate broader perspectives and voices in the design and development of large open source models, including through co-creating a vision for ‘responsible’ open source models and a roadmap to move towards this vision. The workshop will explore the following critical questions: (1) What do ‘responsible’ open source models look like and how might we co-create a vision for ‘responsible’ open source models, including prioritizing AI principles? (2) What tradeoffs exist between identified AI principles for ‘responsible’ open source models and how do we navigate and mitigate these tensions? (3) What future research is needed to advance progress towards the co-created vision for ‘responsible’ open source AI?

The workshop will result in a paper on the results of the workshop including the state of open source foundation models, the co-created vision for “responsible” open source foundation models, and research roadmap. It will also result in an accessible online report summarizing the vision and roadmap, and a Slack channel for participants to connect and collaborate.

The workshop and outputs can also help inform future ReDDDoT priorities as it relates to research, innovation, and capacity building for responsible (open source) AI models.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.