Annual Report 2021
By bringing together diverse voices from across the AI community and around the world, PAI seeks to not just spark new ideas but to turn those ideas into lasting change.
At all points in our process — from convening stakeholders and creating resources to promoting what we’ve learned and practicing accountability — PAI is working to advance responsible governance and best practices so that AI can have the greatest possible benefit for people and society.
PAI believes the best approaches to global challenges are informed by global perspectives. In the spring of 2021, PAI teamed up with human rights non-profit WITNESS to collect insights on one such challenge: the detection of AI-manipulated media. PAI and WITNESS hosted two convenings of journalists, activists, and researchers in South America and Africa, asking them who should have access to synthetic media detection tools and under what conditions.
Informing the Public
Why does the AI field struggle to attract and retain diverse talent? Important answers can be found in the experiences of these professionals themselves, the focus of a three-part series PAI published in the fall of 2021. Drawing from a forthcoming PAI study based on in-depth interviews with present and former tech workers, this series drew attention to an important (but under-examined) barrier to greater diversity in AI.
Encouraging Policy Innovation
Our resources are designed to not only synthesize knowledge but to make that knowledge actionable for policymakers and other stakeholders. In 2021, PAI was pleased to respond to two requests for information from the U.S. government to inform new AI projects. As the U.S. builds its National Artificial Intelligence Research Resource and Artificial Intelligence Risk Management Framework, it will be able draw from PAI’s important work concerning transparency, inclusion, risk documentation, and responsible publication norms.
Fostering Changes in Practice
In 2021, PAI released a number of tools, recommendations, and other resources designed to make ethics in AI actionable. These include the white papers “Managing the Risks of AI Research,” “Responsible Sourcing of Data Enrichment Services,” and “Fairer Algorithmic Decision-Making and Its Consequences,” as well as the ABOUT ML Resource Library and “Redesigning AI for Shared Prosperity: an Agenda.”