AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018 The Partnership on AI Jeffrey Heer (University of Washington; [email protected]) DOI: 10.1145/3284751.3284760 The Partnership on AI to Benefit People and The PAI kick-off meeting was held in Berlin Society (or PAI) is a consortium of industrial, in October 2017, with wide-ranging discus- non-profit, and academic partners for promot- sions organized around the above pillars. A ing public understanding and beneficial use primary of outcome of the meeting and sub- of Artificial Intelligence (AI) technologies. Pri- sequent membership surveys was the forma- mary goals of the PAI are “to study and for- tion of three initial working groups focused on mulate best practices on AI technologies, to “Safety-Critical AI”, “Fair, Transparent, and Ac- advance the public’s understanding of AI, and countable AI”, and “AI, Labor, and the Econ- to serve as an open platform for discussion omy”. Each of these groups are now conduct- and engagement about AI and its influences ing a series of meetings to develop best prac- on people and society.” tices and resources for their respective topics. PAI membership includes representa- I am a member of the Fair, Transparent, and tives from over 70 partner organizations Accountable AI (or FTA) working group, co- (https://www.partnershiponai.org/ chaired by Ed Felten (Princeton) and Verity partners/), spanning nine countries (pri- Harding (DeepMind). The charter of the FTA marily in North America, Europe, and Asia) working group is to study issues and oppor- and including technology companies of tunities for promoting fair, accountable, trans- various sizes, academic institutions, and a parent, and explainable AI, including the de- number of non-profits (including ACM). The velopment of standards and best practices as Board of Directors represents a number a way to avoid and minimize the risk that AI of large tech companies (Amazon, Apple, systems will undermine fairness, equality, due Facebook, IBM, Google/DeepMind, Microsoft) process, civil liberties, or human rights. The as well as non-profits (ACLU, MacArthur FTA group has held two in-person meetings Foundation, OpenAI) and academics (Arizona so far: the first in London in May 2018 and State University, Harvard, UC Berkeley). The the second at Princeton in August 2018. The founding co-chairs are Eric Horvitz (Microsoft) initial meeting resulted in the formation of four and Mustafa Suleyman (DeepMind), with specific projects. day-to-day operations of the PAI run by a dedicated staff in San Francisco, CA. The “Papers” project aims to produce a set of three educational primers, one each for fair- The PAI’s research and outreach efforts center ness, transparency, and accountability. The around six thematic pillars: goal is to provide a starting point for people to learn about fairness in AI, how it is conceived, • Safety-Critical AI and what people are currently doing about it. • Fair, Transparent, and Accountable AI The purpose is not just to come up with def- initions, but provide structure for how to think • AI, Labor, and the Economy about problems of fairness, transparency and • Collaborations Between People and AI Sys- accountability in the context of the work that tems Partner organizations are conducting. • Social and Societal Influences of AI The “Case Studies” project focuses of identi- • AI and Social Good fying multiple case studies that illustrate how FTA concerns can be addressed, as well For more details regarding each thematic pil- as develop a framework for the analysis of lar, see https://www.partnershiponai. such case studies. Meanwhile, the “Grand org/about/#our-work. Challenges” project is working to create chal- lenges or endorse existing contests developed Copyright c 2018 by the author(s). around FTA issues in practice. 25 AI MATTERS, VOLUME 4, ISSUE 3 4(3) 2018 The “Diverse Voices” project seeks to under- stand different communities, how their voices can be involved in tech policy making, and the impact of AI deployment. The project in- tends to engage with these groups in a mean- ingful manner and ensure under-represented groups’ voices are heard. One goal is to pro- vide guidance for outreach, as many compa- nies are interested in deploying ethical AI but are unsure how to address the problem. Each of these projects is a work-in-progress, with additional milestones and meetings planned over the coming year. Jeffrey Heer is a Pro- fessor of Computer Sci- ence & Engineering at the University of Washington and Co-Founder of Tri- facta Inc., a provider of interactive tools for scal- able data transformation. His research spans human-computer interac- tion, visualization, data science, and interac- tive machine learning. He also serves as an ACM representative to the Partnership on AI. 26.