Program of Activities
The AI for Good Foundation addresses three principal issues that exist today, and that are preventing Artificial Intelligence from having its most beneficial impact on society.
- The lack of a common vision in the AI research community, and a fragmented body of scientific work that is difficult to reproduce and build upon;
- The absence of mechanisms and forums of communication between researchers, practitioners, policy-makers, and the public, to educate and engage on the opportunities and threats of emerging capabilities, both from a technology and socio-economic perspective;
- The unbalanced nature of incentives and funding for Artificial Intelligence research, that end up favoring defense and military applications, at the expense of directly socially beneficial projects;
We believe that Artificial Intelligence is misunderstood, that the various stakeholders do not communicate as well as they should, and that the research efforts need more common ground for sharing and benefiting from the good work that is already happening. We also believe that the solutions need to be global, bringing together people from many different backgrounds, identifying the most promising areas for research and implementation, and being the voice of rationality and data-driven conclusions. As such, we have developed a series of complementary programs, engaging the world’s most prominent researchers in the field, and based on a culture of evidence-driven transparency. Our programs are:
- Workshops and Conferences
Host regular workshops and conferences, in partnership with major AI-related conferences, as well as independently in key strategic locations around the world. These events bring together different groups of people from the research, practitioner, and policy communities, along with the general public. We aim to engage these groups to contribute to the research agenda, speak openly about the threats and opportunities, and develop a coherent plan of action through discussions, panels, collaborative projects, presentations, surveys, and other methods as appropriate to furthering the exempt mission of the organization.
- Education Outreach
In collaboration with strategic partners such as videolectures.net (a United Nations Educational, Scientific, and Cultural Organization), among others, we aim for the continuous dissemination of up-to-date studies, statistics, charts, newsletters, and presentations. We try to elucidate the current state of affairs in Artificial Intelligence research and practice, future possibilities, the likely socio-economic impacts of change, and the management of any emerging threats from such capabilities.
We look to harness all relevant social and traditional media platforms, from discussion forums, blogs, social networks, to television, radio, and newspapers. Additionally, we present at appropriate forums nationwide and globally, and host all materials freely on our website and partner websites to encourage access and redistribution.
- Standards and Guidelines
Part of building a shared vision and a culture of transparency, stems from having common principles that can be followed by the community of researchers and practitioners. Any such guidelines need to be crafted with strong oversight from key stakeholder groups, and should outline achievable goals that do not overburden participants.
We aim to propose guidelines for rendering research in Artificial Intelligence more accessible and reproducible. We also aim to propose a code of ethics for the research and practitioner communities that would promote risk-mitigation practices in situations where these would be appropriate. Finally, we may research and offer guidance associated with socio-economic policy direction, where Artificial Intelligence might have an impact.
We believe that these initiatives will promote trust and transparency among all stakeholders.
- Tools and Platforms
One of the biggest difficulties in the scientific research community is building a strong and coherent body of work that can be leveraged, built upon, and trusted. Incentives for academic researchers lie mostly in the publication of novel ideas, and not in the validation of others’ novel ideas. As such, the pressure to publish often overwhelms the need for clarity and robustness. Incentives are misaligned. Many research results cannot and will not be reproduced. The inefficiencies of such incentives go far beyond the research communities, with businesses of all sizes wasting months of R&D effort attempting to leverage promising results, and failing due to under-specified procedures, private data, models that over-fit a particular domain, or that were trained on samples statistically equivalent to their test data.
There is a need for tools, services, and platforms that realign incentives. This will require close collaboration with leading journals and conferences, as well as establishing reputational mechanisms to encourage beneficial behaviors from students and researchers. We have been discussing the blueprints for these systems with various stakeholder groups, and feel ready to begin implementation. These tools and platforms will be managed by the AI for Good Foundation, and access will be free for all.
- Research Program Funding and Support
In order to benefit society to the fullest extent possible, Artificial Intelligence researchers need to be driven to solve problems that are relevant to our everyday social existence. Instead, much funding is driven by military and advertising applications, and as a result, these domains are heavily over-represented.
We intend to build a set of very visible and prestigious research awards that focus on solving some of the biggest issues impacting society in our time. We solicit proposals for each research award, favoring those that have far-reaching consequences, and are scientifically bold, as judged by a panel of top academics, practitioners, and policy-makers. Not only will each research award have a direct impact, but the visibility of these projects will help shift more attention to a class of important problems that are otherwise not so visible.
Furthermore, we may engage in our own internal programs of fundamental research, where we see a strategic opportunity, and where it has proven difficult to solicit such aligned proposals as part of our award program. Such fundamental research will be exclusively in furtherance of our mission.
- Local Chapters
In order to engage the widest possible population to disseminate information, encourage transparency, and build a shared vision, we need to create a global community. We encourage local chapters with their own events, projects, and engagement. Each local chapter is encouraged to apply for grants to help with the costs of running events that further our mission.