The Computing Community Consortium (CCC) has released a draft version of its 20-year roadmap for AI research in the United States. Made after consultation with researchers and tech companies, the roadmap calls for sustained support from the federal government and prescribes a number of steps to ensure the U.S. retains its position as a nation with some of the most advanced AI resources on the planet.
Among the proposed steps:
- Create an open AI platform that includes a collection of data sets, knowledge repositories, and libraries available to and made in part by researchers in academia, government, and business.
- Launch national AI competitions that challenge researchers to solve big problems and push the community to achieve state-of-the-art results.
- Open national research centers and AI laboratories to support the Open AI platform, competitions, and research fellows.
- Support the research of self-aware learning — AI that can learn by example. Self-aware learning is listed among three major areas of investment for impact in the future. Research to better understand human intelligence is also needed.
- Create recruitment programs to identify and attract talented students, as well as people from underrepresented groups in the AI industry, such as women and people of color.
The roadmap also supports the development of lifelong personal assistants to augment human ability in education, health care, and industry.
According to the roadmap: “… lifelong personal assistants will enable an elderly population to live longer independently, AI health coaches will provide advice for lifestyle choices, customized AI tutors will broaden education opportunities, and AI scientific assistants will dramatically accelerate the pace of discovery.”
Organizers believe such efforts could promote universal personalized education, accelerate scientific discovery, and drive business innovation.
The roadmap is meant to identify challenges, opportunities, and pitfalls for AI researchers in the U.S. It’s also meant to let the AI community set the tone for research and funding priorities at a time when artificial intelligence is an increasingly political subject.
“If we don’t address [the challenges of AI], then others will and might force things on us that we actually don’t like,” Cornell University professor and roadmap co-chair Bart Selman said in January, when initial results were shared at a town hall meeting at the Association for the Advancement of Artificial Intelligence (AAAI) conference in Hawaii.
The roadmap release follows a series of workshops with companies and researchers held last fall and in early 2019. Organizers of the roadmap include University of Southern California director of knowledge technologies Yolanda Gil and Dr. Fei-Fei Li, a professor at Stanford University and until last year Google’s chief AI scientist.
The document makes its public debut weeks after the signing of the somewhat vague and less than substantive American AI Initiative by President Trump. Before resigning in protest earlier this year, former Defense Secretary Jim Mattis reportedly urged President Trump to create a national AI strategy akin to that of the Chinese government.
National AI labs could help address a shortfall of resources required to create more advanced systems, such as continuous data collection and social experimentation.
“This requires new facilities that do not exist in academia today. Although major AI innovations have roots in academic research, universities now lack the massive resources that have been acquired or developed by major IT companies. These are fundamental capabilities to build forward-looking AI research programs. This also puts universities at a serious disadvantage in terms of attracting talented graduate students and retaining influential senior faculty,” the roadmap reads.
The report cites interdisciplinary teams that draw knowledge from disciplines like psychology, biology, social science, and public policy as being only rarely available to U.S. researchers today.
It also notes a gap between the demand for well-educated AI talent and the supply coming from schools in the country. Most universities lack the resources to adequately prepare graduates for jobs in the industry, and “the need for AI expertise surpasses current production of university graduates with AI skills at the undergraduate, masters, and PhD levels,” the report notes. To make matters worse, “Many PhD-level AI graduates in the U.S. find attractive opportunities abroad.”
The AI research roadmap follows an approach similar to that taken by the CCC in creating a Robotics Roadmap in 2009.
Director Ann Drobnis plans to invite the AI community to comment on the roadmap later this month, when a larger version of the report is released, she told VentureBeat in a phone interview. The CCC will make a final copy of the report available in April.
The results could help shape AI research in the U.S. in the years ahead. In addition to working with AI researchers, the CCC is closely involved with the National Science Foundation, a major backer of academic research projects.
The 20-year AI research roadmap is running concurrently with another CCC interim report. Released last week, that report is intended to gauge how relationships between large tech companies and research institutions at universities are altering research topics and academic culture.