The President’s Council of Advisors on Science and Technology (PCAST) met on April 23 to receive an update on the report titled Supercharging Research: Harnessing Artificial Intelligence to Meet Global Challenges. An Executive Order on AI issued in October 2023 requested this to address “the potential role of AI, especially given recent developments in AI, [and] in research aimed at tackling major societal and global challenges”.
{Watch the presentation and discussion here}
Dr. Terrence Tao, University of California, Los Angeles, and co-chair of the AI working group presented on key opportunities for AI and provided examples of how it is already transforming science. Dr. Tao identified removing barriers, mitigating errors, and supporting a culture of responsible use through openness and reproducibility as critical areas well served by AI. Dr. Tao provided three points in the scientific process where AI is already changing science:
- Identifying candidate solutions to scientific problems. (materials science, therapeutic drugs, semiconductor design)
- Accelerating and enhancing scientific simulations and models. (Modeling in climate science and simulations; cellular biology; cosmology)
- Analyzing new types of data. (social sciences and health and wellness)
Dr. Tao elaborated on the analysis of new types of data, explaining that these fields are already working with large language models to analyze non-quantitative data, such as large-scale behavioral data in social media posts. In the future, he predicted that data driven social science will enable more effective, responsive, and fair delivery of public services. In health and wellness, AI is aiding physicians to successfully make early diagnosis of cancer and detect potential errors to enhance patient safety. Looking forward, he imagines ultra-personalized medicine will tailor health care to an individual’s specific genetics and medical history.
Dr. Laura Green, Florida State University followed with a vision for how AI could best empower human scientists through the responsible use of shared and open resources. Dr. Green pointed to the potential of AI to manage huge streams of data, facilitate new interdisciplinary collaborations, and support complex workflows. When elaborating on responsible use, the committee recommends that outputs be externally verified and tested to protect against bias. To reduce redundancy, expenses, and take advantage of open resources, the scientific community will need AI infrastructure optimized for efficient research processes.
Following an engaged and enthusiastic discussion, the committee unanimously approved the report.
