PCAST Backs Use of AI to Accelerate Scientific Research
President Joe Biden at a briefing on artificial intelligence from members of the President’s Council of Advisors on Science and Technology in September 2023.
(Cameron Smith / The White House)
The President’s Council of Advisors on Science and Technology published a report on April 29 exploring how artificial intelligence can “supercharge” research and help tackle major societal challenges. The report emphasizes that AI should support, not replace, human scientists.
“The scientific enterprise is an excellent ‘sandbox’ in which to practice, study, and assess new paradigms of collaboration between humans and AI assistants,” the report states. “The objective should not be to maximize the amount of automation, but to allow human researchers to achieve high quality science that utilizes AI assistance responsibly.”
Examples of how AI can support research cited in the report include teams of AI assistants seeking patterns in huge volumes of data, enhancing existing simulations and models, and identifying candidate solutions to pressing research questions.
The report recommends that agencies that fund research should encourage scientists to integrate AI into their workflows. It also proposes that agencies consider requiring their grantees to provide plans describing how they will responsibly use AI tools and assess their risks, noting such a requirement could be aligned with the White House’s Blueprint for an AI Bill of Rights and the National Institute of Standards and Technology’s AI Risk Management Framework.
In light of the huge amount of computing power and data required to make the most of AI, the report recommends fully funding NSF’s National Artificial Intelligence Research Resource, highlighting its potential as a stepping stone for even more ambitious programs. It also expresses support for expanding NSF’s National Secure Data Service Demonstration project and the Census Bureau’s Federal Statistical Research Data Centers to give researchers greater access to federal datasets. In addition, NSF’s Materials Innovation Platforms program is spotlighted as an example of a data-sharing infrastructure that uses AI tools to build community with other agencies and industrial partners to more quickly discover novel materials — a framework that could be replicated in other fields.
PCAST unanimously approved the report last week during a virtual meeting, where lead authors of the report highlighted how AI is poised to accelerate research in various fields.
Terence Tao, co-chair of the PCAST Working Group on AI and a professor of mathematics at UCLA, said during the meeting that AI is being used to identify new molecules for therapeutic drug development, advance personalized medicine, and develop more advanced climate models to understand the potential impact of climate change and catastrophic weather events. He also highlighted AI’s potential to discover future materials such as room-temperature superconductors and new battery materials for electric vehicles, and how AI-assisted models could help discover new fundamental theories in physics.
“We’ve seen that AI is already actually transforming pretty much every scientific discipline and will transform every aspect of the scientific workflow,” Tao said.
The PCAST report is one of several initiatives completed in response to President Joe Biden’s executive order on the responsible use of AI, published in October. The Department of Energy, for example, recently launched a website highlighting DOE-developed AI tools and published a report on emerging applications of AI in energy research, among other actions.
