Supercharging Research:
Harnessing Artificial Intelligence
to Meet Global Challenges

Artificial Intelligence (AI) has the potential to revolutionize our ability to address humanity’s most urgent challenges by providing researchers with tools that will accelerate scientific discoveries and technological advances. Generative AI, which can create content based on vast data sets and extensive computation, is poised to be particularly transformative. Examples of generative AI include large language models, image generating models, and generative scientific models. In his comprehensive Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, issued on October 30, 2023, President Biden charged PCAST to report on “the potential role of AI…in research aimed at tackling major societal and global challenges.” PCAST is pleased to offer this report in fulfillment of this charge. With well-designed, equitably shared, and responsibly used infrastructure, AI will enable scientists to address urgent challenges, including improving human health and enhancing weather prediction in a time of climate change. AI can help explore long-standing scientific mysteries that inspire and stretch human creativity, such as uncovering the origin and evolution of the universe. AI will also help researchers address continuing national needs, from accelerated semiconductor chip design to the discovery of new materials to address our energy needs. Furthermore, AI is starting to remove barriers that make scientific research slow and expensive, for instance by providing the means for rapidly identifying the best drug candidates for testing (thus reducing the number of expensive laboratory trials), helping to optimize experimental designs, and uncovering connections in data much more efficiently than can be done by hand or using traditional data science methods. If basic AI resources, validated data, and scientific tools and training are made broadly accessible, AI technologies have the potential to democratize scientific knowledge, bringing interconnected technical concepts to many more people and enabling diverse researchers to bring their expertise and perspectives to societal and global challenges. Just as with any other new tool or technology, realizing the potentials of AI will require addressing its limitations. These issues include misleading or incorrect results, perpetuation of bias or inequity5 and sampling errors from patterns embedded in the model-training data, limited access to high quality training data, the challenges of protecting intellectual property and privacy, the significant energy required to train or deploy a model or run the AI algorithms, and the risk that bad or nefarious actors will use readily available AI tools for malicious purposes. Many public and private sector activities addressing these issues are already underway, including government efforts tasked under the October 2023 Executive Order on AI. Reproducibility and validation are key principles underlying scientific integrity and the scientific method and must continue to be held at high value as we develop a culture of responsible AI use and expert human supervision of AI applications. AI has the potential to transform every scientific discipline and many aspects of the way we conduct science. Scientists are already employing AI to create new functional materials that we presently do not know how to design; these include superconductors and thermoelectric materials which would not only enhance our energy efficiency but also reduce our carbon footprint. In a similar vein, AI models are helping researchers create new designs for manufacturing processes and products, and develop new drug therapies which in the future could enable individualized treatment of specific cancers and viruses. AI models are also helping engineers design semiconductor chips, producing better designs with less human effort and time. In health care, AI technologies are creating new ways to analyze a broad spectrum of medical data for applications like the early diagnosis of diseases that can lead to timely intervention and the detection of medical errors. PCAST also foresees widely available AI-powered ultra-personalized medicine tailored to a specific individual and disease process that will include details of medical history, genetic information, and signals, such as how healthy and unhealthy cells are behaving. AI is also transforming science by improving our scientific models. In climate science, AI models are starting to enhance weather prediction, as well as advancing whole-earth models for water management, greenhouse gas monitoring, and predicting the impacts of catastrophes. Scientists have already used AI to successfully predict the structure of proteins; new foundation models will unlock more secrets of cellular biology and power computer simulations of intracellular interactions that can be used to explore new therapies. AI models promise to help us understand the origin of our universe by allowing us to test numerous cosmological hypotheses via rapid simulations. Such AI- enabled modeling may even help scientists discover new laws of physics. AI will enable unprecedented advances in the social sciences, complementing qualitative methods with new quantitative techniques for analyzing existing data, as well as the development and analysis of newer types of data, e.g., step counts on smartphones, anonymized data drawn with permission from search and browsing, or images posted on social media. AI could supercharge research using vast data sets, such as those that have long been collected and curated by federal statistical agencies—ideally complemented by those held by the private sector—as input for designing effective federal policies. Application of AI to these long-standing and newer social science data sets could facilitate more effective, responsive, and fairer data-driven policymaking and delivery of services. These few examples of AI-assisted research illustrate that with the responsible use of AI technology, human scientists will be empowered to realize transformational discoveries. Furthermore, PCAST expects that responsible sharing of basic AI resources will help to democratize science and tackle major societal and global challenges. The use of AI for science and technology research is accelerating rapidly across the globe and therefore demands our commitment to U.S. leadership in the applications of this powerful new tool. Building on the work of the Biden Administration, the United States must act boldly and thoughtfully to maintain our nation’s lead in research, in the innovative applications of AI, and in establishing frameworks and norms for the safe and responsible use of AI. In this report, PCAST offers five specific findings and recommendations for action that will help the U.S. to harness the full potential of AI to equitably and responsibly supercharge scientific discovery. Recommendation 1: Expand existing efforts to broadly and equitably share fundamental AI resources Recommendation 2: Expand secure access to federal data sets for approved critical research needs, with appropriate protections and safeguards Recommendation 3: Support both basic and applied research in AI that involves collaborations across academia, industry, national and federal laboratories, and federal agencies as outlined in the vision for the NAIRR developed by the NAIRR Task Force. Recommendation 4: Adopt principles of responsible, transparent, and trustworthy AI use throughout all stages of the scientific research process. Recommendation 5: Encourage innovative approaches to integrating AI assistance into scientific workflows.