Introduction

Artificial intelligence is driving new discoveries and advancements across many fields of science and engineering. Modern applications of AI and machine learning have the potential to change the practice of research and development (R&D) in the United States over the next decade. This data brief is a first step toward understanding how modern AI and ML have begun accelerating growth across various science and engineering disciplines in recent years.

The application of AI to science is not new. Pioneers at Stanford University in the 1960s used early AI systems to automate analyses of chemical structure and mass spectra, supported by the Advanced Research Projects Agency (ARPA), the predecessor to the Defense Advanced Research Project Agency (DARPA), and NASA. This work grew out of an interest in using AI systems for complex reasoning in scientific problems.

Today, deep learning methods, a specific subset of ML, can predict the 3D structure of proteins to within the width of an atom— substantial progress on a 50-year challenge achieved with modern machine learning. This tool is important because it may contribute new capabilities for disease treatment and pandemic response. In the years ahead, modern ML may help advance the engineering design of viable, cost-effective fusion reactors, a large-scale source of clean energy that has also been out of reach for more than 70 years. AI applications can drive advancements in fields as disparate as gene editing and chip design.

American institutions have begun to adapt to this future. MIT, for example, announced its largest structural change since the 1950s and invested $1 billion to found a new college focused on the intersections between computing, AI, and the university’s existing disciplines of science and engineering. Students will be encouraged to be “bilingual” between computing and other disciplines.

As modern AI applications begin to transform science and engineering, a key question for high-level leaders shaping science and technology strategies is what disciplines will be affected first, and will these early impacts be narrowly confined to a few disciplines (like materials science and biomedical fields) or broadly applicable? Accelerated emergence and growth of new fields is especially noteworthy: bioengineering, computer science, and materials science have all emerged as enabling disciplines that changed science and engineering broadly in the last century—and were followed by national and global changes.

This data brief provides an initial look at how modern AI and ML have begun accelerating growth across a wide array of research disciplines in recent years, and focuses on direct applications of AI to making new scientific and engineering breakthroughs. We first summarize illustrative examples of AI driving innovation in medicine, fundamental sciences, and engineering research. We then analyze global scientific publications data to see if these recent achievements represent real trends in published research.