Photos by Getty Images and Unsplash, design by Kekeli Sumah/RAND | |
|
|
|
Don’t overreact: The United States still holds the AI competitive edge over China
|
|
|
The January 2025 release of a Chinese large language model by DeepSeek prompted questions in the United States about whether China had pulled ahead in AI leadership, as politicians and analysts highlighted the model’s potential speed and efficiency advantages. In a commentary for the ChinaTalk blog, RAND researcher Lennart Heim makes the case that while Chinese models are narrowing key gaps, the United States still leads on a critical metric: total compute capacity, or the combined processing power of advanced AI chips. With roughly 10 times more compute than China, Heim suggests the U.S. could effectively command 10 times more “virtual employees,” offering a long-term edge in productivity and innovation. While the United States holds this advantage today, Heim warns against complacency or overreacting to China’s recent gains. He calls for building the capacity to act swiftly — reinforcing export controls, assessing their impacts, and harnessing U.S. compute power to drive economic transformation and achieve global AI dominance. Read more: China's AI Models Are Closing the Gap—but America's Real Advantage Lies Elsewhere
|
|
|
|
|
Halting the wealth gap: RAND expert pitches policy options to manage AI-driven income inequality
|
|
|
As AI systems become more advanced, their potential to automate and even replace human labor grows — raising significant concerns about widening income inequality. In the new RAND report, “Managing AI’s Economic Future,” expert Tobias Sytsma explores how smart policy choices today can help manage AI-driven economic challenges. To formulate these policy options, Sytsma differentiates between “vertical” automation — improving productivity within already automated tasks, where AI enhances capital — and “horizontal” automation, where AI expands into new tasks, displacing human labor. To identify policies likeliest to achieve preferred outcomes — such as maintaining traditional growth patterns while addressing inequality — Sytsma simulated thousands of economic scenarios, varying key factors like automation rates and productivity improvements. By analyzing the scenarios, he identified which scenarios “succeeded” in achieving managed growth — like achieving at least 2 percent income growth and at most 2 percent inequality growth, or creating growth rates in excess of 10 percent while allowing inequality growth to expand as well. He then determined the probability of different strategies achieving these targets under various policy mixes and goals. In the end, Sytsma was able to tailor recommendations based on policymakers’ priorities. For example, policymakers aiming to sustain traditional growth while managing inequality should pursue a policy that promotes vertical automation while limiting horizontal automation. To achieve transformative growth, however, policymakers may need to fully embrace automation — which may involve higher risks and uncertainty. Read more: Managing AI's Economic Future: Strategic Automation Policy in an Era of Global Competition
|
|
|
|
|
|
|
Requests received: RAND researchers inform federal inquiries on AI policy
|
|
|
In recent months, the U.S. federal government issued several Requests for Information (RFIs) to gather public input that can inform robust AI policies and strengthen U.S. global AI leadership. RAND researchers provided rapid responses to two key RFIs:
|
|
|
- Government land for AI: A RAND research team, led by Ismael Arciniegas Rueda and Henri van Soest, responded to a U.S. Department of Energy RFI on using DOE-owned lands for AI data centers. Their working paper outlines key considerations on site suitability, technology for onsite energy development, and supply chain challenges to guide policy planning. Read more: Locating AI Data Centers on DOE Land.
- In-house AI talent: The White House Office of Science and Technology Policy requested public input to help shape a national “AI action plan” that supports U.S. leadership in AI while minimizing unnecessary regulatory burdens. In response, RAND researchers Gregory Smith, Elina Treyger, Elika Somani, and Margaret Siu developed a working paper focused on strengthening the U.S. government’s AI workforce, offering six policy options, such as expanding special hiring authorities, launching fellowships and exchange programs, and improving retention and career development for AI experts. Read more: Enhancing In-House U.S. Government AI Talent.
|
|
|
|
|
|
|
Why Some Scientists Become Rockstars After International Moves
|
|
|
Is it possible to find future high-output scientists early enough to guide visa or other policy decisions? When researchers pack up their labs to relocate internationally, some excel in their new environments while others struggle to adapt. This report by Jon Schmid, Prateek Puri, Tobias Sytsma, John Vahedi, Nathaniel Edenfield, and Amy M. Nice examines why some thrive and others just survive following an international move. The findings offer valuable insights for both researchers studying migration patterns and policymakers seeking to identify which scientists are most likely to contribute meaningfully to their new host countries. The research reveals several interesting patterns. High-performing scientists with strong track records, experience working in large teams, and existing international connections tend to remain successful even after relocating. Machine learning models can predict post-move success with 15% better accuracy than baseline approaches, and straightforward factors like past impact and collaboration history are surprisingly powerful indicators. The data also shows that different countries attract researchers at different career stages - the US attracts early-career researchers while losing experienced ones, while China experiences the opposite. Read more: Finding International Rockstar Researchers: Identifying Predictors of Success Following an International Move
|
|
|
Other TASP Publications This Quarter
|
|
|
We’ve had a busy spring! The Geopolitics of AGI team launched a Substack in June: Geopolitics of AGI | RAND | Substack Other publications in the past three months include research reports (original, peer-reviewed research), working papers (research released in draft form), “expert insights” (researcher perspectives on current topics), and commentaries and articles (short pieces such as blog posts).
|
|
|
Research Reports
|
|
|
- Algorithmic Advancement in Artificial Intelligence by Carter C. Price, Brien Alkire, and Mohammad Ahmadi
- On the Extinction Risk from Artificial Intelligence by Michael J.D. Vermeer, Emily Lathrop, and Alvin Moon
- Automated Grading for Efficiently Evaluating the Dual-Use Biological Capabilities of Large Language Models by Bria Persaud, Ying-Chiang Jeffrey Lee, Jordan Despanie, Helin Hernandez, Henry Alexander Bradley, Sarah L. Gebauer, and Greg McKelvey, Jr.
- Insights from a Workshop on Hardware-Enabled Governance Mechanisms by Gabriel Kulp, Margaret Siu, and Lennart Heim
- Strategies and Detection Gaps in a Game-Theoretic Model of Compute Governance by Alvin Moon, Padmaja Vedula, Jesse Geneson, and Simon Bar-on
- Finding International Rockstar Researchers by Jon Schmid, Prateek Puri, Tobias Sytsma, John Vahedi, Nathaniel Edenfield, and Amy M. Nice
- Assessing the United States' Additional AI Power Capacity by 2030 by Ismael Arciniegas Rueda, Henri van Soest, Hye Min Park, Austin Smidt, David Gill, Robin Wang, Kelly Klima, and Aisha Najera
- How Artificial General Intelligence Could Affect the Rise and Fall of Nations by Barry Pavel, Ivana Ke, Gregory Smith, Sophia Brown-Heidenreich, Lea Sabbag, Ashwin Acharya, and Yusuf Mahmood
- Managing AI's Economic Future: Strategic Automation Policy in an Era of Global Competition by Tobias Sytsma
|
|
|
Working Papers
|
|
|
- An AI Revolution in Military Affairs? How Artificial Intelligence Could Reshape Future Warfare by Zachary Burdette, Dwight Phillips, Jacob L. Heim, Edward Geist, David R. Frelinger, Chad Heitzenrater, and Karl P. Mueller
- Navigating the AI-Energy Nexus with Geopolitical Insight by Nidhi Kalra, Robin Wang, and Ismael Arciniegas Rueda
- What Even Superintelligent Computers Can’t Do by Edward Geist and Alvin Moon
- The Risks of Preventive Attack in the Race for Advanced Artificial Intelligence by Zachary Burdette and Hiwot Demelash
- Averting a Robot Catastrophe: Preparing for Converging Trends in Robotics and Frontier AI by Michael J. D. Vermeer, Timothy M. Bonds, Emily Lathrop, and Gregory Smith
- Riding the Coming Wave: The Urgent Need for Contingency Plans for AGI Futures by Joel Predd
- Mutual Dependence and Vulnerability in Human and AI Futures by Benjamin Boudreaux
- Locating AI Data Centers on DOE Land by Ismael Arciniegas Rueda, Henri van Soest, Aisha Najera, Kelly Klima, David Gill, Hye Min Park, Austin Smidt, Robin Wang, and Rahim Ali
|
|
|
Expert Insights
|
|
|
- Charting Multiple Courses to Artificial General Intelligence by William Marcellino, Lav Varshney, Anton Shenk, Nicolas M. Robles, and Benjamin Boudreaux
- Documenting Cloud Labs and Examining How Remotely Operated Automated Laboratories Could Enable Bad Actors by Ying-Chiang Jeffrey Lee, Bria Persaud, Barbara Del Castello, Allison Berke, and Gustavs Zilgalvis
- Preliminary suggestions for rigorous GPAI model evaluations by Patricia Paskov, Michael J. Byun, Kevin Wei, and Toby Webster
- Insights from Nuclear History for AI Governance by Benjamin Boudreaux, Gregory Smith, Edward Geist, and Leah Dion
- Who Could Fund Future Artificial Intelligence Development? by Edward Parker, Benjamin M. Miller, and Colin Levaunt
- Toward Best Practices for AI Evaluation and Governance by Patricia Paskov, Lisa Soder, and Everett Smith
- RAND Europe's Response to the EU Cloud and AI Development Act Call for Evidence by Afek Shamir and Maximilian Negele
- Artificial Intelligence, Cybersecurity, and National Security: The Fierce Urgency of Now by Richard Danzig; in prepublication
|
|
|
Commentaries and Articles
|
|
|
|
|
|
|
|