American AI Leadership Should Not Be Defined by Machine Learning 

This report is part of the larger compendium “Future-Proofing U.S. Technology: Strategic Priorities Amid Chinese Tech Advancement.”

Read the full report

Read the full compendium

Executive Summary

The United States’ leadership in developing artificial intelligence should not be defined just by machine learning. This paradigm, in which artificial neural networks learn via data, is a critical step in making progress with this technology. Yet machine learning is a fundamentally limited paradigm, whose shortcomings cannot be overcome by doubling down on its existing techniques. U.S. policymakers should instead reconceive American AI leadership as investing in and pushing the boundaries of the next dominant paradigm in AI, whose ideal candidate is neuro-symbolic AI. Neuro-symbolic AI synthesizes techniques from both traditional and contemporary approaches to AI research. Thus, it demonstrates the most promising path to ameliorating shortcomings in state-of-the-art models without sacrificing what came before. 

The Next Wave of AI: Neuro-Symbolic AI  

Policymaking efforts to retain and expand the United States’ AI leadership should not concede the future of this technology merely to control its present — because its present is fundamentally limited. Machine learning is not the paradigm that will, once fully realized, secure for the U.S. an enduring leadership position in AI.  

This paper explores a new paradigm that is needed to ensure U.S leadership: neuro-symbolic AI. Rather than repeat the mistakes of the past, the U.S. government’s role should instead be relatively targeted and complementary, and thus it should prioritize shortcomings in state-of-the-art machine-learning systems ripe for improvement in the next paradigm. Rather than pursue artificial general intelligence, the federal government should invest in frontier neuro-symbolic AI research by laying its foundations through existing offices and programs like the National Artificial Intelligence Initiative Office (NAIIO) and the National Science Foundation’s National AI Research Institutes.  

Policy Recommendations  

1. The NAIIO, together with the Subcommittee on Machine Learning and AI, should therefore direct the AI R&D Interagency Working Group to prioritize investments in neuro-symbolic techniques. 
2. An institute for neuro-symbolic AI should engage in public-private collaboration in earnest, prioritizing those actors willing to collaborate on innovative research in this emerging paradigm. 
3. Budget cuts for basic research funding should be reversed to allow agencies to invest in foundational research of a sufficiently interdisciplinary nature for the Third Wave of AI development.  
4. The U.S. Congress and the Commerce Department should ensure that U.S. export controls on hardware or models are aggressively proactive yet targeted, are proportional to the actual capabilities of the AI systems they enable or constitute and are implemented in coordination with partners and allies. 

Vincent J. Carchidi is an analyst focusing on critical and emerging technologies in U.S.-China technology competition and the Middle East. He is a Non-Resident Scholar at the Middle East Institute’s Strategic Technologies and Cyber Security Program. He is also a Non-Resident Fellow at the Orion Policy Institute, specializing in AI policy. His tech policy work appears in outlets including Defense One, National Interest, The Hill, Trends Research & Advisory, and the Foreign Policy Research Institute. Carchidi maintains a background in cognitive science, applying this research to the trajectories and limitations of frontier AI models. His opinions are his own. You can follow him on LinkedIn, BlueSky, and X. 

The views expressed in this article are those of the author and not an official policy or position of New Lines Institute.