Michael Kratsios on the New Golden Age of American Science | EP #276

Michael Kratsios serves as the 13th director of the White House Office of Science and Technology Policy and is the principal architect behind America's AI action plan, the Genesis Mission, and the Golden Age of American Science report. The Genesis Mission is framed as a Manhattan Project style effort to apply AI across all of government with the stated goal of doubling the productivity of the entire US scientific enterprise over the next decade, though Kratsios now believes the target should probably be 10X given transformations in AI over the last six months.

Kratsios distinguishes between technologies born free, like the internet, which benefit from government restraint, and technologies born in captivity, like commercial drones and AI-powered medical diagnostics, which cannot reach commercialization without affirmative government action. He cites the EU AI Act as a cautionary example, noting it was finalized before ChatGPT was invented and therefore cannot properly apply to today's large language models. The administration's stated priority since day one has been maintaining US leadership in AI by creating a permissive regulatory environment rather than imposing heavy restrictions.

Three quarters of Americans currently fear AI, and 71 percent oppose data centers near their homes, a higher percentage than those who oppose nearby nuclear power plants, compared to 80 percent of Chinese citizens who are pro-AI. Kratsios attributes poor American public perception partly to the prior administration's narrative fixating on job losses and danger, and to the first AI safety summit at Bletchley Park framing AI almost entirely around fear. He identifies healthcare as the domain where Americans most positively connect with AI and considers it the best starting point for shifting the national narrative. He also acknowledges that current data on AI's actual impact on the labor force is poor, and notes that some companies label layoffs as AI-related when those layoffs would have occurred regardless because doing so boosts stock price.

The administration has set several concrete science and space targets. The United States plans to return a man to the moon in 2028 and build the first elements of a lunar base by 2030. A nuclear reactor with sufficient propulsion power to send people to Mars is planned for space by 2028. The president directed through executive order the creation of a scientifically relevant quantum computer by the end of his term, with pharmaceuticals identified as the most transformative application due to its relevance to molecular calculations in drug development. Private sector investment in fusion energy is at a historic high, with approximately 37 venture-backed fusion companies currently operating.

On competition with China, Kratsios argues that EUV lithography export controls implemented in 2019 are among the most impactful export controls in US history and have significantly throttled China's ability to produce leading-edge chips. He rejects the argument that Nvidia chip export restrictions were a mistake, maintaining that the US lead over the best Chinese chip continues to increase year over year partly as a result. China has over 150 humanoid robot companies compared to a countable few in the United States, and Kratsios warns that Beijing is exporting open source AI as an instrument of influence, particularly to the global south, with some cash-strapped American entrepreneurs already using Chinese open source models because they are among the cheapest available.

The existing federal grant system has structural problems the Golden Age report aims to correct. The NIH budget has grown to nearly 45 billion dollars yet drug costs continue to rise, consistent with Eroom's Law of declining scientific productivity. Approximately 45 percent of a researcher's time is spent on administrative work associated with their grant. Most grants run roughly 18 months driven by academic calendar logistics, scientists begin applying for their next grant before finishing their current one, and least-common-denominator ideas get funded while high-risk ideas are not even submitted. The median age of an intramural NIH scientist is 71, while the average age at which Nobel laureates did their prize-winning work is in their mid-20s.

The report proposes several corrective mechanisms including a golden ticket allowing any single reviewer to unilaterally fund a grant regardless of the rest of the committee, fast grants modeled on the Covid-era approach where decisions were made in hours, and the use of prediction markets and decentralized autonomous organizations to fund research directly. Autonomous self-driven labs where AI proposes hypotheses, robots run experiments, and the model designs the next experiment with no humans in the loop represent one of the most significant near-term opportunities Kratsios identifies, with Lila Sciences building out a million square feet of such space. US national labs hold 70 years of scientific data that has not been made AI ready, and Kratsios argues making that data available as a public good represents a major opportunity to accelerate discovery.