Key Moments
Michael Kratsios: Inside the White House's AI Strategy
Y CombinatorWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
The US government supports open-source AI but needs input from the startup community to avoid rigid regulations that stifle innovation, as seen with the EU AI Act's failure to account for LLMs.
Key Insights
The US AI strategy, released in July last year, explicitly states a commitment to fostering a vibrant closed and open-source AI ecosystem.
The EU AI Act was finalized before ChatGPT's release, highlighting the challenge of regulating rapidly evolving technologies with fixed rules.
The administration's priority is to reduce regulatory barriers for 'born-free' technologies and streamline processes for 'born-in-captivity' ones, such as commercial drone operations.
The government's role is to 'shape the arena, not direct discovery' in scientific advancement, recognizing AI's transformative impact on research.
Quantum information science is a critical future technology that requires significant prioritization, with an ambitious goal for a quantum computer by 2028.
The White House advocates for a national standard for AI regulation to prevent a patchwork of state laws that disproportionately burdens startups.
The US commitment to open-source AI
The White House strongly supports open-source AI, a stance articulated in the nation's AI strategy released in July of the previous year. Michael Kratsios, Director of the White House Office of Science and Technology Policy, emphasizes that a thriving open-source ecosystem is crucial for American leadership in artificial intelligence. He highlights the importance of open dialogue between policymakers and the tech community, including startups, to inform effective policy. Recent advocacy from companies and organizations for open-weight models demonstrates the dynamic nature of this conversation, forcing policymakers to internalize community beliefs. Kratsios finds these interactions not only inspiring but also practically helpful in ensuring policy decisions are well-informed and aligned with the nation's innovation goals.
The challenge of regulating rapidly evolving tech
Regulating a field as dynamic as AI, which reinvents itself every six months, presents a significant challenge. Kratsios points to the EU AI Act as a prime example of how rigid, pre-defined regulations can become obsolete. This act was finalized years before the advent of LLMs like ChatGPT, meaning it now applies to technologies that didn't exist at the time of its creation. Similarly, the US administration recognizes that hard-capped compute thresholds for disclosures can quickly become outdated. The strategy, therefore, is to avoid setting firm red lines and instead ensure that policy actions can adapt to the technological frontier. This cautious approach aims to prevent regulations from hindering rather than supporting innovation.
Balancing AI risks: cyber threats and biological concerns
When considering AI risks, Kratsios identifies two primary concerns: cyber risks and biological risks. He notes that the sophistication of AI models, while posing potential cyber threats, also offers significant benefits in hardening existing systems. This inherent trade-off means that the same powerful models capable of nefarious actions can be invaluable in bolstering cybersecurity. Regarding biological risks, Kratsios believes that while the potential exists, the alarm raised in previous years (2021-2022) has been somewhat overblown. Nevertheless, he stresses the necessity of building robust infrastructure and evaluation processes to test models as they advance. The focus remains on understanding and mitigating genuine risks while acknowledging the dual-use nature of many AI capabilities.
Championing 'little tech' in policy decisions
Kratsios acknowledges that policy conversations in Washington can often be dominated by the five largest tech companies. To counteract this, the White House is committed to ensuring that smaller tech companies, or 'little tech,' have a voice. He emphasizes the importance of mechanisms like requests for proposals and information, which allow for broad input. A key example of this commitment is the push for a single national standard for AI regulation. Kratsios explains that a patchwork of state-level regulations, while manageable for large corporations with extensive legal resources, would be prohibitive for startups. The President's stance on a unified national framework aims to create regulatory clarity and ease of compliance, fostering an environment where any entrepreneur can build an AI company without being hampered by complex, fragmented rules.
Shaping the scientific arena with AI
The White House's recent report, 'A New Golden Age for Science,' outlines a vision for revitalizing the nation's R&D ecosystem, with artificial intelligence at its core. Historically, the government funded the majority of R&D, but this has inverted, with the private sector now dominating. The report posits that AI will fundamentally transform scientific discovery across fields like material science, pharmaceuticals, and chemistry. Kratsios stresses that the government's role is to 'shape the arena, not direct discovery.' This involves fostering infrastructure for autonomous experimentation, where AI agents can design experiments, operate robotic labs, and analyze results. The administration aims to support this by making R&D funding more agile, establishing prize competitions, and ensuring all major agencies develop AI action plans, thereby enabling scientists to make groundbreaking discoveries more efficiently.
Prioritizing quantum information science
Beyond AI, quantum information science, including quantum computing, is identified as a critical technology that warrants significant attention. Kratsios draws a parallel to the early days of AI advocacy, noting that quantum computing is on the cusp of a similar surge in importance. While basic scientific questions remain about its application in computing and sensing, the government is actively prioritizing this field. An executive order on quantum information science has been signed, and the Department of Energy is tasked with building a scientifically relevant quantum computer by 2028. Kratsios predicts that in three to five years, the progress in quantum technology will be substantial and widely recognized, mirroring the current explosion in AI capabilities.
The role of Congress and intellectual property
While executive orders can quickly shape AI policy, Kratsios notes the challenges in legislative action. Laws passed by Congress are more durable than executive orders, which can be revoked by a new administration. The White House has proposed legislative solutions, urging Congress to act on preemption to establish a national AI standard. Another critical area for congressional action is intellectual property, particularly concerning AI's impact on creators' names, images, and likenesses. Kratsios emphasizes the need for protections and clarity for creators, stating that AI models outputting copyrighted characters, like Mickey Mouse, are unacceptable. He also suggests that the market can play a role in developing revenue-sharing models for creators, citing early examples like Yelp's deal with ChatGPT.
Encouraging public service in technology
Kratsios strongly encourages young people passionate about technology and its impact on the country to consider public service. He describes government work as deeply fulfilling and rewarding, offering the opportunity to tackle significant national challenges. While acknowledging the bureaucratic hurdles, he highlights the profound impact that successful policy initiatives can have on citizens' lives. The White House Office of Science and Technology Policy actively seeks talented individuals to help create an environment where startups can thrive and innovation can flourish. For those in the audience interested in contributing to the nation's technological future, he suggests exploring opportunities within the office.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Concepts
●People Referenced
Navigating AI Policy: Key Takeaways for Innovators
Practical takeaways from this episode
Do This
Avoid This
R&D Funding Breakdown: Government vs. Private Sector
Data extracted from this episode
| Time Period | Federal Government Funding (%) | Private Sector + Philanthropy Funding (%) |
|---|---|---|
| Circa 1950 | ~70% | ~30% |
| Present Day (approx.) | ~30% | ~70% |
Common Questions
Kratsios believes American leadership in emerging technologies is critical and that the U.S. government provides a unique platform to work on problems of immense scale, offering a rewarding and fulfilling experience.
Topics
Mentioned in this video
Director of the White House's Office of Science and Technology Policy and President's top advisor on science and technology. Key architect of America's national AI strategy.
Former President of the United States, under whom Michael Kratsios served as Chief Technology Officer and later worked to revitalize the science ecosystem.
Former CEO of Apple, whose keynotes Michael Kratsios followed obsessively, indicating an early passion for technology.
Venture capitalist and co-founder of PayPal, Michael Kratsios worked for him for nearly seven years, gaining experience with his investments.
Head of the Office of Management and Budget, whose life mission is to eliminate regulations.
U.S. President who wrote a letter to his science advisor, Vannevar Bush, in 1945, asking about the post-WWII science ecosystem, inspiring the 'Science, A New Golden Age' report.
Science advisor to FDR, author of the seminal report 'Science: The Endless Frontier', which laid out the government's role in funding basic research.
Founder of The Walt Disney Company, mentioned in the context of intellectual property rights concerning AI model outputs, stating that only Disney should be able to output 'Mickey Mouse'.
The executive branch of the U.S. federal government where Michael Kratsios serves. Involved in shaping AI policy and strategy.
Company where Michael Kratsios served as Chief Operating Officer between government roles, gaining experience in a hyper-growth startup.
The Federal Aviation Administration, mentioned in the context of SpaceX's challenges with launch permits and its role in regulating commercial drone operations.
A U.S. government department mentioned in relation to regulatory challenges faced by companies like SpaceX.
The current U.S. administration mentioned for setting a specific hardcap compute threshold for AI disclosures, which Kratsios suggests may not stand the test of time.
A trade association formed by Y Combinator and other companies to represent the interests of smaller tech firms in policy discussions.
The Department of Justice, mentioned in relation to antitrust cases against big tech companies initiated under the Trump administration.
The Federal Trade Commission, mentioned in relation to antitrust cases against big tech companies initiated under the Trump administration.
Department of Health and Human Services, mentioned as an example of a government agency focused on a specific policy area.
A government department focused on defense, one of the agencies that needs to be coordinated by the White House on technology policy.
Government department responsible for energy, including running national labs, and a key player in quantum computing initiatives.
The National Aeronautics and Space Administration, involved in coordinating space policy across the government.
The first smartphone from Apple, its release in college was a memorable moment for Michael Kratsios, highlighting his early interest in technology.
A copyrighted character mentioned as an example of an AI model output that would infringe on intellectual property rights, emphasizing the need for clear statutes.
A ride-sharing company that Michael Kratsios mentions as an example of a company facing regulatory challenges at state and local levels.
Space exploration company co-founded by Elon Musk, which Michael Kratsios cites as an example of a company facing challenges with launch permits from government agencies.
Startup accelerator that organized a letter advocating for open-weight models and helped spin up the 'Little Tech Association'.
A major tech company that can navigate complex state-by-state regulations, contrasted with smaller startups that would struggle with a patchwork of rules.
Company focused on supersonic flight, used as an example of technological stagnation due to regulatory barriers (noise limits instead of speed limits).
A company that recently entered a deal with ChatGPT to use some of its reviews, indicating a developing market for licensing IP to AI companies.
European Union's regulation on artificial intelligence, cited as an example of regulation written before major advancements like ChatGPT, highlighting the challenge of regulating fast-moving tech.
The first executive order prioritizing AI in U.S. history, signed by President Trump in February 2019, which set out the first national strategy on AI.
A recent executive order signed by the President to prioritize quantum information science and computing.
A large language model developed by OpenAI, its release significantly impacted the AI landscape and highlighted the challenges of regulating technology developed after legislation is enacted.
Mentioned as a powerful AI model that presented cybersecurity risks, requiring careful guardrail implementation during its release.
A field of technology that Michael Kratsios believes Washington will care enormously about in five years, currently discussed but not receiving enough airtime.
A rapidly advancing field of computation that Michael Kratsios highlights as a key future technology that the government needs to prioritize, similar to AI's rise.
More from Y Combinator
View all 626 summaries
52 minSusan Kare: Designing Icons & Graphics For the Original Mac
59 minChelsea Finn: This is the State of the Art in Robotics
42 minPeter Steinberger: What Happens When 4.7 Million People Let It Cook
85 minWhy Robotics Still Isn't Solved - But Could Be Soon | YC Paper Club
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free