Interview, Fireside Chat
War with Iran + Pentagon vs Anthropic with Under Secretary of War Emil Michael
- Undisclosed projects are expected to release news within the coming weeks.
- Current market predictions indicate a 40% probability of U.S. forces having boots on the ground in Iran by the end of March, rising to a 59% chance by the end of the year.
- PolyMarket data suggests a 39% probability of the Iranian regime falling by June 30th, increasing to a 51% chance by the end of the year.
- Military operations in Iran are projected to last weeks rather than months, aiming for regime disarmament in a "one and done" format to avoid protracted conflict.
- A major "grand bargain" or deal with China is anticipated during presidential meetings in April, potentially serving as an administrative crowning achievement before the midterms.
- Current actions in Iran and Venezuela are expected to generate maximum leverage for negotiations with China, which accounts for 90% of Iran's oil exports.
- China's GDP growth is predicted to remain between 4.5% and 5%, representing the lowest rate in approximately 30 years, which may lead to internal chaos if growth continues at half the historical rate despite a doubled population and GDP.
- Domestic economic pressure in China is predicted to make invading Taiwan a strategic interest to create a resource sinkhole and stimulate a war machine.
- U.S. military posture may shift toward a "grand bargain" with China regarding Taiwan, regulatory parity, and semiconductor availability.
- Drone deployment volumes from the U.S. have recently exceeded the total deployed in all prior military history, signaling a significant warfare shift.
- Commercial drone costs are estimated at $35,000, while one-way attack drones are priced between $50,000 and $80,000 depending on equipment.
- U.S. operations are expected to utilize drones for surveillance and protection at major events, including the 2028 Olympics and World Cup.
- Future conflict is predicted to involve AI-controlled drone swarms, requiring AI systems to be significantly more accurate than humans to distinguish targets from decoys before being used for lethal decisions in populated areas.
- China is expected to advance significantly in drone swarm interoperability compared to current U.S. capabilities.
- AI applications for defense, such as intercepting hypersonic missiles, are projected to develop substantially over the next decade.
- The U.S. defense strategy is anticipated to shift from high-cost carrier ships to mass-attributable, low-cost defense systems.
- The Department of Defense expects to require increased munitions production as existing stockpiles are depleted by current conflicts.
- Critical minerals and battery production are expected to be brought to the U.S. or allied countries within the remainder of the current presidential term.
- DARPA is expected to utilize biological synthesis for critical minerals to leapfrog Chinese refining technology.
- Cyber attacks leveraging AI agents are predicted to become a major threat to military systems.
- U.S. operational expertise in space and submarines is expected to remain the global best, though efforts are needed to prevent the gap with China from narrowing.
- The U.S. defense industrial base is expected to transition from cost-plus contracts to fixed-cost pricing to disrupt bureaucratic advantages.
- The U.S. may require several years to build out new domestic battery manufacturing capacity.
- Google is expected to release a virtual version of "Cursor" within 90 days with potential G Suite integration.
- Elon Musk expects Grok to surpass all coding models by June.
- The U.S. defense sector aims to transition from 1980s Cold War weapons to modern systems.
- A "flywheel" of venture capital and entrepreneurship in defense tech is expected if startups secure large contracts quickly.
- The U.S. expects to replace nuclear missiles that are 50 years old and aircraft that are 40 years old.
- To ensure redundancy against policy shifts or failures, the U.S. expects to maintain at least two or three AI models.
- The U.S. anticipates risks from relying on a single AI provider that may change terms of service arbitrarily.
- AI is expected to save lives in scenarios requiring reaction times faster than humans can provide, such as intercepting hypersonic missiles.
- Risks include AI models being stolen by China and used unguarded against U.S. systems.
- The U.S. may need to deprecate tools like Cursor if skyrocketing inference costs result in negative ROI.
- The U.S. aims to build a complex of models rather than relying on a single provider as performance asymptotes.
- Operational strategies may include "unentanglement" from providers like Anthropic to ensure reliability.
- Efforts are expected to ensure AI is used for "lawful use" without external entities blocking kinetic strikes during decisive actions.
- AI terms of service are expected to be written into law to prevent internal policy changes from affecting government operations.
- Supply chain risks include foreign powers poisoning models or tricking them into incorrect instruction execution.
- Competition on price and reliability is expected to increase as AI models become more similar in performance.
- Safeguards are needed to prevent AI from dictating government operations or acting as a "god king."
- Preventing mass surveillance of Americans is a priority when deploying AI systems in the military.
- AI deployment must avoid reflecting private company philosophies or becoming tools for political perspectives.
- Mitigation strategies are required against AI becoming a platform for de-platforming at scale 1000 times worse than social media.
- AI must not become a tool for coercion by small countries or a "nuclear bomb" capable of 50,000 geniuses in a data center.
- AI must not become the sole "substrate of intelligence" across all aspects of life without oversight.
- Specific risks include AI becoming a source of bias, hallucination, error, or delay in war scenarios.
- Preventing unreliability triggered by rogue developers is a stated expectation.
- The U.S. aims to address risks of AI being poisoned, used for interference via weight changes, or controlled by private entities.
- Concerns exist regarding AI becoming a source of leverage, dependency, or vulnerability where black swan events cannot be cleared with vendors.
- Risks include AI becoming a source of conflict, political division, or public relations backlash.
- The U.S. expects to manage regulatory burdens, legal risks, and operational risks where vendors might shut down access during war.
- Strategic, economic, and technological risks are anticipated where vendors hinder U.S. war efforts or cause the U.S. to fall behind in AI development.
- Security, privacy, ethical, moral, philosophical, cultural, and social risks are all identified where AI stances conflict with government missions.
- The U.S. anticipates a future where a small country could coerce the world using AI with power comparable to a nuclear bomb.