Eric Schmidt on AI, the Battle with China, and the Future of America

AI Landscape and Strategic Outlook

• Eric Schmidt characterizes the current AI revolution as "underhyped," predicting the imminent arrival of autonomous agents capable of self-directed planning and collaboration.
• Schmidt identifies a new "non-human intelligence" likely to surpass human reasoning capabilities, emphasizing that Western nations must prioritize winning in this competition.
• Schmidt rejects the "San Francisco narrative" that predicts artificial general intelligence (AGI) within three years, arguing that the concept relies on "scale-free" recursive self-improvement for which there is currently no evidence.
• Schmidt defines AGI as the ability of a system to set its own objective function, a capability he claims has not yet been demonstrated in current LLMs.
• He posits that current AI is "middle-to-middle" rather than "end-to-end," requiring human prompting, validation, and iterative loops to solve problems, making immediate replacement of human workers unlikely.
• Schmidt predicts that AGI may be achieved within six to seven years, primarily through solving the "non-stationarity" of mathematical proofs and mastering cross-domain analogy.

US-China Geopolitical Competition

• Schmidt observes that China is pivoting away from "crazy" AGI strategies due to hardware restrictions and capital market limitations, focusing instead on applying AI to consumer apps, robotics, and drones.
• A critical geopolitical divergence is identified: China is aggressively pursuing open weights and open training data models, while the US remains predominantly focused on closed models.
• Schmidt warns that the majority of the "Belt and Road" nations will likely adopt Chinese open-source models, potentially eroding Western influence if the US does not match this strategy.
• He notes that Chinese AI developers are achieving high performance with lower precision (4-bit and 8-bit), contrasting with the US reliance on 16-bit precision (FP16).
• Schmidt supports the US strategy of releasing smaller, open-weight models (like the smallest O3 version from OpenAI) to ensure handheld device proliferation remains under American control.
• He highlights that Chinese robotics companies are successfully applying their manufacturing dominance to the sector, mirroring their success with electric vehicles.

Defense, Warfare, and Robotics

• Schmidt argues that traditional military infrastructure, such as fixed tank formations, is obsolete due to the high cost-efficiency of $5,000 drones destroying $30 million tanks (a significant "kill ratio" advantage).
• He predicts a future military doctrine where drones act as the "rifle" and "artillery," with humans positioned behind the lines to command autonomous systems via networks like Starlink.
• Schmidt envisions a future battlefield where opposing sides deploy millions of drones using reinforcement learning AI strategies, creating a state of high deterrence because neither side can predict the other's AI-generated battle plans.
• He clarifies that this mutual inability to predict AI plans creates deterrence, not detente, as the cost of conflict would result in the total destruction of both sides' drone supply chains and infrastructure.
• Schmidt dismisses the near-term viability of humanoid robots in warfare, forecasting that future conflict will rely on air-based mobility, hypersonics, and unmanned underwater vehicles (UUVs).
• He acknowledges that human soldiers will eventually be required to clear enemy lines after automated forces have depleted opposing defenses.
• Schmidt cites the Ukrainian conflict as proof that a nation without a traditional navy or air force can achieve success through automation and drone innovation while outnumbered.

Capital Markets, Industry, and Societal Trends

• Schmidt discusses the return to in-office work, arguing that early-career professionals require in-person exposure to senior leadership to learn organizational dynamics effectively.
• He dismisses the "996" work schedule (9 AM to 9 PM, six days a week) as illegal and unsustainable in China, yet notes it remains a competitive reality they are forced to match.
• Regarding Relativity Space, Schmidt describes rocketry as a "hard" art and science where 80% of weight is propellant, 18% is structure, and only 2% is payload.
• He mentions Relativity Space has a full order book for Low Earth Orbit (LEO) satellite launches and aims to compete with SpaceX despite the technical difficulty.
• Schmidt identifies declining birth rates in the West (and even more severely in Asia, with South Korea at 0.78) as a critical business and societal risk that necessitates immigration or technological innovation.
• He views the decline in Western population as a "revenue decline" scenario for the global economy, as shrinking, aging populations reduce the capacity for innovation and market growth.
• Schmidt advocates for investing in the US "chaotic" strengths, including deep capital markets, a robust industrial base, and a university system, to maintain American exceptionalism.

Corporate and Technical Developments

• Schmidt acknowledges Meta's missteps in open-source execution but believes Alex Wang will steer the company toward a more balanced approach rather than a fully closed model.
• He notes that DeepSeek has achieved significant progress in reinforcement learning and forward reasoning with lower precision numeric formats than US counterparts.
• Schmidt confirms that Sam Altman plans to release an open-weight version of the O3 model, which is significantly smaller than the 10^26 parameter class and fits on mobile devices.
• He states that Google's Gemini 2.5 has recently overcome competitors to lead the current rankings, with Demis Hassabis focusing heavily on scientific discovery as a path to AGI.
• Schmidt served as a top-secret clearance advisor to the Pentagon, helping reorganize the department prior to the Ukraine war, which he views as a "spectacular success" in innovation.