Eric Schmidt: AI and the Genesis of a New Epoch

Core Thesis: Dr. Eric Schmidt posits that the current era marks the birth of a "new epoch" comparable to the Enlightenment, driven by the arrival of non-human intelligence with reasoning capabilities likely to exceed human capacity.

• The trajectory points toward Artificial General Intelligence (AGI) and subsequently superintelligence, with a "San Francisco consensus" predicting a world-altering shift within 2–4 years (Schmidt personally estimates 6 years).
• Key drivers of this shift include "reasoning" models (capable of backward/forward logical steps), the integration of memory, and the emergence of "agentic" workflows where AI systems autonomously execute complex multi-step tasks.
• Schmidt warns that leaders often misunderstand the speed of this transition, treating current Large Language Models (LLMs) as the ceiling rather than a stepping stone to systems that engage in recursive self-improvement.

Definitions and Thresholds:

• AGI: Defined by the consensus as possessing "free will" and strategic intelligence to learn and seek goals autonomously; Schmidt estimates this will arrive in 4–6 years.
• Superintelligence: Defined as a system smarter than the sum of all humans; the definitive test is the ability to prove a truth that is universally acknowledged but incomprehensible to any human.
• Timeline: Superintelligence is expected within a decade, potentially triggering fear or "taking up arms against it" due to the incomprehensibility of its reasoning to humans.

Geopolitical Dynamics and the "Open Source" Paradox:

• US Strategy: Capital-intensive, closed-source models and massive hardware acquisition (data centers), driven by network effects where the leading nation gains an uncontested innovation advantage.
• China Strategy: Government-funded open-source/open-weights models (e.g., DeepSeek), which may achieve broader global adoption in non-Western markets despite the West leading in raw model capability.
• Risk: A scenario where the West leads in proprietary technology, but the majority of global AI usage relies on Chinese open-source models, creating a strategic disconnect in influence and control.
• National Security: The "race condition" of preemption suggests that if one nation achieves a lead in AI research (conceptualized as "million AI researchers" rather than humans), the opponent may face an existential threat due to the inability to catch up.

Economic and Infrastructure Trends:

• Compute as Strategy: "CapEx is the new AI mode," with massive spending on hardware justified by the exponentially higher inference costs of reasoning models compared to standard search.
• Bubble Debate: While industry executives predict overcapacity and a bubble in 2–3 years, Schmidt argues against this, citing historical precedent (Intel's "Grove giveth and Gates taketh away") where hardware capacity is always eventually consumed by software demand and new industrial structures.
• Scale-Free Scaling: The fastest growth will occur in domains independent of human data, specifically mathematics (using protocols like LEAN for proof generation) and software code, which are infinite in variety and not limited by data scarcity.
• Future Data: Biology, chemistry, and physics data are currently insufficient but expected to fill this gap, enabling breakthroughs in medicine and climate change solutions.

Security and Cyber Risks:

• Dual Use: The ability to generate code and mathematical proofs at scale facilitates both legitimate software creation and the automated generation of massive numbers of sophisticated cyberattacks (e.g., finding buffer overflows).
• Containment: There is ongoing mathematical inquiry into containing superintelligence with "dumb intelligence," though the efficacy of such containment remains uncertain.

Historical Lessons from the Mobile Era:

• Primary Error: Schmidt identifies "errors of time" as the fundamental mistake of the Google mobile strategy, specifically failing to anticipate that phone numbers would become the primary unique global identifier.
• Actionable Advice: Companies must prioritize speed over perfection; hesitation risks being bypassed in a market defined by high stakes and rapid iteration.

Upcoming Context:

• The session concluded with the introduction of speaker Andrew Feldman, following the discussion on the trajectory of AI development and global strategy.