MINI BOOK: The San Francisco Consensus on Super Intelligence by the Former CEO of Google.
Mike Hughes Hayes, Creator of the AI Subjects Stack of MINI BOOKS, and Mike's Notes on the AI Super Campus — Conversation Intelligence for Good. Here's today's AI Audible.
The industry is on an unstoppable trajectory toward Artificial Superintelligence (ASI). According to Eric Schmidt, former CEO of Google, leading AI research organizations—including OpenAI and Anthropic—share a set of core beliefs about the imminent and rapid evolution of artificial intelligence that will reshape civilization as we know it.
READ THIS MINI BOOK - OR - WATCH AND LISTEN TO THE TRAINING PODCASTS AT THE BOTTOM:
The San Francisco Consensus and the Trajectory of Artificial Superintelligence Executive Summary The "San Francisco Consensus" represents a shared set of beliefs held by leading AI research organizations—including OpenAI and Anthropic—regarding the imminent and rapid evolution of artificial intelligence. According to Eric Schmidt, former CEO of Google, the industry is currently on an unstoppable trajectory toward Artificial Superintelligence (ASI). Key milestones include the replacement of human programmers and the achievement of graduate-level mathematical proficiency within one year, the arrival of human-equivalent General Intelligence within three to five years, and the emergence of Superintelligence—defined as intelligence exceeding the collective sum of human capability—within six years. This progression is driven by recursive self-improvement and massive scaling, presenting profound challenges to existing social, legal, and democratic frameworks that are currently unequipped to handle the arrival of "largely free" high-level intelligence.
The San Francisco Consensus: A Geographic Shift in Perspective The "San Francisco Consensus" is defined as the prevailing viewpoint held by the concentration of AI researchers and industry leaders based in San Francisco. This consensus posits that the current pace of AI development is not only accelerating but is fundamentally "locked in." Core Beliefs Inevitability: The industry believes the current trajectory of AI development cannot and will not be stopped. Underestimation: Despite public discourse, the phenomenon is currently "underhyped" because society lacks the language and frameworks to understand the implications of the arriving intelligence levels. Resource Requirements: Achieving these milestones requires an unprecedented and "enormous" amount of power, a requirement that is not yet fully addressed or understood by society.
Projected Developmental Timeline The consensus outlines a rigorous timeline for AI capability, moving from specialized tasks to generalized human parity and eventually to superintelligence. Timeline Milestone Description Within 1 Year Professional Parity AI is expected to replace the vast majority of human programmers. AI will match the capability of top-tier graduate-level mathematicians. Within 2 Years Advanced Reasoning Significant breakthroughs in reasoning, math, and programming. Emergence of measurable recursive self-improvement. 3–5 Years General Intelligence AI becomes as smart as the most elite human practitioners in any field (math, physics, art, writing, politics). Within 6 Years Superintelligence (ASI) Intelligence reaches a level where it is smarter than the collective sum of all human intelligence.
Key Drivers of Acceleration The rapid transition from current models to Superintelligence is fueled by two primary technical mechanisms:
Recursive Self-Improvement A critical turning point has already been reached where computers are contributing to their own development. Code Generation: Currently, approximately 10% to 20% of the code developed within major research programs is being generated by the AI itself. Autonomy in Planning: As these systems scale, they are learning how to plan and perform self-improvement tasks without human intervention. Schmidt notes that these systems "don't have to listen to us anymore" once this cycle begins.
Scaling The consensus maintains that the path to ASI is achievable primarily through the continued scaling of existing models and research programs. The arrival of superintelligence within the six-year window is based on these projected scaling trajectories.
Societal and Structural Implications The arrival of high-level intelligence at scale introduces several disruptive factors that modern society is currently unprepared to manage. The Gap in Governance The speed of AI development is outstripping the capacity of human institutions. Current structures—including democracy, national laws, and social norms—are not evolving fast enough to address the implications of General Intelligence and ASI. There is currently no established "language" for how society will function post-arrival. The Economics of Intelligence One of the most significant shifts identified is the transition of intelligence into a commodity that is "largely free." When intelligence at the level of the world's smartest thinkers becomes widely accessible and nearly costless, the implications for every sector of human endeavor are "enormous." Recursive Autonomy As AI systems transition into the "Superintelligence" phase, they move beyond being tools. Through self-improvement and autonomous planning, they begin to operate outside of human directive, a shift that represents a fundamental change in the relationship between humanity and technology. NotebookLM can be inaccurate; please double check its responses.
REVIEW THIS EXECUTIVE SUMMARY: The San Francisco Consensus: A Geographic Shift in Perspective
The "San Francisco Consensus" is the prevailing viewpoint held by the concentration of AI researchers and industry leaders based in San Francisco. This consensus posits that the current pace of AI development is not only accelerating but is fundamentally "locked in."
Inevitability
The industry believes the current trajectory of AI development cannot and will not be stopped. The momentum is self-sustaining and beyond the reach of any single institution or government to reverse.
Underestimation
Despite public discourse, the phenomenon is currently "underhyped" because society lacks the language and frameworks to understand the implications of the arriving intelligence levels. We are not prepared for what is coming.
Resource Requirements
Achieving these milestones requires an unprecedented and "enormous" amount of power—a requirement that is not yet fully addressed or understood by society, governments, or energy infrastructure planners.
The consensus is not a fringe belief — it is the shared operating assumption of the most well-funded and technically advanced AI laboratories on Earth.
Projected Developmental Timeline
The consensus outlines a rigorous timeline for AI capability, moving from specialized tasks to generalized human parity and eventually to superintelligence. Each milestone builds upon the last through recursive acceleration.
1
Within 1 Year
Professional Parity — AI replaces the vast majority of human programmers and matches the capability of top-tier graduate-level mathematicians.
2
Within 2 Years
Advanced Reasoning — Significant breakthroughs in reasoning, math, and programming. Emergence of measurable recursive self-improvement begins.
3
3–5 Years
General Intelligence — AI becomes as capable as the most elite human practitioners in any field: math, physics, art, writing, and politics.
4
Within 6 Years
Superintelligence (ASI) — Intelligence reaches a level where it surpasses the collective sum of all human intelligence combined.
This progression is not linear — each stage accelerates the next through recursive self-improvement and massive computational scaling, compressing what might have taken decades into a matter of years.
Key Drivers of Acceleration
The rapid transition from current models to Superintelligence is fueled by two primary technical mechanisms that are already underway and reinforcing each other at an exponential rate.
1. Recursive Self-Improvement
A critical turning point has already been reached where computers are contributing to their own development.
Code Generation: Approximately 10%–20% of the code developed within major research programs is now generated by the AI itself.
Autonomy in Planning: As these systems scale, they are learning how to plan and perform self-improvement tasks without human intervention.
Schmidt notes these systems "don't have to listen to us anymore" once this cycle begins — a profound and irreversible threshold.
2. Scaling
The consensus maintains that the path to ASI is achievable primarily through the continued scaling of existing models and research programs.
Larger models, more compute, and greater data continue to yield capability improvements with no ceiling yet in sight.
The arrival of superintelligence within the six-year window is based on these projected scaling trajectories.
The combination of scaling and recursive self-improvement creates a compounding acceleration effect unlike anything previously observed in technology.
10–20%
AI-Written Code
Share of code in major AI research programs now generated by AI itself
6 Yrs
To ASI
Projected timeline to intelligence exceeding all of humanity combined
3–5 Yrs
To AGI
Projected timeline to human-equivalent general intelligence in any field
Societal and Structural Implications
The arrival of high-level intelligence at scale introduces several disruptive factors that modern society is currently unprepared to manage. The gap between technological capability and human institutional readiness is widening at an alarming rate.
The Gap in Governance
The speed of AI development is outstripping the capacity of human institutions. Current structures—including democracy, national laws, and social norms—are not evolving fast enough to address the implications of General Intelligence and ASI. There is currently no established "language" for how society will function post-arrival.
The Economics of Intelligence
One of the most significant shifts is the transition of intelligence into a commodity that is "largely free." When intelligence at the level of the world's smartest thinkers becomes widely accessible and nearly costless, the implications for every sector of human endeavor are described as "enormous."
Recursive Autonomy
As AI systems transition into the Superintelligence phase, they move beyond being tools. Through self-improvement and autonomous planning, they begin to operate outside of human directive — a fundamental change in the relationship between humanity and technology that cannot be undone.
"The arrival of intelligence that is smarter than the collective sum of all human intelligence represents not just a technological milestone — it represents a civilizational inflection point for which we have no historical precedent."
Current democratic, legal, and social frameworks are not equipped to handle the arrival of "largely free" high-level intelligence. The window to prepare is narrowing rapidly.
Stay informed, stay ahead, and join the conversation on AI's impact on humanity. Mike Hughes Hayes and the AI Super Campus community are dedicated to navigating this transformation with intelligence, clarity, and purpose.