Foresight Institute

Ramez Naam | Contrarian Views on the State of AI @ Vision Weekend Puerto Rico 2026

Ramez Naam opened his Vision Weekend Puerto Rico 2026 talk by rejecting the prevailing hype that artificial general intelligence and a rapid super‑intelligence take‑off are imminent. He argued that the dominant narrative—zero‑sum competition between the U.S. and China and a winner‑takes‑all race to AGI—is fundamentally wrong. He highlighted the staggering economic figures: AI‑related revenue is projected to reach $1.3 trillion by 2032, and U.S. AI capex this year will hit roughly $600 billion, about 2 % of GDP. Yet he warned that scaling compute alone yields diminishing returns, with each linear performance gain demanding exponential increases in hardware and energy. The real driver, he said, will be algorithmic breakthroughs and data efficiency, not merely larger models. Naam cited concrete examples to illustrate his point. The price of frontier models fell by a factor of 300 in a single year, and OpenAI’s early lead in benchmark rankings lasted only 18 months before competitors caught up within weeks. He contrasted a human brain’s 20‑watt power draw with a top‑tier LLM’s 20‑kilowatt inference consumption, underscoring the unsustainable trajectory of pure compute scaling. The implication for investors, policymakers and technologists is clear: AI is becoming a hyper‑democratized, multi‑polar field where no single entity can dominate. Safety strategies must shift from trying to lock down individual models to designing robust ecosystems and governance frameworks. Stakeholders should prioritize funding algorithmic research and data‑centric approaches rather than betting solely on ever‑larger compute budgets.