Marc Andreessen: Intelligence Is Only 0.4 Correlated With Success

Why Marc Andreessen Thinks AI Won’t Rule the World

Marc Andreessen and Ben Horowitz rarely do public conversations together, making this AI Engineer fireside chat particularly valuable. They tackle the biggest philosophical questions around AI—creativity, intelligence, the bubble thesis—with characteristic contrarianism and surprising optimism.

On AI clearing the bar: “If these things clear the bar of 99.99% of humanity, that’s pretty interesting in and of itself… how many actual real conceptual breakthroughs have there ever been actually ever in human history as compared to sort of remixing ideas?” The argument: even Beethoven was remixing Mozart and Haydn. If AI can match everyone but the handful of generational geniuses, that’s already transformative.

On intelligence not being sufficient: “Intelligence is 0.4 correlated to basically everything—educational outcomes, professional outcomes, income, life satisfaction, non-violence. But that still leaves 0.6 unexplained.” The social science finding that even raw IQ only explains 40% of outcomes undermines the “superintelligence will rule” thesis.

On theory of mind limits: “If the leader is more than one standard deviation of IQ away from the followers, it’s a real problem. And that’s true in both directions.” The US military finding: being too smart to model subordinates’ thinking is as disabling as not being smart enough. A 1000-IQ AI might be so alien it couldn’t effectively lead.

On the bubble question: “The fact that it’s a question means we’re not in a bubble. In order to get to a bubble, everybody has to believe it’s not a bubble.” The psychological dynamic of bubbles requires capitulation—widespread skepticism is itself evidence against bubble formation.

On LLM theory of mind: “It turns out that language models now are good enough—they can correctly and accurately reproduce a focus group of real people inside the model.” Political focus groups can now run inside LLMs with accurate persona modeling, eliminating weeks of logistics for voter research.

5 Insights From a16z on AI, Creativity, and Intelligence

• AI already matches 99.99% of human creativity - The Beethoven bar is extraordinarily high; matching everyone below generational genius is already transformative
• Intelligence is necessary but only 0.4 correlated - Raw IQ explains less than half of success outcomes; courage, theory of mind, and situational judgment matter as much
• Too-smart leaders fail at theory of mind - Military research shows leaders more than 1 SD above followers struggle to model their thinking, suggesting superintelligent AI may have coordination limits
• Skepticism disproves bubble - The psychological mechanics of bubbles require universal capitulation; widespread questioning means we’re not there
• LLMs can simulate accurate focus groups - Political research can now run persona-based voter groups entirely inside models, matching real focus group accuracy

What a16z’s AI Thesis Means for Builders

Andreessen and Horowitz offer a counterintuitive case for AI optimism that doesn’t depend on the “superintelligence takes over” thesis. Their argument: intelligence isn’t sufficient for leadership or success, most human creativity is remixing anyway, and the things that actually matter—theory of mind, situational judgment, physical embodiment—are either already achievable or represent limitations that keep AI as tool rather than ruler. The bubble skepticism is similarly grounded: if everyone’s questioning it, the psychological preconditions for bubble formation haven’t been met.