Where this was said
Creativity, Synthesis, and What AI Can't Do
At 35:02 · chapter starts 29:44
Ben Horowitz introduces the idea that current AI is really 'Artificial Human Intelligence' — trained not on reality but on language, a human structure for organizing the world. [1] — Rick Rubin "The AI can't invent flight for the Wright brothers. It can only regurgitate what the Wright brothers did. And we do that not by knowing mor…" 35:02 Andreessen then poses his deepest question: if all human creativity is synthesis — the iPhone built on 20 years of failed smartphones, television built on 40 years of mechanical spinning blocks — is there anything distinctively human about creativity, or will AI eventually match it? Rubin's answer is precise: if AI has reason, it won't do what humans can do, because we are not reasonable. All the genuine breakthroughs — flight, the 4-minute mile, the unthinkable Go move — came from believing in something that can't be, from a kind of productive delusion that AI, optimized for consistency and rationality, cannot replicate. 'The AI can't invent flight for the Wright Brothers. It can only regurgitate what the Wright Brothers did.' The magic of human creativity, Rubin argues, lies exactly in its irrationality.
Ben Horowitz argued current AI is better described as 'Artificial Human Intelligence' because it reflects only the human linguistic structuring of the world, not first-principles observation of reality.
The best creative discoveries happen by accident. Set out in a direction, pay attention to where the magic appears, and follow it — even if it leads somewhere you never planned to go. Being invested in what it's 'supposed to be' is the enemy of finding what it actually is.
Marc Andreessen noted that IBM released something that looks a lot like an iPhone in 1987, suggesting the smartphone was not Steve Jobs' original invention but a 20-year refinement.
AI can't invent flight. It can only regurgitate what the Wright Brothers did. Real breakthroughs happen not by knowing more, but by believing in something that can't be — a quality AI, trained on what is reasonable, cannot replicate.
