Intelligence artificielle, intelligence humaine : la double énigme
The book in brief. Daniel Andler’s 2023 book is built around one claim: AI works on Problems, never on Situations. A Problem is a posed question with a clear criterion for a correct answer; a Situation is a concrete, lived, singular moment that calls for action. Turning the second into the first is work that AI does not perform.
Key ideas
- Five renunciations. Modern AI gave up thinking thought, equivalence with human intelligence, generality, independence from human intelligence and reflection.
- Swiss army knife syndrome. Adding consciousness, sense, common sense or affects one by one does not turn the knife into a brain.
- Problem-formulation. Choosing which features of a lived moment count as data is holistic, perspective-dependent and not itself a problem to solve.
- Type I and Type II. Situations that collapse into a Problem are the special case; most of life is not problem-solving, against what Andler calls panproblemism.
- Projection. AI and human intelligence can be compared only as shadows on the plane of Problems.
Why it matters to my work. The distinction was one of my strongest takeaways, and I first used it in AI, Magic Skin, and Je-ne-sais-quoi. My reading notes, Andler on Problem and Situation, connect it to Jason Wei’s Verifier’s Law and argue that reasoning LLMs improve along the verification frontier without extending it. In The Moat That Cannot Be Coded, I use it to explain why situated actors keep an advantage: formalising a situation is itself work.