A Mind-Blowing Way of Looking at Math (with David Bessis)
What this covers
David Bessis, a mathematician and author, joins Russ Roberts to argue that mathematics is fundamentally a practice of training intuition rather than manipulating logic. The conversation charts how great mathematicians work: they build and rebuild their sense of what should be true through repeated cycles of guessing, testing against logical formalism, discovering error, and retraining their intuition. Bessis traces this pattern across history—from René Descartes to Alexander Grothendieck—showing that mathematicians themselves consistently denied being innately gifted, instead describing a learnable method. Andrew Wiles's correction of his Fermat's Last Theorem proof and Ramanujan's dream-revealed theorems anchor the discussion of how intuition and proof interact in the real practice of mathematics.
The conversation ranges across education, psychology, and cognition. Bessis introduces System 3—a third mode of thought beyond Kahneman's System 1 and System 2—that describes the super-slow, iterative retraining of intuition that characterizes mathematical work. He argues that mathematical ability is developmentally acquired, like swimming, and that schools wrongly write children off as "not math people" when they would never apply such verdicts to language learning. The pair examine how intuition operates non-visually even for those with aphantasia, why math books should not be read sequentially, and how asking "stupid" questions—as Jean-Pierre Serre famously does—unlocks genuine understanding. They also touch on Language as Low-Bandwidth Tool, arguing that rationality functions as a technique to consolidate intuition rather than replace it, and briefly address how AI's arrival will collide with mathematics itself.
Bessis argues that the real engine of mathematics is not logic but intuition, and that great mathematicians achieve their power by repeatedly retraining their malleable intuition against formal logic—a third mode of thought ('System 3') that anyone can practice and that constitutes a learnable technique for becoming smarter rather than a fixed innate gift.
- Logical formalism is a low-bandwidth validation device used to correct intuition, not the source of mathematical meaning
- Math books, like the best mathematicians, reveal that intuition is built through back-and-forth iteration, asking 'stupid' questions, and not reading proofs linearly
- Mathematical ability is developed over time like a sport, not an innate IQ trait, with implications for education and rationality
Mathematical reasoning is the only tool allowing rationality to venture far from concrete experience because it is engineered for logic.
- All reasoning, even the most solid, comes apart the further it gets from day-to-day experience—not from lack of rigor but because language itself rests on a base of sand and mud; the sole exception is mathematical reasoning in the official formal language, which is engineered to be compatible with logic and is therefore the only tool that lets rationality venture far from concrete experience.
“All reasoning, even the most solid, ends up coming apart the further it gets from day-to-day experience. Not for lack of rigor, but because our language itself is built on a base of sand and mud.”
We wrongly apply fixed-ability verdicts to mathematics while treating language learning as developmentally open-ended.
- We never conclude that a babbling two-year-old will never learn to speak and so stop teaching them language, yet we routinely write children off as 'not math people' who are 'not good at math,' applying a fixed-ability verdict to mathematics that we would never apply to language acquisition.
“Just because a two-year-old is babbling, you don't say: Well, I guess he'll never learn to speak... And yet, we do that with math. We say, 'Well, he's not a math person. He's not good at math.'”
Language is inherently vague and ambiguous, as evidenced by repeated translation failures even among skilled translators.
- Repeated failures to produce 'good' translations between languages, even by skilled human translators, indicate a genuine non-transformation between languages—each may carry its own unique syntax that does not map cleanly onto another—illustrating that language is inherently vague and ambiguous.
“When it happens over and over again, you realize there's a certain non-transformation between two different languages.”
The speaker found Bessis's book electrifying and aims to encourage listeners to read it.
- Roberts states the goal of the interview is to get listeners to read Bessis's book, which he found electrifying and read twice.
“the goal of this interview is to get you to read this book. I've read it twice.”