What this covers
Hamkins, a set theorist, anchors this conversation in the distinction between Truth vs proof distinction—the idea that mathematical truth concerns objective reality independent of what we can prove, while proof is our finite, computable interaction with that reality. He weaves this philosophical backbone through a sweeping technical landscape: Gödel's incompleteness theorem and its connection to the halting problem, the nature of transfinite ordinals and Cantor's revolutionary discovery of multiple infinities, the continuum hypothesis and its independence from known axioms, and Conway's surreal numbers. Throughout, he builds toward a central claim: mathematical truth is fundamentally plural. Hilbert's program to axiomatize all of mathematics failed because no finite computable theory can answer every question about itself. This incompleteness is not a gap in our knowledge but evidence of a Set-theoretic multiverse—multiple equally valid mathematical universes where statements like the continuum hypothesis can be true in some and false in others.
The conversation branches into specific technical turns: how forcing lets mathematicians toggle the continuum hypothesis "like a light switch," how the halting problem becomes asymptotically decidable despite being undecidable in principle, and how diagonalization—the Diagonal argument method pioneered by Cantor—underlies almost every major result in logic, from Russell's paradox to the recursion theorem. Hamkins also revisits practical questions like P versus NP through this lens: such questions are fundamentally asymptotic and independent of particular axioms, which shifts how we should interpret their answers. He defends the axiom of choice as a logical principle rather than a distinctly mathematical assumption, and argues that set theory serves two roles—foundational and intrinsic—that mathematicians often conflate. The overall effect is to reframe independence results not as embarrassments but as celebrations: they mark the exact points where mathematical reality branches.
Hamkins argues that mathematical truth is fundamentally plural rather than singular: the independence phenomena pervading set theory (continuum hypothesis, axiom of choice) reveal a multiverse of equally valid set-theoretic universes rather than gaps in our knowledge of one true universe, and that the distinction between truth and proof is the core insight that organizes our understanding of mathematical reality.
- Gödel's incompleteness theorems show no consistent computable theory can answer all questions or prove its own consistency, refuting Hilbert's program
- The continuum hypothesis can be turned 'on and off like a light switch' via forcing, and is independent of all known large cardinal axioms
- Abstract mathematical existence is more clearly understood than physical existence, supporting mathematical realism
Cantor's discoveries of transfinite sets and diagonalization revealed infinity's internal structure and multiplicity.
- Cantor proved that the set of real numbers is an uncountable infinity that is strictly larger than the infinity of natural numbers, establishing that there is more than one size of infinity, which broke and then rebuilt mathematics.
“that's the profound achievement that Cantor made is proving that the set of real numbers is not a countable infinity. It's a strictly larger infinity, and therefore there's more than one concept of infinity, more than one size of infinity.”
- The most beautiful idea in mathematics is the transfinite ordinals invented by Cantor: counting beyond infinity by always being able to add one (omega, omega+1, ... omega+omega, ... omega-squared as the first limit of limit ordinals), which never ends and forms the foundation for transfinite recursive constructions throughout set theory.
“The most beautiful idea in mathematics is the transfinite ordinals... just the idea of counting beyond infinity.”
- Cantor proved the general fact that for any set whatsoever, the power set (the set of all its subsets) is strictly larger than the set itself, demonstrated by the diagonal set of elements not in their associated subset, which cannot be associated with any element without contradiction.
“Cantor actually proved a much more general fact, namely that for any set whatsoever, the power set of that set is a strictly larger set.”
- Cantor's diagonalization idea is an extremely fruitful proof method, with almost every major result in mathematical logic—including Russell's paradox, the halting problem, and the recursion theorem—using diagonalization at its core in an abstract way.
“this diagonalization idea has proved to be an extremely fruitful proof method, and almost every major result in mathematical logic is using in an abstract way the idea of diagonalization.”
- By assuming the real numbers can be listed and constructing a number Z whose Nth digit differs from the Nth digit of the Nth listed number (avoiding zeros and nines to dodge the dual-representation problem), Cantor produces a real number not on the list, contradicting the assumption and proving the reals are uncountable.
“the Nth digit of Z is different from the Nth digit of the nth number. But now it follows that Z is not on the list”
- Hilbert's Hotel demonstrates that a countably infinite set can accommodate additional elements without becoming larger, because a one-to-one correspondence still exists; this violates Euclid's principle that the whole is always greater than the part.
“It's a property of infinity that sometimes when you add an element to a set, it doesn't get larger. That's what this example shows.”
- Russell's paradox is better understood as Russell's theorem—there is no universal set of all sets—because if such a set existed, one could form the set of all sets not members of themselves, which would be a member of itself if and only if it is not, a contradiction following exactly the same logic as Cantor's power set argument.
“I don't call it the Russell paradox. Actually, when I teach it, I call it Russell's theorem. There's no universal set.”
- The theorem that not everyone can be more pointed-at than pointing fails for infinitely many people; with countably many people each holding one dollar, you can arrange a pattern of passing dollars (mapping the Nth train car to the Nth hotel room) so that afterward everyone has infinitely many dollars while each paid out only one.
“if there were infinitely many people in your group? Then it's not true anymore. The theorem fails... afterwards everyone has infinitely many dollar bills.”
Independence from ZFC reveals genuine mathematical reality's plural nature rather than foundational failure.
- When a question turns out to be independent of ZFC, you have asked exactly the right question because you have found a cleavage in mathematical reality—worlds where it is true and worlds where it is false—which is to be celebrated rather than lamented as a failure.
“when you ask a question that turns out to be independent, then you asked exactly the right question because this is the one... It's carving nature at its joints.”
- Paul Cohen's method of forcing (1963) provides a powerful tool for building alternative mathematical realities: from any model of set theory one can construct a forcing extension in which the continuum hypothesis is false, complementing Gödel's 1938 constructible universe where CH is true, together establishing CH's independence.
“Paul Cohen invented the method of forcing and proved that if there's a model of set theory, then there's a model of set theory in which the continuum hypothesis is false.”
- The continuum hypothesis is independent of all known large cardinal axioms—none of them can settle it—even though large cardinal axioms do resolve the CH for progressively more complex hierarchies of definable sets (Borel, projective) given enough large cardinals.
“it's also known that the continuum hypothesis is independent of all of the known large cardinal axioms. So none of the large cardinal axioms can settle the continuum hypothesis.”
- It does not make sense to say a statement like P vs NP is independent all by itself; every statement is independent over a sufficiently weak theory, so independence is always relative to a specified theory.
“every statement is going to be independent over an extremely weak theory. So that's, you know, it doesn't make sense to say it's independent all by itself. You're only independent relative to a theory”
- The developments of set theory over the past half-century are evidence that there is no unique set-theoretic reality; rather there are multiple alternative set-theoretic universes in which fundamental truths differ, so set-theoretic truth has a fundamentally plural character with no singular meaning to fundamental terms.
“what we've been doing for decades now is producing more and more alternative set-theoretic universes in which the fundamental truths differ from one to the other.”
Set theory serves distinct foundational and mathematical roles, grounding both abstract existence and recursive transfinite constructions.
- Set theory serves two distinct roles that are often conflated: it is its own mathematical subject concerned with transfinite recursive constructions, and it separately serves as a foundation of mathematics by providing a way to think of a collection of things as one abstract thing.
“Set theory really has two roles that it's serving... set theory is its own subject of mathematics, with its own problems... But set theory has also happened to serve in this other foundational role.”
- The main goals of logicism are essentially completely fulfilled by the rise of set-theoretic foundationalism, because the principles of ZFC—including the axiom of choice and even the axiom of infinity—are fundamentally logical in character, so set theory's foundational success means mathematics can be founded on logic.
“we should view the main goals of logicism are basically completely fulfilled in the rise of set-theoretic foundationalism.”
- A robust rigorous foundation is apparently not required to make enduring mathematical insights, as shown by Newton and Leibniz proving all the fundamental theorems of calculus using infinitesimals that were not rigorously founded until non-standard analysis in the 1950s; thus the choice of philosophical foundation does not directly drive mathematical progress.
“do you need a robust, rigorous foundation of mathematics to make enduring insights in mathematics? And the answer, regrettably, is apparently not”
- We have no satisfactory account of what it means to exist physically—the more physics we learn the more mysterious physical existence becomes (billiard balls to atoms to quarks to wave functions)—whereas the nature of abstract mathematical existence becomes only clearer the more we say about it, so abstract existence is better understood than physical existence.
“I don't think we have such a clear existence of the nature of physical objects, actually... It is a profound mystery. In fact, it becomes more and more mysterious the more physics we know.”
- A major cultural difference between mathematics and philosophy is the idea of progress: mathematics shows continual growth in understanding (we understand infinity better than 100 years ago), with fields moving on to harder questions, whereas philosophy retains eternal questions for thousands of years, with some philosophers holding the field's contribution is asking questions rather than answering them.
“mathematics has huge progress. We simply understand the mathematical ideas much, much better... Whereas in philosophy... there are these eternal questions that have been with us for thousands of years”
Gödel and Turing's results definitively refute Hilbert's program and establish fundamental limits on axiomatizable knowledge.
- Had Hilbert been right, the nature of mathematical investigation would be reduced to rote computation—turning the crank of a theorem-enumeration machine that eventually outputs every theorem—devoid of creative thinking or imagination, so Hilbert was in effect claiming the fundamental nature of mathematics is rote procedure.
“the nature of mathematical investigation in Hilbert's world is one of just turning the crank of the theorem enumeration machine. Devoid of creative thinking or imagination, it's just getting the answer by rote procedure.”
- Gödel's incompleteness theorem follows immediately from the undecidability of the halting problem without needing the Gödel sentence: a complete computable axiomatization of elementary mathematics would let a theorem-enumeration machine solve the halting problem by waiting for either the halt or non-halt statement, which is impossible.
“I view this as the simplest proof of Godel's theorem. You don't need the Godel sentence to prove Godel's theorem. You can do it with the halting problem.”
- There is no computable procedure that correctly answers, for all programs, whether the program will halt; one can always get the yes answers by running the program, but giving the no answers requires deep insight into the program's behavior, and a diagonal argument shows a contradiction arises if a halting decider existed.
“The halting problem is computably undecidable. There is no computable procedure that will correctly answer all instances of whether a given program will halt.”
- Gödel's two incompleteness theorems decisively refute Hilbert's program: the first shows no computably axiomatizable theory containing enough arithmetic can answer all questions (every such consistent theory is incomplete), and the second shows no such theory can prove its own consistency.
“the first incompleteness theorem says you cannot write down a computably axiomatizable theory that answers all the questions. Every such theory will be incomplete, assuming it includes a certain amount of arithmetic. And secondly, no such theory can ever prove its own consistency.”
Truth and proof are fundamentally distinct concepts distinguishing objective reality from justified assertion.
- The most beautiful idea in philosophy is the distinction between truth and proof: truth is about objective reality and the way things are apart from our knowledge, while proof is about our interaction with that reality and how we come to know what we know—a fundamental philosophical distinction even though born in mathematical logic.
“The truth is about the nature of the world and the way things are. It's about objective reality in a sense, whereas proof is about our understanding of the world and about how we come to know the things that we know about the world.”
- Truth (on the semantic side, concerning whether statements hold in a mathematical structure, formalized by Tarski's disquotational theory) is fundamentally distinct from proof (on the syntactic side, a finite computably-checkable sequence of statements following logical rules), a distinction that was sloppily conflated before Gödel and Tarski but is now core to mathematical logic.
“this is a really core distinction... they were totally sloppy about this distinction between truth and proof. It wasn't clear at all until Gödel, basically.”
Philosophical perspective in set theory determines which research questions and programs mathematicians pursue without changing core theorems.
- The choice of philosophical perspective in set theory does not change the mathematical theorems but determines which questions and research programs are pursued: a universe view drives efforts like Woodin's ultimate L to find the one true universe, while a pluralist view leads to studying interactions between universes, such as set-theoretic potentialism and geology.
“the choice of the philosophical perspective doesn't actually have to do with the mathematical developments directly at all. Rather, it tells us, 'Where should set theory go? What kind of questions should we be asking?'”
Axiom of choice, rigor's role, and proof decidability raise deep logical and foundational questions mathematics addresses.
- Structuralism holds that what matters about mathematical objects is not their substance or essence but how they function in a mathematical structure; we should only care about structures up to isomorphism, so questions about what the number four 'really is' (or whether Julius Caesar is a number) are irrelevant to mathematics because any isomorphic copy works equally well.
“what's important about mathematical objects is not what they're made out of or what their substance or essence is, but rather how they function in a mathematical structure.”
- P versus NP is fundamentally an asymptotic question about behavior as input size goes to infinity, not about practical problems; polynomial-time algorithms can have enormous coefficients and high degrees, so P=NP would not necessarily bring practical benefits, and we already have approximation algorithms (like SAT solvers) that solve almost all real engineering instances.
“P and P equal NP or P not equal NP are not about these practical things at all because of the asymptotic nature of the question itself.”
- The axiom of choice should be regarded as a principle of logic rather than a distinctly mathematical assumption: if you have a collection of nonempty sets, there is a function selecting an element from each, even when no explicit rule can specify the choices, as in choosing one sock from each of infinitely many indistinguishable pairs.
“the axiom of choice principle is completely obvious that we want this to be true, that it is true. A lot of people take it as a law of logic.”
- Current AI language models are a dangerous source of error for mathematical reasoning because they are designed to produce arguments that look like proofs rather than arguments that are proofs; they are grounded in how words appear on a page rather than genuine mathematical understanding, and Hamkins finds them basically zero help for mathematical questions.
“the AI is trying to give me an argument that sounds like a proof rather than an argument that is a proof. The motivation is misplaced.”
- Beyond the commonly cited soundness and completeness, proof systems require a hidden third property: it must be computably decidable whether a given sequence of statements is a proof, so that all claims to having a proof can be correctly adjudicated.
“there's a hidden third adjective that they should always be talking about in any such case, which is that you should be able to recognize whether or not something is a proof or not.”
- The union of countably many countable sets is still countable; using prime factorization (3 to the C times 5 to the S), the passengers of an infinite train with infinitely many cars each with infinitely many seats can all be assigned unique room numbers.
“We've proved that if you have countably many countable sets, then the union of those sets, putting all those sets together into one giant set, is still countable.”
- There is no boring natural number, because if there were uninteresting numbers there would be a smallest uninteresting number, but being the smallest uninteresting number is itself an interesting property, which is a contradiction.
“if there was an uninteresting number- … then there would have to be a smallest uninteresting number. But that's a contradiction, because the smallest uninteresting number is a super interesting property to have.”
- A powerful proof method is to prove something true in the ground model by first constructing a forcing extension, using its features to realize what must already be true, then discarding the extension—analogous to early mathematicians solving real algebraic equations by passing through imaginary square roots of negative numbers that cancel out to give correct real answers.
“You start in set theory and you go to this land of nonsense in the forcing extension, this imaginary world. And you argue and you come back.”
- A theory proving its own consistency gives no real reason to believe it is consistent, because even an inconsistent theory would prove its own consistency—like a used car salesman telling you he is trustworthy, the statement provides no logical grounds for trust.
“even an inconsistent theory would prove its own consistency. And so it doesn't seem to be a logical reason to believe in the consistency, if you have a theory that proves itself consistent.”
Computability theory shows that deep understanding requires running systems not merely examining program structure.
- The main lesson of computability theory, made robust by Rice's Theorem, is that you can never have a thorough understanding of a program's behavior by examining the program; in the most general case the content of what you learn is obtained only by running it and observing the behavior.
“it's never the case that you can have a thorough understanding of the behavior of a program by looking at the program... is always obtained just by running it and looking at the behavior.”
- Although the halting problem is undecidable, it has a 'black hole': there is a computable procedure that correctly decides almost every instance, because the asymptotic density of programs whose head falls off the tape (via random-walk behavior, by Pólya's recurrence theorem) before repeating a state goes to 100% as the number of states grows.
“there's a computer procedure that decides correctly almost every instance of the halting problem. Even though the halting problem is not decidable, we can decide almost every instance.”
- Conway's Game of Life is a playground for computably undecidable questions: the question of whether a given cell will ever become alive in the evolution of a configuration is equivalent to the halting problem and is computably undecidable, though semi-decidable (if it becomes alive you find out at a finite stage).
“you can prove that the question of whether a given cell will ever become alive is computably undecidable.”
Concrete mathematical play like infinite chess and surreal numbers reveals unexpected transfinite structures and unified number systems.
- In infinite chess (played on an integer board infinite in all directions with no promotion and infinite play counting as a draw), positions can have transfinite game values; white can have a winning strategy that wins in infinitely many moves with no fixed N (game value omega), and it is now known that every countable ordinal arises as the game value of some position.
“we know now that every countable ordinal arises as the game value of a position in infinite chess”
- Conway's surreal number system unifies all other number systems—integers, rationals, reals, ordinals, and infinitesimals all sit inside it—generated from nothing by a single rule of dividing existing numbers into a left and right set and creating a new number in the gap, forming a real closed field.
“it's providing us a number system that unifies all the other number systems. So it extends the real numbers... but also the ordinals and the infinitesimals.”
- The surreal numbers are fundamentally discontinuous: no non-trivial set of surreal numbers has a least upper bound and there are no convergent sequences, so calculus based on limits and convergence does not work, though calculus can still be done via non-standard infinitesimal methods.
“they don't have the least upper bound property for subcollections... and there are no convergent sequences in the surreal numbers. And so for the sort of ordinary use in calculus based on limits and convergence, that method does not work”
Mathematical practice thrives through social engagement and simple clear arguments over isolated technical difficulty.
- A productive habit in mathematics is to anthropomorphize mathematical objects—imagining them as people or agents with goals—which often makes problems easier to understand and lets one leverage human intuition, as in proving you cannot all be more pointed-at than pointing by imagining everyone pays a dollar to whoever they point at.
“It's one of my habits that I suggest in the book: to anthropomorphize your mathematical ideas... it often makes the problems easier to understand”
- Hamkins's mathematical style favors simple, clear, easy-to-understand arguments that prove surprising results over difficult mathematics, partly because he is naturally skeptical of complicated arguments that he cannot hold entirely in his head, which might be wrong—a preference that draws him off the beaten path.
“What I love is simple, clear, easy-to-understand arguments that prove a surprising result... I'm kind of naturally skeptical about any complicated argument because it might be wrong.”
- For Hamkins, mathematics is often a social activity; putting forth mathematical ideas to other people who respond to them is not merely pleasant but is the nature of mathematical investigation, productive because it helps everyone learn—contrasting with the isolated multi-year grind of someone like Andrew Wiles.
“that's the nature of mathematical investigation as I see it, is putting forth mathematical ideas to other people and they respond to it in a way that helps me learn, helps them learn”
Mathematical progress and knowledge growth involve luck and simultaneous discovery rather than singular achievement.
- It is relatively common for disparate mathematicians to prove essentially similar results at approximately the same time because certain ideas are in the air and being thought about but not yet fully articulated; this is the nature of growth in knowledge, and credit going to whoever was first involves luck.
“certain ideas are in the air and being thought about but not fully articulated, and so this is the nature of growth in knowledge.”
- Because mathematics genuinely progresses and we understand things much better now, some achievements of the greats were things that someone else could easily have done first, so there is a luck aspect to who gets credit; Hamkins prefers to gain insight from whoever provides it rather than ranking mathematicians.
“Sometimes the greats are doing things that are just first and not... You know, somebody else could have easily been first. So there's a kind of luck aspect to it”