146 | Emily Riehl on Topology, Categories, and the Future of Mathematics
What this covers
Emily Riehl and Sean Carroll discuss the philosophical foundations and modern trajectory of pure mathematics, with particular attention to how topology and category theory have shifted mathematical thinking away from the intrinsic properties of objects and toward the relationships between them. Riehl traces this shift through concrete examples: how homotopy—the study of spaces under continuous deformation—leads naturally to algebraic invariants like groups that can distinguish genuinely different spaces; how the fundamental infinity-groupoid encodes the full structure of a space by recording not just paths between points, but paths between paths, and so on; and how the Yoneda Lemma proves that any mathematical object can be completely characterized by its relationships to other objects of the same type, making relational thinking the deepest way to understand mathematics.
The conversation covers a wide conceptual range. Riehl illustrates notions of sameness in topology through examples like homeomorphism versus homotopy equivalence, and explains how homotopy groups measure the obstruction to filling higher-dimensional spheres into a space. She works through concrete cases—robots on tracks (whose configuration space reveals whether they can swap positions), the five Platonic solids (derived from symmetry groups), and the Klein bottle (which cannot embed in three dimensions)—to build toward infinity-groupoids and category theory as the language of structure without essence. Throughout, Riehl argues that pure mathematics justifies itself not through applications but by making the world a richer place to think in, and predicts that foundational mathematics will eventually shift to better accommodate modern work in these abstract structures.
Riehl argues that the value of pure mathematics lies in making the world a more interesting place to think in, and that the deepest modern understanding of mathematical spaces comes from forgetting their intrinsic essences and instead studying the maps and relations between objects — culminating in category theory and infinity-groupoids.
- Topology studies properties invariant under continuous deformation, and algebraic invariants (groups) let us prove spaces are genuinely different
- The fundamental group/infinity-groupoid encodes a space's full homotopy type via paths, paths between paths, and so on
- The Yoneda Lemma shows any mathematical object can be fully characterized by its relationships to other objects of the same type, making relations more fundamental than intrinsic essence
Mathematical objects are defined by their role and relationships within mathematics, not by intrinsic essence.
- A topological space is formally a set of points together with a notion of when points are nearby or far apart; this can be a distance function (a metric space) or a more abstract structure that captures nearness without a distance formula.
“what turns a set into a topological space is also a way to understand distances between points, or more generally understand when points are nearby and when points are far apart”
- Under mathematical structuralism, what a mathematical object really is, is determined by the role it plays within mathematics; for example, the natural numbers are essentially 'the thing you can recursively define functions on' (a universal discrete dynamical system), and recursion being possible tells us something deep about their role.
“what a mathematical object really is, is determined by the role that it plays within mathematics”
- A group fixes a single base point and records only loops (or, equivalently, automorphisms of one fixed object), whereas a groupoid allows multiple objects/base points and records paths between different points; a groupoid is the multi-object analog of a group.
“In a groupoid you have different objects, so there's not just one object anymore, there are different objects, and it's exactly the mini-object analog of a group”
- The Yoneda Lemma, the fundamental theorem of category theory, states that any mathematical object — a space, ring, field, group, etc. — can be completely characterized by considering all the other objects of the same type and the maps between them; this works in any mathematical context whatsoever, making relations rather than intrinsic essence the way to understand objects.
“it's completely independent of the mathematical context. So the same theorem is true for rings... This works in any mathematical context whatsoever.”
- A category is a general template for a mathematical theory consisting of a collection of objects (the 'nouns', e.g. all rings) together with the structure-respecting maps between them (the 'verbs', e.g. ring homomorphisms or continuous functions) and a composition operation, following Barry Mazur's noun/verb metaphor.
“A category is given by a collection of objects, these are the nouns... And then, you also have the sort of functions between them”
Symmetry groups and category theory are foundational organizing principles across all mathematics.
- By studying the symmetries (orthogonal and special orthogonal groups) of hypothetical regular polyhedra, you can prove there are only finitely many — exactly five — Platonic solids, even before knowing they exist, and a duality relationship pairs the cube with the octahedron and the dodecahedron with the icosahedron (with the tetrahedron self-dual), reflected in their isomorphic symmetry groups.
“you can prove in fact that there are only finitely many, in fact, there are five of them using group theory by studying the symmetries, the orthogonal groups, the special orthogonal groups”
- A field, like a ring, has two binary operations, but unlike a ring it requires that every nonzero element have a multiplicative inverse (zero is treated as a special case and never has one); the integers form only a ring because most integers have no multiplicative inverse, whereas the rationals and reals are fields.
“In a field you do have an inverse to both the multiplication and the addition operation, assuming you're not trying to divide by zero”
- The category of vector spaces and linear transformations is equivalent to a much simpler category whose objects are natural numbers and whose arrows are matrices (with matrix multiplication as composition); this category equivalence explains why computational linear algebra (matrices) and theoretical linear algebra (bases, linear independence) are really the same subject despite feeling very different.
“those categories are equivalent. So in other words, you can think of the natural number as a stand-in for vector space, the number five corresponds to five dimensional Euclidean space”
- A group is a set together with a composition operation satisfying natural axioms; in the topological context it is the collection of loops with the operation of performing one loop then another, and every element has an inverse (reversing the loop) that returns you to the starting state.
“every element in a group has an inverse that if you compose with it, it gets back to where you started, and a few simple axioms like that”
- A ring is a setting with two interacting operations — an addition-like operation and a multiplication-like operation that obey familiar laws like distributivity — generalizing the integers; polynomials with real coefficients also form a ring, and there is a deep analogy between the ring of integers and the ring of polynomials (both have a division algorithm and principal ideals).
“a ring is a setting where you have two operations, an addition operation or an addition-like operation, and a multiplication-like operation, and they interact in ways that are familiar for the integers”
- A group can also be understood as an axiomatization of the symmetries of an object; the four ways to flip a mattress form the Klein four-group, in which every flip is order-two and any two distinct flips compose to the third, and crucially this group has no single generator, which is why it is hard to remember how to rotate a mattress correctly each month.
“a group is an axiomatization of the symmetries of an object”
Topology uses algebraic invariants like homotopy and homology groups to distinguish spaces when intuition fails.
- To prove two topological spaces are genuinely different when geometric intuition is insufficient, one assigns algebraic structures (homology or homotopy groups) to spaces in a way that yields identical answers for homotopy-equivalent spaces; therefore if the assigned invariants differ, the spaces must be different.
“a strategy is to bring algebra into the story somehow and give a way to assign a number or some other kind of algebraic structure to a topological space, and in a way that would give the same answer if the spaces were homotopy equivalent”
- The fundamental infinity-groupoid of a space forgets the topology and remembers the set of points, every path between points, every homotopy (path between paths), every path between paths between paths, and so on to infinity; this algebraic structure captures the full homotopy type of the space and also invites generalization to settings where some paths are non-invertible one-way paths, which is the world of infinite-dimensional category theory.
“we could take paths between paths, between paths, between paths, paths between paths, between paths, between paths, between paths, and all the way up and that goes to infinity, that's the infinity-groupoid”
- No single group captures the full data of a topological space; instead you need infinitely many groups (the homotopy groups in every positive dimension) where the nth group measures whether n-dimensional spheres in the space can be filled in, and together this family describes the full homotopy type of the space.
“a single group will not capture the full data of a topological space... really, what you have to do is introduce lots of different groups that are measuring lots of different things”
- The fundamental group counts the essentially different (homotopically inequivalent) loops an ant can make from a base point; on a torus these loops are enumerated by a pair of integers (times around the short loop and times around the long loop), whereas on a sphere every loop can be continuously shrunk to a point so the group is trivial, which proves the torus and sphere are different spaces.
“What you need are two different integers, essentially. One integer that's describing the number of times you go in a clockwise or counterclockwise direction around the short loop, and then another integer is describing the number of ways to go... around the long loop”
- Topology has multiple notions of sameness: homeomorphism requires an invertible continuous one-to-one correspondence of points (so a coffee mug equals a torus), while homotopy equivalence is weaker and allows continuous deformation that changes the number of points, so that all of n-dimensional Euclidean space is homotopy equivalent to a single point even though they are not homeomorphic.
“two spaces are homotopy equivalent if you can continuously deform one to another... all of Euclidean space, three-dimensional Euclidean space is homotopy equivalent to a single point”
Configuration spaces reveal how topology constrains what physical or abstract systems can actually do.
- On the two-sphere all one-dimensional paths between two points are essentially the same, but two-dimensional paths between paths can be fundamentally different (e.g. a square mapping the prime meridian to the date line going east over Asia versus west over the New World cannot be continuously deformed into each other on the surface), showing the fundamental groupoid alone misses structure that higher-dimensional data captures.
“in two dimensions, paths can be quite different... there is no three-dimensional path, no path between paths between paths, that continuously deforms the one to the other. If you could pass through the core of the earth, you could do that. But we have to stay on the surface.”
- A Klein bottle is constructed like a torus but with one of the gluing circles reversed in orientation; it cannot be embedded in three-dimensional space (only in four dimensions), and as a result it has no distinction between inside and outside, so liquid 'inside' it can spill onto the outside.
“you can't do this in three dimensional space. And so if we were in a fourth dimension, you could actually embed a Klein bottle into four dimensional space”
- A Mobius band is built by cutting a cylinder lengthwise into a flat strip, twisting it, and gluing it back together; the result is a non-orientable space where there is no way to decide whether you are on the inside or outside, unlike a cylinder.
“it's a non-orientable space. It's just confusing in a Mobius strip, whether you're on the inside or the outside, there's no way to decide that”
- The configuration space of two robots on a straight one-dimensional track is a square with the diagonal removed (because the robots cannot collide), yielding two disconnected triangular components corresponding to which robot is on the left; this means the robots can never swap order, whereas using a circular track makes the configuration space connected so all configurations become reachable.
“you have to remove a diagonal segment from that square, and what you're left is with these two triangular components... it's disconnected. Once you've put your robots on the track, you're not going to be able to swap their orientation with respect to each other.”
- The product of two circles (S1 cross S1) corresponds to the points on the surface of a torus, which is why the configuration space of two robots on a circular track is naturally a torus.
“the product of two circles corresponds to the points on the surface of a torus again. So S1 cross S1. The product of two circles are the points on the torus.”
Mathematics improves by making proven results human-usable through exposition and foundational reform.
- Mathematics can make the world a better place not primarily through practical applications but by making it a more interesting place to be a conscious being, which is what inspires Riehl to be a mathematician.
“a way that mathematics can make the world a better place is by making it a more interesting place to be a conscious being, and that's how I... Or that's what inspires me to be a mathematician.”
- Riehl hopes that in the future mathematics will change its foundational system to be more suitable for the complicated 'up to homotopy' structures studied today, and predicts that 22nd century mathematics would greatly surprise every living mathematician, analogous to the shift from classical to quantum mechanics.
“what I'm hoping happens in the future is that we change our foundation system of mathematics, so it's kind of more suitable to these complicated up to homotopy structures that we're thinking about today”
- In modern mathematics everyone has to be everything; even if your heart is with geometry, you have to use some algebra, reflecting the increasing complexity of the field.
“this is just the tale of complexity in the modern world. Everybody has to be everything these days. So even if you really... Your heart isn't with the geometry, you have to use some algebra.”
- Following Bill Thurston, Riehl argues mathematicians have a valuable role in making proven mathematics understandable to humans by tidying up, repackaging, and streamlining results, since a theorem being true does not make it usable; this is how once-controversial cutting-edge discoveries (like calculus) become standard undergraduate material.
“the role that mathematicians need to play in making mathematics understandable to humans... it is worthwhile to do a bit of tidying up and repackaging, streamlining”
Mathematical truths remain eternally valid and do not depend on current proof methods or foundations.
- A nice feature of mathematics is that theorems proven 2000 years ago are equally true today.
“the theorems that were proven 2000 years ago are equally true today”