Lex Fridman PodcastJordan Ellenberg: Mathematics of High-Dimensional Shapes and Geometries | Lex Fridman Podcast #190
CHAPTERS
- 0:00 – 4:18
Math as a cognitive process: language, visuals, and “Behold” proofs
Lex and Jordan explore whether mathematical thinking is fundamentally linguistic or whether it can be genuinely non-verbal. Jordan uses dissection proofs (including the legendary Bhāskara “Behold” diagram) to show how pictures can function as rigorous reasoning. They converge on the idea that math (like language) is an active process of manipulation and transformation, not just static symbols.
- •Math often feels like producing language: propositions, structured statements, and argument flow
- •Dissection proofs as purely visual reasoning; Bhāskara’s Pythagorean proof and the “Behold” story
- •Question of whether a diagram/movie counts as “language”
- •Math and language framed as processes of transformation and composition
- 4:18 – 9:15
Why geometry feels different: the “cilantro of math” and an early symmetry epiphany
Lex describes geometry as the gateway to meaning and certainty; Jordan notes geometry polarizes people—some love it, others feel alienated. Jordan recounts a childhood moment counting holes in a 6×8 speaker grid, “seeing” commutativity as a geometric fact. The discussion sets up geometry as a bridge between algebraic truths and spatial intuition.
- •Geometry as emotionally distinctive: love-it-or-hate-it subject
- •Jordan’s 6×8 vs 8×6 realization as embodied proof of commutativity
- •Geometry as the intertwining of algebra and spatial structure
- •Teaching goal: helping students directly access ‘why’ a fact is true
- 9:15 – 18:35
Symmetry as invariance: from axes to ‘scronches’ and the modern question of sameness
Jordan defines symmetry broadly as any transformation under which key properties remain invariant, not only mirror/rotational symmetry. He connects this to the central modern mathematical theme: deciding when two objects count as ‘the same.’ Lex ties this to AI perception (MNIST) and category boundaries between digits.
- •Classical symmetry (axes/rotations) generalized to arbitrary transformations
- •“Scronch” example: stretch one way, shrink another—still a symmetry in a broader sense
- •Modern math focus: equivalence via invariants under transformations (e.g., translation invariance)
- •AI analogy: what transformations preserve digit identity vs change it
- 18:35 – 31:14
Poincaré, the three-body problem, and why dynamics forces higher-dimensional geometry
Jordan introduces Henri Poincaré via the three-body problem and the birth of chaotic dynamics. He explains why adding one body turns solvable orbital mechanics into a sensitive, tangled system. This motivates phase space: describing position and velocity together as points in higher-dimensional spaces, driving the need for a geometry that works in any dimension.
- •Historical context: national ambition and modernization shaping math institutions
- •Three-body problem: stability vs instability; tiny initial changes yield huge outcome differences
- •Poincaré as a pioneer of chaotic dynamics and qualitative geometry of motion
- •Phase space as 6D per body (position + velocity), pushing beyond 3D intuition
- 31:14 – 41:52
Topology (‘analysis situs’) and the Poincaré Conjecture via loops you can’t shrink
Jordan frames topology as an intrinsic geometry—properties you can detect without ‘stepping outside’ the space. Using a mug, he explains simple connectivity: every loop can be contracted to a point in ordinary space, but not on a mug’s handle. This becomes the accessible doorway into what the Poincaré Conjecture is trying to characterize in 3D manifolds.
- •Topology as intrinsic vs extrinsic description of shape
- •Simple connectivity: shrink every loop vs loops trapped by a handle
- •Universe-shape question: what if your ‘string’ is cosmic in scale?
- •Poincaré Conjecture as recognizing when a 3D space is the ‘standard’ one
- 41:52 – 46:37
Flatland, extra dimensions, and what math can know without visualization
They use Flatland to discuss whether humans could ‘reason’ about higher dimensions even if we can’t picture them. Jordan argues we can compute and prove properties of high-dimensional objects (e.g., corners of an n-cube) despite limited imagination. They also touch on the religious subtext of Flatland and the philosophical move from perception to abstract reasoning.
- •Flatland’s sphere as a metaphor for transcending cognitive limits
- •Possible extra dimensions in physics vs higher dimensions as analytic tools
- •Reasoning beats visualization: we can derive facts about tesseracts and n-cubes
- •Religious allegory and the sphere’s own inability to imagine a 4th dimension
- 46:37 – 56:10
How many holes does a straw have? From arguments to homology arithmetic
A deceptively simple question becomes a tour of how mathematicians formalize ‘holes.’ Jordan surveys competing intuitions (0, 1, 2 holes), then escalates to pants to reveal relationships among holes. This leads to the homology mindset: holes behave like quantities with addition/subtraction—‘two holes, but one is the negative of the other.’
- •Why groups disagree: ‘one region through the straw’ vs ‘top and bottom holes’
- •Engineering vs topological intuitions (bagel analogy; manufacturing doesn’t ‘poke’ a hole)
- •Pants example: waist hole as ‘leg + leg’—introducing relations among holes
- •Homology viewpoint: holes have algebraic structure (addition/negatives)
- 56:10 – 1:01:50
Math communication and the rise of visual explanation: 3Blue1Brown and beyond
They discuss what it means for math writing/teaching to contain real mathematical action rather than only commentary. Jordan praises the explosion of math YouTube (3Blue1Brown, Numberphile, Vi Hart) as a new channel with massive reach. They also note academia is still learning how to value visualization tools and computational infrastructure as legitimate mathematical contributions.
- •Difference between ‘a book about math’ and ‘math happening on the page’
- •YouTube as a higher-reach teaching medium; different engagement trade-offs vs classroom/books
- •Grant Sanderson’s programmatic visualization as a distinct kind of mathematical work
- •Institutional shift: increasing recognition of visualization and computational math tooling
- 1:01:50 – 1:08:42
Will AI win a Fields Medal? Computation, conjectures, and the Conway knot breakthrough
Jordan doubts an AI will win a Fields Medal soon, but he’s deeply interested in AI’s role in pure math. He notes many tasks once counted as research are now ‘just computation’ via systems like Sage/Magma/Macaulay. The conversation highlights AI’s current strengths (e.g., finding counterexamples) and contrasts them with human insight through the Lisa Piccirillo solution of the Conway knot ‘slice’ problem.
- •Shifting boundary: what counts as ‘math research’ vs routine computation over time
- •Computer algebra systems as standard tools in modern math practice
- •Neural nets can help search for counterexamples; usefulness even when problems aren’t famous
- •Piccirillo’s short, picture-heavy proof prompts: what does ‘difficulty’ mean in math?
- 1:08:42 – 1:33:23
Fermat’s Last Theorem: myth of a simple proof, deformation theory, and p-adic distance
Lex presses on beauty and simplicity in famous theorems; Jordan argues Fermat likely didn’t have the claimed proof. They explain how the Fermat problem fueled vast new machinery (unique factorization failure, modern number theory). Jordan sketches deformation theory’s intuition and introduces p-adic distance—where ‘closeness’ is about divisibility by high powers of a prime—and links this to how Wiles’ methods ‘move’ objects infinitesimally in a nonstandard metric.
- •Fermat’s marginal note as legend; historical reasons we think Fermat lacked a full proof
- •Fermat work catalyzed deep developments (Kummer and beyond)
- •Deformation theory intuition: local infinitesimal motion revealing global structure (R=T)
- •p-adic metrics: numbers are ‘close’ if their difference has large prime power factors
- 1:33:23 – 1:48:00
Prime numbers: pseudo-primes, twin primes, and why randomness is a powerful model
Jordan defines primes as the ‘atoms’ of integers and explains Fermat-style primality tests, including pseudo-primes that pass despite being composite. They discuss prime gaps and the twin prime conjecture, emphasizing that modern heuristics treat primes as if they were randomly scattered. The key methodological point: pretending a deterministic object is random can generate productive conjectures and intuition.
- •Primes as irreducible building blocks; why ‘1’ is excluded by convention and usefulness
- •Fermat test: if 2^n mod n isn’t 2, n is composite; pseudo-primes fool the test
- •Twin primes and prime gaps: unknown whether infinitely many pairs differ by 2
- •Random-model heuristics: deterministic primes can be studied ‘as if random’ to predict truths
- 1:48:00 – 2:06:46
Conway, Game of Life, and when simple rules create complexity (and when they don’t)
Jordan and Lex celebrate John Conway’s playful genius and how personality and mathematics intertwine. They explain the Game of Life as a simple local update rule yielding emergent ‘organisms’ like gliders and guns. A central open mystery is why only certain rule sets generate interesting complexity and why prediction can be computationally intractable (Wolfram’s irreducibility theme).
- •Conway’s style: inventing games, analyzing them, and building structures like surreal numbers
- •Game of Life basics: births/deaths on a grid; rich emergent behavior from short rules
- •Complexity isn’t typical: most rules are dull; ‘tuned’ rules create lifelike dynamics
- •Limits of prediction: why even simple deterministic systems can resist simplification
- 2:06:46 – 2:18:05
Group theory and physics symmetries: from squares to infinite groups and (a glimpse of) gauge ideas
Jordan introduces group theory as the abstract study of symmetries—transformations that can be composed and include an identity ‘do nothing’ move. He illustrates with symmetries of people, rectangles, squares, and permutations (shuffling a deck). They briefly connect this to physics: changing what counts as a symmetry (e.g., Lorentz group) reshapes how we interpret reality and invariance.
- •Groups encode symmetries + composition; always include the identity operation
- •Examples: bilateral symmetry (2), rectangle (4), square (8)
- •Permutation groups: shuffles as symmetries; relevance to computation and sorting
- •Physics lens: laws as invariants; relativity as a shift in the symmetry group
- 2:18:05 – 2:28:17
Perelman’s proof of the Poincaré Conjecture: Ricci flow and the ‘geometry of geometries’
Jordan describes the conceptual heart of Perelman’s solution: to answer a question about a 3D space, study the space of all 3D geometries and follow a deformation path through it. Ricci flow acts like a smoothing evolution that pushes a manifold toward canonical form, with the technical battle being controlling singularities. The story highlights a recurring theme: real progress often comes from stepping up a level of abstraction.
- •Key idea: not just geometry of a manifold, but geometry of the space of geometries
- •Ricci flow as a controlled deformation process toward standard forms
- •Technical challenge: managing singularities (non-smooth kinks) along the flow
- •Perelman as culmination of Hamilton and a broader program, not an isolated miracle
- 2:28:17 – 2:41:47
Learning math and giving life advice: self-knowledge, motivation, and problem-driven study
The conversation turns reflective: how to learn math well and how to persist through difficulty. Jordan argues there’s no single path—textbooks, contests, Gardner-style play, videos—but sustained progress usually requires caring about a problem. Lex connects this to embracing hardness as part of growth, while Jordan emphasizes accepting dead ends even if they’re frustrating.
- •Learning approaches vary: textbooks, Martin Gardner, contest problems, videos
- •Problem-first learning: you learn tools when you need them for something you care about
- •Grad school lesson: reading huge tomes without a driving problem rarely works
- •Hardness, frustration, and dead ends as unavoidable parts of mathematical development
