Interview, Fireside Chat, Podcast
Jordan Ellenberg: Mathematics of High-Dimensional Shapes and Geometries | Lex Fridman Podcast #190
- Current classical symmetries are deemed insufficient for capturing human digit recognition in the MNIST problem, with the construction of novel symmetries by the brain identified as an open question for AI to replicate, likely involving complex image transformations rather than simple rotations.
- AI is expected to eventually prove theorems and identify counterexamples to conjectures, potentially making predictions by invoking notions of "distance" to determine similarity to unseen cases, while mathematical proofs may one day be recognized as describing real physical entities in space-time.
- The study of higher-dimensional spaces is anticipated to reveal that the universe possesses more dimensions than perceived, provided effects can be detected, while mathematical tools will enable reasoning about objects like six-dimensional cubes despite human visualization limits.
- Complex structures are expected to emerge from simple underlying laws, as demonstrated by cellular automata like the Game of Life, though the specific rules required to generate interesting behavior remain a mystery, and the three-body problem persists as a significant unresolved issue.
- Mathematical progress is projected to increasingly value computational infrastructure and programmatic visualization, potentially applying novel distance functions to economics or sociology, while the field aims to prioritize understanding over the mere production of theorems.
- Philosophical views such as ultrafinitism, which challenges the existence of infinity, are linked historically to the dawn of formal computation in the 1930s and 1940s, while the heuristic of treating prime numbers as random remains a valuable tool for generating insights despite their deterministic nature.
- The International Mathematical Olympiad is characterized as misrepresenting the reality of mathematical research due to its focus on speed and canned problems, whereas effective learning is expected to stem from personal investment in specific problems rather than abstract curricula.
- Future solutions to complex problems in fields like consciousness and life are hoped to be found in simple laws, with the perception of numerical transitions viewed as fundamental to cognition, while the proof of Fermat's Last Theorem is noted for its execution size despite the potential simplicity of its underlying truth.