After Math
Rapid advances in AI-generated proofs for major problems like Navier–Stokes are forcing mathematicians to confront what it really means to “solve” a problem: is a formally correct but unintelligible proof enough, or is human understanding central to mathematics? Commenters weigh the value of AI as a powerful assistant against fears of lost purpose, broken credit and career systems, and potential shifts in public funding if machines outpace human researchers. Many argue that the deeper stakes go beyond employment to questions of whether mathematics—and scientific inquiry more broadly—should remain oriented around human comprehension rather than opaque oracles.
AI “solutions” vs mathematical understanding
- Many commenters argue that a formally correct, machine-checked proof is not enough; mathematics is about human-understandable arguments that generate new concepts and insight.
- Others counter that a logically valid proof does count as solving the problem, and redefining “solution” to require an extra “intelligible” property is seen as goalpost-shifting.
- There is debate whether distinguishing “formal” vs “intelligible” proofs is essential or just philosophical flourish.
Formal proofs, readability, and attribution
- The Navier–Stokes and other recent AI-generated proofs are described as unreadable, meandering, and possibly not reflecting real “understanding.”
- Attribution is a major concern: systems are trained on others’ work and chats but cannot give credit, leading to resentment about where recognition and prizes should go.
- Some view this as a continuation of earlier worries about opaque computer proofs (e.g., large case-checking), now amplified.
Impact on mathematicians, careers, and funding
- Anxiety that problem-solving, long a key metric for hiring, tenure, and grants, is being automated. Some fear departments shrinking and ideas being hoarded rather than shared.
- Others suggest new roles: interpreting AI proofs, curating questions, and guiding models, but this feels to some like a “participation trophy” compared to traditional achievements.
- There is explicit questioning of whether public funding for math should change if AI can generate many key results.
Parallels with software and productivity claims
- Mixed reports: some small teams claim ~10x code output with AI; others see little visible payoff in mainstream software quality or novelty.
- Several note that large organizations are limited by coordination and taste, not coding speed, so AI’s impact will show first in small, less-visible groups.
- This analogy is used to suggest AI might both supercharge and trivialize aspects of math: more prototyping, less deep investment.
Purpose, meaning, and “purpose death”
- One thread focuses on “purpose death”: the crisis when a long-cultivated skill is automated, producing existential dread and grief-like stages.
- Some see this as a widespread future experience; others note many people already ground their purpose in religion, family, or nature and will be largely unaffected.
- There is pushback against pathologizing or psychoanalyzing specific mathematicians and against AI “doomerism.”
Capabilities and limits of current AI in math
- Commenters note AI’s current strengths in brute-force existence proofs, counterexamples, and formal verification, but relative weakness in big-picture creativity and question-posing.
- Some expect rapid progress to creative, agenda-setting systems; others insist current architectures fundamentally lack human-like reasoning and abstraction.
Future of mathematics with AI
- One view: math’s ultimate goal is human understanding; AI must be a tool that accelerates that, including systems that can explain prior AI work.
- Another view: it’s acceptable if AI, not humans, becomes the main engine of mathematical progress; humans may shift to steering, interpreting, or simply pursuing math as a personal, aesthetic practice.
- Several emphasize that mathematics is likely infinite; AI will change the game but not exhaust the subject.