@algeboy
YouTube
×
📁 Articles
💾 Software
✏️ Posts
🎓 Professional
5000 Linear DataKing
Wilson

MathChat

What are people saying about AI & Math?

Current viewpoint map

AI and mathematics viewpoints A scatter plot. The horizontal axis is evidence basis from speculative to data-supported. The vertical axis is outlook from anxious to hopeful. Circle size represents provisional source reliability. 0255075100 0255075100 Evidence basis: speculative (0) → data-supported (100) Outlook: anxious (0) → hopeful (100) Francis Su — outlook 82, evidence 35, reliability 60The Economist — outlook 28, evidence 86, reliability 72Conrad Wolfram — outlook 72, evidence 68, reliability 78Yahoo report — outlook 30, evidence 70, reliability 55Jaron Lanier — outlook 60, evidence 36, reliability 64 David Bessis — outlook 54, evidence 38, reliability 68Geordie Williamson — outlook 78, evidence 62, reliability 80Terence Tao — outlook 61, evidence 78, reliability 88Max Weinreich — outlook 12, evidence 48, reliability 65Timothy Gowers — outlook 58, evidence 82, reliability 90Jeremy Avigad — outlook 64, evidence 76, reliability 88Jacob Tsimerman — outlook 88, evidence 55, reliability 76Emily Riehl — outlook 56, evidence 86, reliability 90 Leiden Declaration — outlook 24, evidence 72, reliability 82 AI Snake Oil — outlook 38, evidence 78, reliability 82Yann LeCun — outlook 65, evidence 65, reliability 84Gary Marcus — outlook 32, evidence 65, reliability 84Stephen Wolfram — outlook 78, evidence 55, reliability 74 Jordana Cepelewicz / Quanta — outlook 64, evidence 76, reliability 82Steven Strogatz / Quanta — outlook 60, evidence 70, reliability 84Sean Carroll / Mindscape — outlook 55, evidence 54, reliability 76Curt Jaimungal / Theories of Everything — outlook 76, evidence 55, reliability 70Sabine Hossenfelder — outlook 58, evidence 45, reliability 64Brian Keating / Into the Impossible — outlook 61, evidence 58, reliability 74 SuEconomistConrad WolframYahooLanierBessisCarrollJaimungalHossenfelderKeatingWilliamsonTaoWeinreichGowersAvigadQuanta: CepelewiczQuanta: StrogatzTsimermanRiehlLeidenAI Snake OilLeCunS. WolframMarcus
Circle size = source reliability
Add your own source for scoring Join the discussion board

On this first map, hopeful means that the source expects AI to have a net positive effect on mathematics or mathematics education—for example, by assisting discovery, explanation, accessibility, or formal verification. It does not simply mean “AI can do mathematics,” confidence in artificial general intelligence, or approval of every AI use. Conversely, an anxious score reflects concern about the net effect on learning, proof, research culture, or public institutions.

Each source is reviewed along three provisional dimensions:

  1. Outlook: anxious (0) to hopeful (100)
  2. Evidence basis: speculative (0) to data-supported (100)
  3. Source reliability: lower (0) to higher (100), based on relevant expertise, transparency, methods, primary sourcing, and relevance

The numbers are deliberately open to revision. They describe a particular source’s argument, never an author’s worth or status.

AI openness by source category

This strip plot avoids treating categories as a numerical axis. Rows show the source’s primary public role for this review; horizontal position shows openness to AI use. Categories describe source context, not motive or argument quality.

AI openness by source category Sources are grouped into educator, journalist, and AI-industry rows, and positioned horizontally by openness to AI use. 0255075100 EducatorJournalistAI industry Openness to AI use: reject (0) → actively embrace (100) Francis Su — 65David Bessis — 48Geordie Williamson — 80Terence Tao — 70Max Weinreich — 5Leiden Declaration — 45Timothy Gowers — 60Jeremy Avigad — 70Jacob Tsimerman — 90Emily Riehl — 65 The Economist — 40Yahoo report — 35AI Snake Oil — 40Jordana Cepelewicz / Quanta — 70Steven Strogatz / Quanta — 65Sean Carroll / Mindscape — 55Curt Jaimungal — 75Sabine Hossenfelder — 55Brian Keating — 65 Conrad Wolfram — 90Yann LeCun — 80Jaron Lanier — 65Gary Marcus — 40Stephen Wolfram — 92

The category assignments and openness scores are reviewable in openness-by-category.csv.

What the map currently suggests

  • There is no simple pro-AI/anti-AI divide. Several optimistic sources also insist on verification, disclosure, and human responsibility.
  • The clearest empirical concern in this collection is educational: AI can improve visible homework performance while weakening unaided performance. That supports careful course and assessment design, not a blanket ban.
  • Research-mathematics sources focus on a different risk: generated claims and proofs may outpace the community’s ability to verify, understand, attribute, and teach them.
  • The broadest common ground is conditional adoption: use AI for explanation, exploration, routine tasks, and formal assistance; preserve independent practice and require transparent checking for consequential mathematical claims.

Current review map

Source Outlook Evidence basis Reliability Main contribution
Stephen Wolfram 78 55 74 Computation-augmented AI: pair generative models with exact computation.
Conrad Wolfram 72 68 78 Mathematics-education reform for the AI age.
Geordie Williamson 78 62 80 AI may contribute to mathematical discovery.
Max Weinreich 12 48 65 Argues against AI-generated mathematics.
Jacob Tsimerman 88 55 76 Strongly future-facing research-mathematics forecast.
Terence Tao 61 78 88 Conditional analysis of mathematical values, verification, and practice.
Steven Strogatz (Quanta) 60 70 84 Podcast interview on black-box models, uncertainty, and statistical reasoning.
Francis Su 82 35 60 Humanistic case for mathematics and learning.
Emily Riehl 56 86 90 Tests and verification for meaningful AI contribution to mathematics.
Gary Marcus 32 65 84 LLM reasoning can be brittle; plausible output is not robust abstraction.
Yann LeCun 65 65 84 Broad-AI baseline: limits of current language models and future architectures.
Leiden Declaration 24 72 82 Governance, responsibility, and peer-review proposals.
Jaron Lanier 60 36 64 Immersive mathematical visualization, paired with cautions about treating learners as data.
Brian Keating 61 58 74 Tao interview on AI as a complementary research tool requiring verification.
Curt Jaimungal 76 55 70 Yang-Hui He interview on AI-assisted mathematical discovery and its limits.
Sabine Hossenfelder 58 45 64 Explainer on claimed AI mathematics breakthroughs; full transcript pending public access.
Timothy Gowers 58 82 90 Separates verified mathematical progress from AI hype.
The Economist 28 86 72 Reports recent evidence on AI use and secondary-school learning.
Jordana Cepelewicz (Quanta) 64 76 82 Quanta’s reported synthesis of AI-assisted proof and changing mathematical practice.
Sean Carroll 55 54 76 Podcast discussion of neural-network mathematics and the limits of data-hungry models.
David Bessis 54 38 68 Mathematical understanding matters beyond theorem production.
Jeremy Avigad 64 76 88 Formalization, proof, and responsible mathematical practice.
AI Snake Oil 38 78 82 Evidence-oriented education baseline, not mathematics-specific.
Yahoo report 30 70 55 Secondary reporting on student-learning evidence.

Explore, challenge, or extend the review

The complete public record includes the source ledger, scoring methodology, prompt and model-assisted review record, contribution guide, and a small interactive HTML version of the map.

Open the MathChat review repository on GitHub

A note on evidence

“Evidence basis” is not a measure of author prestige. It asks whether the source’s central claim is directly supported by relevant data, transparent methods, primary sources, and appropriate caution about uncertainty and causation. A thoughtful philosophical essay can be valuable while still scoring lower on this particular axis.

Add an author or correct the map

Please use the author or source suggestion form to propose a specific source, challenge a score, or point to better evidence. Keep submissions tidy, polite, and apolitical. Source snapshots in the repository preserve links, access notes, and review summaries without republishing third-party articles.

Leave a message

Questions, source leads, score challenges, and constructive corrections are welcome. Keep the discussion tidy, polite, and apolitical.