/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Q&A with Terence Tao on AI-generated Erdős solutions, “cheap wins”, hybrid human AI contributions, push-of-a-button workflows, new ways of doing math, and more

Terence Tao, the legendary mathematician, explains the promise of generative AI.

The Atlantic Matteo Wong

Context & Ripple Effects

Tao’s earlier discussion of how OpenAI’s o1 might assist mathematicians framed AI as a tool to be evaluated within expert practice, not as a substitute for proof standards. This interview extends that arc toward hybrid workflows and lower-friction exploration.

The surrounding coverage shows research AI moving from broad scientific applications toward practical use by researchers, while later reporting on a GPT-assisted solution to a long-standing Erdős problem tests what an apparently simple workflow can actually produce. The unresolved issue is not whether models can generate leads, but how experts validate, interpret, and build on them.

First-order effects

  • Mathematicians gain a more explicit model for using generative AI: delegate inexpensive exploratory work and retain human judgment for selecting, checking, and developing results.
  • AI-generated solutions to Erdős-style problems put greater immediate weight on verification, provenance, and expert assessment before a model output can count as useful mathematical work.

Second-order effects

  • Tool builders will face pressure to improve proof checking, traceability, and interfaces that support iterative human-AI collaboration rather than merely producing polished answers.
  • A lower cost of generating candidate approaches can broaden participation in problem solving, but it also increases the review burden on experts and institutions that establish correctness.

Third-order effects

  • If hybrid workflows repeatedly yield validated results, mathematical research may shift toward a division of labor in which model-generated search and human-led formalization, judgment, and agenda-setting are distinct inputs.
  • The lasting advantage may accrue less to systems that produce a one-shot answer than to workflows that make reasoning auditable and reliably turn abundant candidates into accepted knowledge.

The trend: This is one instance of reasoning economics: AI is reducing the cost of generating research leads while raising the importance of validation and expert orchestration.

Discussion

  • @nxthompson @nxthompson on x
    I love this metaphor from Terence Tao—widely considered the world's greatest living mathematician—about one of the drawbacks of using AI to solve hard math problems. https://www.theatlantic.com/ ... [image]
  • @matteowong Matteo Wong on bluesky
    I checked with Terence Tao about all the excitement around AI for mathematics research: www.theatlantic.com/technology/ 2...
  • @mikecaulfield Mike Caulfield on bluesky
    “a lot of AI companies have this obsession with push-of-a-button, completely autonomous workflows where you give your task to the AI, and then you just go have a coffee, and you come back and the problem is solved.  That's actually not ideal.” www.theatlantic.com/technology/ 2...