[CLSTR]
API
< Back to all clusters
[TECHNOLOGY] · United States · 39 sources

started · updated

Part of situation: OpenAI mathematical breakthroughs and transparency debate (8 clusters)

OpenAI releases 722 math manuscripts from unreleased AI model

OpenAI mathematical proofs frontier AI model Lean programming language AGMAI

OpenAI has released 722 mathematical manuscripts to a public GitHub repository, generated by an unreleased internal frontier AI model. The collection is organized into 372 families of related results, covering diverse fields such as number theory, algebraic geometry, and theoretical computer science. OpenAI reports that the model was tested against approximately 4,000 problems, with the average result requiring computational power equivalent to three hours of ChatGPT Pro reasoning.

While the release includes formalizations in the Lean programming language for approximately 42% of the results, many manuscripts remain unformalized. OpenAI’s own documentation warns that unformalized results may contain errors. The release has sparked significant debate within the mathematical community. While some researchers express excitement over the potential for AI to drive new discoveries, others, including mathematicians from MIT and the Institute for Advanced Study, have raised concerns regarding transparency and verifiability.

Critics have noted that OpenAI did not disclose the specific prompts used or the model's name, contrary to recommendations from the Advisory Group on Mathematics and AI (AGMAI). AGMAI has urged AI laboratories to use independent academic repositories rather than company-controlled platforms and to avoid using mathematical breakthroughs as marketing tools. Some researchers also expressed concern over the potential for AI to leverage existing human research without proper attribution.

Entities

Advisory Group on Mathematics and AI · Advisory Group on Mathematics and Artificial Intelligence · Andrew Sutherland · Clay Mathematics Institute · Duke University · GitHub · IBM · Institute for Advanced Study · Johns Hopkins University · Lean · Lean 4 · OpenAI

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

Най-интересните математически открития в 722-те нови научни статии на OpenAI | IT-BG. org [it-bg.org]
OpenAI, Yayımlanmamış Yapay Zekâ Modelinden Çıkan 722 Matematik Makalesini Paylaştı [dijitaliyidir.com]
OpenAI Publishes 722 Math Papers; Only 162 Are Lean-verified [bitnewsbot.com]
OpenAI pubblica 722 manoscritti matematici prodotti da un modello frontier interno non ancora rilasciato [ihal.it]
OpenAI Unveils Hundreds of Math Solutions From Unreleased Model [timesng.com]
OpenAI veröffentlicht Hunderte Mathe-Resultate: IAS kritisiert Blackbox-Tests [www.it-boltwise.de]
OpenAI Says Its AI Made Progress on 372 Unsolved Math Problems [www.sofx.com]
OpenAI rešil na stotine matematičnih problemov [slo-tech.com]
OpenAI : 377 résultats de maths, ce qui est vérifié [www.1001web.fr]
Dall'irrazionalità di pi greco alla fisica: OpenAI rilascia 722 manoscritti matematici open source [www.hwupgrade.it]
OpenAI pubblica 722 manoscritti matematici prodotti da un modello interno non rilasciato [ipaddisti.it]
OpenAI 釋出 722 篇數學論文手稿，未公開模型展現強大研究能量 | TechNews 科技新報 [technews.tw]
OpenAI Releases 722 AI-Generated Math Manuscripts After Criticism | TokenPost [www.tokenpost.com]
OpenAI、社内AIがリーマン予想の関連難問を証明と主張　論文722本をGitHubで公開 [www.itmedia.co.jp]
OpenAI Is Pissing Off a Bunch of MathematiciansAgain - IndiaVision India News & Information [www.indiavision.com]
All the drama around AI’s takeover of mathematics [www.theverge.com]
Mathematicians are combing through OpenAI's AI-generated results on over 300 problems [cryptobriefing.com]
ChatGPT Aced Its Hardest Math Test. Now the Exam Is Unsolved Math [businessmodelanalyst.com]
OpenAI AI Solves Math Problems: Automated Scientific Discovery [list25.com]
OpenAI Publishes Biggest AI Math Dump to Date [www.techjuice.pk]
OpenAI Makes 81% of Top Math Discoveries of the Last Three Years – Released in one Day [www.nextbigfuture.com]
OpenAI podría haber resuelto los enigmas matemáticos más complejos de la historia, pero nadie puede comprobar si es verdad [www.xataka.com.mx]
OpenAI, yapay zekanın gerçekleştirdiği 722 matematik keşfini yayımladı [tr.euronews.com]
OpenAI shakes up mathematics again, debuting dozens of new math solutions [me.mashable.com]
OpenAI shakes up mathematics again, debuting dozens of new math solutions [sea.mashable.com]
OpenAI lancia un'API per prendere decisioni rapide in base a immagini e test: cos'è, come funziona e quanto costa Decisions API [www.smartworld.it]
OpenAI objavio rješenja za gotovo 400 matematičkih problema [www.bug.hr]
OpenAI matematikte yeni dönemi başlattı: Yapay zekadan 722 bilimsel keşif [www.karar.com]
OpenAI 一次公開 722 篇數學手稿，數學家轟 AI 公司「黑幫行徑」 [www.blocktempo.com]
Navier-Stokes equation and beyond: How OpenAI is changing mathematics and the debate over AI creative thinking [www.theage.com.au]
OpenAI Says a Secret AI Model Cracked Hundreds of Open Math Problems in One Prompt—Mathematicians Want Receipts [decrypt.co]
OpenAI maths release: 722 papers from a model it has not released [thenextweb.com]
OpenAI, yapay zekanın ürettiği 722 matematik çalışmasını yayınladı [webrazzi.com]
OpenAI divulga 722 manuscritos matemáticos criados por IA ainda não lançada [exame.com]
Navier-Stokes equation and beyond: How OpenAI is changing mathematics and the debate over AI creative thinking [www.smh.com.au]
OpenAI publishes solutions to more than 370 outstanding math challenges. Math may never be the same [fortune.com]
OpenAI shares dozens of new math problem solutions [mashable.com]
OpenAI unleashes hundreds more math results upon a field already in shock [www.scientificamerican.com]
OpenAI Is Pissing Off a Bunch of Mathematicians—Again [www.wired.com]