Artificial intelligence may have crossed another important threshold: from explaining established mathematics to generating new mathematical arguments.
OpenAI has published a collection of ten results that it says either resolve or make substantial progress on long-standing open problems in mathematics and theoretical computer science. The problems span high-dimensional geometry, coding theory, group theory, operator algebras, quantum complexity, lattice cryptography and extremal combinatorics.
The company says the arguments were generated by an internal version of Astra, described as its next major model. Humans then worked with the same system to prepare the manuscripts, while the model formalised each argument into a Lean certificate—a machine-checkable representation designed to strengthen verification.
Among the reported results are new bounds for high-dimensional sphere packing, a construction establishing the existence of non-sofic groups, progress related to post-quantum cryptography, and resolutions of several problems associated with mathematician Paul Erdős.
The cost claim is almost as striking as the research claim. OpenAI estimates that the model tokens used to find all ten solutions would cost roughly $2,000 at current Sol API rates. If independent mathematical scrutiny confirms the work, that would suggest high-level research exploration can become dramatically cheaper and faster.
But the most important question is not whether AI replaces mathematicians. It is how the role of the mathematician changes.
AI can generate candidate proofs at enormous speed, but humans still have to examine significance, identify hidden assumptions, connect results to existing literature and decide which questions are worth pursuing. Formal verification can check logical structure, yet mathematical judgement remains essential.
OpenAI also addressed attribution directly, arguing that human authorship should not be claimed for arguments generated entirely by an AI system. That debate reaches far beyond mathematics. Research institutions will increasingly need language that distinguishes human-originated work, AI-originated work and genuine human–AI collaboration.
The emerging model may be a research partnership: machines explore an immense possibility space while humans provide direction, interpretation, responsibility and meaning.
Source: OpenAI’s official publication, “Ten advances in mathematics and theoretical computer science.”
Discover more from Marychuks.com AI, Psychology, Business & CreativeVerse
Subscribe to get the latest posts sent to your email.
