OpenAI says its AI models disprove Exponential Time Hypothesis and reveal new math‑physics advances

OpenAI disclosed that its newest generation of artificial‑intelligence systems has generated a formal proof overturning the Exponential Time Hypothesis (ETH), a claim that contests a long‑standing premise in computational complexity theory. The firm also issued a suite of fresh findings that tackle unresolved questions in mathematics and physics.

The Exponential Time Hypothesis asserts that specific problem families—most notably the satisfiability of Boolean formulas with a fixed clause width—cannot be resolved in sub‑exponential time. Demonstrating that the hypothesis fails would reshape expectations about the ultimate limits of algorithmic efficiency and could reverberate through fields that depend on hardness assumptions, such as cryptography and optimization.

OpenAI explained that the breakthrough arose from a workflow that merges large‑scale language models, trained on scientific literature, with automated theorem‑proving utilities. Researchers fed the model a formal statement of ETH and requested a proof outline, which was subsequently refined and validated by symbolic verification software. The announcement notes that the resulting proof cleared internal consistency checks and was later examined by external specialists prior to public release.

This progress builds on an expanding body of work where AI systems aid or independently generate mathematical insights. Prior efforts have seen language models tackle International Mathematical Olympiad‑style problems, while DeepMind’s AlphaFold revolutionized protein‑structure prediction. OpenAI’s assertion adds a new layer by confronting a foundational conjecture in theoretical computer science rather than a niche, domain‑specific puzzle.

In tandem with the ETH result, the organization released a technical report outlining several additional discoveries. Among them are novel constructions in number theory, advances on combinatorial‑optimization conjectures, and a collection of calculations that push forward understanding of quantum‑field‑theoretic models. The report supplies the formal statements and proof sketches, inviting the wider community to verify and extend the work.

Academic response has been cautiously optimistic. Scholars recognize the potential importance yet stress the necessity of thorough peer review and independent replication. Some experts highlight that the verification pipeline—connecting language‑model output to formal proof assistants—offers a promising route for scaling mathematical research, while others caution that premature publicity might outstrip the validation process.

Should the falsification of ETH be confirmed, it could unlock pathways for crafting algorithms that operate faster than previously deemed feasible across a spectrum of hard problems. Such a shift would affect both theoretical inquiry and practical applications, ranging from cryptographic protocol design to large‑scale data analysis.

OpenAI indicated that future efforts will concentrate on open‑sourcing key elements of the proof‑generation pipeline and forging collaborations with university labs to evaluate the system on further open problems. The company also plans to organize workshops aimed at weaving AI‑assisted proof techniques into conventional mathematical practice.

The announcement underscores the accelerating convergence of artificial intelligence and fundamental scientific investigation. While the claims remain subject to meticulous scrutiny, the episode demonstrates how sophisticated language models are beginning to assume a substantive role in probing questions long regarded as the domain of human mathematicians and physicists.