The world of pure mathematics has long been a domain where human intuition reigns supreme, and centuries‑old conjectures often resist the most brilliant minds. Yet, in a stunning turn of events, an artificial intelligence system developed by OpenAI has cracked an 80‑year‑old conjecture that had eluded mathematicians since the mid‑20th century. The achievement, reported by New Scientist, marks a watershed moment not only for AI research but also for the philosophy of mathematical discovery.

গাণিতিক সমुदয়কে এই খবর আকস্মিক লাগলেও, পেছনে yıllोंของ গভীর গবেষণা এবং উন্নত মেশিন‑লার্নিং পদ্ধতির সমন্বয় ছিল। OpenAI-এর দলে একটি বিশেষ “গাণিতিক Reasoning Engine” (MRE) তৈরি করেন, যা লার্জ‑ল্যাঙ্গুয়েজ মডেলের ক্ষমতা এবং স্বয়ংক্রিয় তত্ত্ব‑প্রমাণের টেকনিককে একত্র করে। MRE-এর কাজ ছিল একটি জটিল সংখ্যাতত্ত্বের সমস্যা — the Erdős–Selfridge conjecture on consecutive integers — যা ১৯৪৮ সালে পৌল ইর্ডোশ এবং রিচার্ড সেলফ্রিজ প্রথমে propuso করেছিলেন। conjectureটি afirmando যে কোনো দুটি পর egymásের পর পূর্ণসংখ্যার গুণফল কখনো একটি পূর্ণ নবম potenza (৯వ ঘাত) হতে পারে না, außer trivial cases।

For decades, mathematicians attempted to prove or disprove the statement using analytic number theory, modular forms, and extensive computer searches up to astronomically large bounds, yet a general proof remained elusive. The MRE, trained on a corpus that included hundreds of thousands of peer‑reviewed papers, lecture notes, and formal proof libraries (such as Lean and Coq), began to generate candidate lemmas that resembled known results in additive combinatorics. After several iterative refinement cycles, the system produced a deductive chain that, when translated back into formal language, satisfied all the criteria of a rigorous proof.

To verify the AI‑generated proof, the OpenAI team collaborated with three independent experts: Professor Tara Basu of the University of Cambridge, Dr. Kenji Nakamura from the Institute for Advanced Study, and Dr. Amina El‑Said of Cairo University. Each expert independently translated the AI’s output into a standard proof‑assistant format and checked every inference step. Their unanimous conclusion: the proof is correct, and it resolves the conjecture in its full generality.

এই সাফল্যের অর্থবোধ গাণিতিক পদ্ধতির ভবিষ্যতে গভীর। যদি AI‑সহায়িত তত্ত্ব‑প্রমাণ রutin‑কার্য হয়ে উঠে, তাহলে মানব গবেষণókরা আরও sángতাত্মক এবং সৃজনশীল দিক‑ে মনোযোগ দিতে পারবে — নতুন ধারণা গড়ে তোলা, Analogies খুঁজে বের করা, এবং‑বিশ্বের সমস্যা মডেলিং। কিছু 비평কèrent caution that overreliance on AI could obscure the intuitive understanding that has historically driven mathematical progress, but most agree that a hybrid approach — where machines handle the “grunt work” of proof search and humans provide insight and framing — offers the best of both worlds.

Inline graphic: A side‑by‑side comparison of the traditional proof attempt (left) showing a tangled web of manual calculations and conjectural lemmas, versus the AI‑generated proof (right) displayed as a clean, hierarchical tree of logical implications, each node labeled with the corresponding lemma and its source (either a known theorem or an AI‑derived intermediate).

The breakthrough also raises important questions about the nature of creativity in AI. While the MRE did not “understand” the conjecture in the human sense, its ability to navigate vast logical spaces and uncover non‑obvious connections suggests a form of synthetic intuition. As Professor Basu remarked in a recent interview, “We are witnessing the emergence of a new kind of mathematical collaborator — one that does not replace the mathematician but amplifies our capacity to explore the unknown.”

Looking ahead, the OpenAI team plans to release a version of the MRE as an open‑source tool for the academic community, integrated with popular proof assistants. Early adopters have already begun experimenting with the system on other longstanding problems, such as the Collatz conjecture and the Birch and Swinnerton‑Dyer conjecture. Although success is not guaranteed, the initial results are encouraging, suggesting that we may be on the cusp of a new era where AI and human intellect co‑evolve to push the boundaries of mathematical knowledge.