Using its in-house model, OpenAI says it has produced three groundbreaking research results in mathematics, computer science and physics.

Vài tuần sau khi công bố bản thảo dài hơn 700 trang trình bày lời giải và những tiến triển trong 372 bài toán lớn, ngày 26/10, OpenAI tiếp tục công bố những phát hiện mới. Lần này, kết quả không giới hạn ở toán học, mà còn mở rộng sang khoa học máy tính và vật lý.

In the field of mathematics, OpenAI claims that its model has successfully proven the Local Smoothing Conjecture (LSC) hypothesis in all dimensions, based on an auxiliary partial differential equation of a new type. LSC is one of the most central and fundamental problems in the field of harmonic analysis and partial differential equations (PDE).

Proposed by mathematician Christopher Sogge in 1991, the hypothesis concerns the way waves propagate and converge: the strongest concentrated states of waves are predicted to be short-lived.

According to Engadget, the local smoothing hypothesis (LSC) has fundamental significance in harmonic analysis, with far-reaching consequences for mathematical analysis. If OpenAI's results are confirmed by experts, it will lead to solutions to many other classic problems, such as the Kakeya conjecture - a problem that involves rotating a needle in all directions in space, or the Stein restriction hypothesis, which are major challenges that have troubled mathematicians for decades.

In the field of computer science, OpenAI claims that AI has proven to disprove the Exponential Time Conjecture (ETH), which has been trusted by researchers for decades in many fields such as graph theory, computational biology, databases and computational geometry. According to OpenAI, the model proved the hypothesis wrong on October 16. The results were then validated in Lean, an AI tool for functional programming and theorem proving, on October 17.

OpenAI said it will establish a group of independent expert collaborators to interpret the results, and make the product public so anyone can review and develop based on it. This time, the work will be led and supervised by a team of experts including Russell Impagliazzo and Ramamohan Paturi of the University of California at San Diego, and Ryan Williams of the Massachusetts Institute of Technology (MIT).

In physics, OpenAI said its model has mathematically proven the existence of several quantum field theories interacting in four-dimensional space, including Quantum Chromodynamics (QCD) and a unified theory that completes the Standard Model of particle physics. The work is based on the methods of mathematician and physicist Abraham Balaban in the field of tectonic quantum field theory.

When dealing with QCD and the Standard Model, OpenAI's model also solves the sub-problem of proving the existence of pure Yang-Mills theory - part of the Clay Mathematics Institute's Millennium Prize on Yang-Mills theory problem. The company said the model produced natural language results for the existence of Yang–Mills on September 9, followed by QCD on September 18.

Due to the length and complexity of the project, OpenAI has postponed publication until formalization using Lean is complete.

With three new results as well as future achievements, OpenAI said it "does not seek academic recognition" for its AI findings. Instead, experts from independent research groups will name the project and take responsibility.

In addition, OpenAI said it has added a $5 million contribution to the Simons Foundation - a fund that helps create conditions for researchers to analyze and develop results generated by AI. The company is committed to securing funding for independent scientists to evaluate, interpret and expand, and expects the collaboration between humans and AI to unlock new insights, clarifying capabilities and the limits of computational complexity.

OpenAI's announcement continues a series of "achievements" of AI in scientific discovery, according to Wired. In the middle of this month, researchers at the Massacusetts Institute of Technology predicted that artificial intelligence will revolutionize statistical probability, economic theory, cosmology, quantum physics or mathematics. However, it also caused a series of mixed opinions.

Particularly in the field of mathematics, after OpenAI announced 372 mathematical results in early October, the Mathematical Humanities Association with more than 800 members issued a protest notice, saying this does not actually promote mathematics forward, and even called for a boycott of OpenAI. In September, 25 mathematicians who received the Fields Medal signed an open letter, arguing that the mass "dumping" of mathematical research could harm the field, disrupting the process of research and knowledge transmission.

Bảo Lâm - Huy Đức tổng hợp