Fields Medalist Zelmanov: OpenAI and Anthropic Claiming to Solve Math Problems Does Not Equal Advancing Mathematics

Article body

"This has not advanced mathematics, nor has it educated any new generation (of mathematicians)." On November 2, at the "Hong Kong World Youth Science Conference — Into Beiwai and the 2026 Beiwai Nobel Summit Dialogue" event, when a Daily Economic News reporter asked how he viewed the recent situation in which OpenAI and Anthropic frequently claimed to have solved mathematical problems, 1994 Fields Medalist Yefim Zelmanov said bluntly.

While acknowledging that the related progress is "exciting," Zelmanov stressed that these "solutions" currently remain at the stage of "claimed solutions" and have not yet been verified through actual texts. "If they can all be confirmed, then that would indeed be astonishing." In Zelmanov's view, this kind of "problem-solving" is fundamentally different from the work of human mathematicians — it has neither expanded the boundaries of mathematics nor produced an educational effect.

At this dialogue event, top scientists including several Nobel Prize and Fields Medal laureates gathered in the Zhongguancun Beiwai community. Participating scholars generally agreed that AI (artificial intelligence) is accelerating scientific discovery and boosting productivity, but they also warned about issues such as black boxes and security risks.

Omar Yaghi, winner of the 2025 Nobel Prize in Chemistry and pioneer of reticular chemistry, delivered a keynote speech, proposing that in the AI era the real bottleneck is no longer "how many possibilities can be generated" but "how to choose." The core capability of future scientific research is to pose high-quality big questions that can change systems, and to make choices based on facts and evidence from the infinite possibilities generated by AI — what he calls "the science of choice."

The emerging discipline of "the science of choice" proposed by Omar Yaghi systematically studies how people decide which things are worth conceiving, creating, testing, developing, and ultimately putting into use when possibilities far exceed the scope people can actually explore.

Omar Yaghi said that the science of choice can become the core intellectual foundation of AI for Science.

Photo provided by the event organizer

At this dialogue event, Omar Yaghi pointed out that the scientific community currently has serious gaps in records of "how discoveries are made" and "the failed paths before discovery." The greatest value of AI lies in its ability to capture and build databases of this tacit knowledge, thereby better training models.

Omar Yaghi believes that in the future, laboratories will all become automated laboratories driven by AI. Automation can also free humans to think about new questions, and AI, even when giving wrong answers, can inspire humans to think about problems from more different dimensions.

When looking ahead to how to use AI to tackle difficult problems, this explorer in the materials field aimed his criticism directly at the "naysayers": whenever a new idea or new material is proposed, someone always obstructs it with doubts lacking data support, such as "it cannot be scaled up, cannot reduce costs, will not have economic benefits, and will not be reliable." This kind of emotional denial seriously delays the process of scientific research moving toward commercialization.

According to Omar Yaghi, his team is advancing some work from the laboratory to commercial application. For example, the MOF (metal-organic framework) materials developed by his team can extract water from the air. The industrial prototype is designed to produce 4,000 liters per day, and desert scenario testing has already been carried out to advance the commercialization of this "water from nothing" technology.

In an interview with a Daily Economic News reporter, Omar Yaghi said that every step in the commercialization of new materials is accompanied by a large number of experiments, trial and error, and choices, and these choices mostly rely on researchers' intuition and experience, lacking hard rules. This is precisely why the commercialization cycle for materials is so long. "The science of choice" is expected to greatly shorten this journey.

"In this process, AI is not only an accelerator, but can also connect different fields, enabling people to conduct holistic and integrated consideration," Omar Yaghi said. "The science of choice" is not just about making research and technology implementation faster, but about serving as a "connector" between different fields, forming cross-disciplinary dialogue from chemists to economists and posing transformable questions. "When different perspectives are connected through choice, magic happens."

We should cherish the unique value of human intuition and critical thinking

At the "Frontier Discipline Research and Cross-Domain Cooperation" forum, in the face of AI's strong intervention in generating hypotheses, analyzing data, and accelerating computation, participating scholars agreed that the threshold for solving problems is lowering, but "asking good questions" and "maintaining passion" are becoming more important.

"In the AI era, what is scarce? It is passion," pointed out Julio Cebral, academician of the Hong Kong Young Academy of Sciences and young academician of the Hong Kong Academy of Engineering Sciences. AI can quickly provide answers and offer new capabilities for solving problems, but passion and enthusiasm for science are the most indispensable fuel. He reminded young people that they must not rely on AI just to publish one more paper, and that scientific research cannot lack true love arising from "the pain of failure and human experience."

Konstantin Novoselov further stated that the significance of scientific exploration lies not only in obtaining a discovery, but also in providing a "stepping stone" for the next discovery. "I am not quite sure whether the discoveries brought by AI are complete, whether further exploration of the data is needed, or whether human wisdom is needed to interpret them. If AI is left unchecked, it may generate many hallucinations, as well as some knowledge that humans cannot understand. Whether this knowledge can promote human progress — these are questions that urgently need to be considered and resolved."

Ryad Chellali, academician of the Asian Academy of Sciences and senior scientist at the Italian Institute of Technology, warned about the AI black box problem: "AI is an irreversible system. We obtain results and digest knowledge, but because of the existence of the black box, we cannot reconstruct the intermediate knowledge system. Therefore, it is very important to study and explain this black box, and at the same time AI also needs to be shared globally."

At the "Scientific Iteration Reshaping Humanity's Future" forum, the participating guests looked ahead to the next decade and likewise outlined a common core vision: the "equalization" of technology and education.

Zheng Zijian, founding academician of the Hong Kong Young Academy of Sciences and young academician of the Hong Kong Academy of Engineering Sciences, most hopes to see that in the next decade all students worldwide can use AI tools for free to acquire knowledge, while Omar Yaghi set his sights on young people in developing countries, stressing that these people must not be left behind in AI development.

As Shao Bin, dean of the Zhongguancun Institute of Artificial Intelligence and professor at Beijing Zhongguancun College, summarized, in the future science will not only mean faster discovery, but also better questions, connecting disciplines and institutions, so that scientific progress ultimately creates lasting value for all humanity.

(Source: Daily Economic News) Middle of the body Embedded advertisement