At the Intelligent Manufacturing Research Institute of Hefei University of Technology, staff are debugging an AI chemical management robot.

The Chinese Academy of Sciences has released the “Panshi 100” model system, which targets eight major disciplines to create a large model cluster across fields.

Currently, AI is intervening in scientific research with unprecedented breadth and depth, from predicting protein structures to discovering new materials. AI has become a “universal engine” for accelerating science, showcasing the immense potential of intelligent scientific paradigms.

As a new partner for researchers, how is AI changing the paths and rhythms of scientific research? How can we use AI responsibly and effectively? How can we stimulate the role of open scientific intelligence platforms? In this educational edition, we invite several experts to discuss these questions.

1. How is the path of scientific discovery changing?

Traditional research begins with “hypothesis-validation,” but now, the path of scientific discovery is gradually shifting to “data-pattern discovery-intelligent generation-closed-loop iteration.”

Wang Xijun, a distinguished professor at the University of Science and Technology of China, states that in traditional research, researchers often propose questions based on experience and intuition. However, for some disciplines, AI can proactively discover patterns in vast amounts of data. The new paradigm of scientific discovery is evolving into “data-pattern discovery-intelligent generation-closed-loop iteration,” where AI can even design desired materials according to specific goals.

For example, in my research on framework materials, these materials can be created through combinations of different metal nodes, organic ligands, and connection methods, resulting in trillions of structures, far exceeding human exploration limits. In this context, AI provides a breakthrough. On one hand, machine learning can quickly predict material properties, saving significant trial-and-error costs in real experiments. On the other hand, AI can extract patterns from data, transforming past experiential “intuition” into computable and transferable models, making material design more rational.

Based on this, generative AI can further push research from “filtering the known” to “creating the unknown”—directly generating new material structures beyond the training data, achieving “reverse design” around target performance. This means AI is not only accelerating problem-solving but also expanding the boundaries of the problems themselves.

Thus, the role of AI in research is continuously evolving: from an initial computational tool to an analytical research assistant, and now to a “research partner” that can participate in and even drive autonomous exploration.

Of course, AI will not replace scientists. Understanding key scientific questions and mechanisms still relies on human judgment and insight. Humans are responsible for posing questions and guiding direction, while AI searches for possible answers within vast data and complex spaces. The collaboration between the two will provide a more solid and expansive space for future scientific innovation.

2. Has the efficiency of scientific innovation improved?

AI excels at handling tasks with clear answers that require extensive repetitive calculations.

Mo Bofeng, a professor at the Oracle Bone Research Center of Capital Normal University, explains that AI significantly enhances research efficiency in literature review, experimental design, and data analysis. Even when dealing with oracle bones from over 3,000 years ago, AI can play a substantial role. Tasks like oracle bone splicing (reassembling broken bones) and restoration (recovering missing images) previously relied on a few experts’ experience. Now, AI offers new solutions.

To truly harness AI’s potential, it’s crucial to identify the right integration points. As oracle bones are archaeological documents, the core research goal is to restore textual materials and information, and AI is particularly adept at handling tasks with clear answers and extensive repetitive calculations. It can identify subtle features that humans may overlook, such as the curvature of break edges and the angle of brush strokes, providing key clues for splicing and restoration.

However, AI is not omnipotent. The total number of oracle bones exceeds 160,000, with over a million characters, which may seem substantial but is still insufficient for training large AI models. Therefore, human experts are still needed for deep semantic judgments. A more effective approach is human-machine collaboration: using AI as a speed tool while having experts review and correct its results.

Currently, splicing and restoration are just the beginning of AI-assisted oracle bone research. As technology advances, tasks like classification, aggregation, and translation of oracle bones will gradually break through. Future researchers will need to not only understand their specialized knowledge but also enhance their data processing capabilities and effectively leverage technology to amplify their research advantages.

3. Will AI influence scientific judgment?

While lowering some research barriers, risks such as false citations and erroneous reasoning deserve attention.

Yang Yaodong, a researcher at Peking University’s Institute of Artificial Intelligence, notes that AI is not just helping researchers write code, review literature, and create charts; it is transforming the entire research process: from humans proposing hypotheses, conducting experiments, and analyzing results in a linear flow, to a closed-loop system of human-machine collaboration, model prediction, automated experiments, and feedback iteration.

This change brings several benefits. First, efficiency has significantly improved. In fields like materials, drugs, and energy, there are numerous candidate solutions, making it challenging for traditional methods to cover all possibilities. AI can quickly filter options, liberating researchers from repetitive trial-and-error to focus on critical issues. Second, it promotes interdisciplinary integration, as a scientific problem often involves physics, chemistry, biology, engineering, and computation, and AI can establish connections between multi-source data. Third, it lowers some research barriers; with open-source models and tool platforms, small teams can undertake large projects.

However, it is important to note that AI does not equate to genuine scientific understanding. Scientific research must not only be accurate in predictions but also answer “why.” If a model is a black box, data sources are unclear, and experimental processes are non-reproducible, the conclusions drawn by AI may introduce new risks. Especially with generative AI, issues like false citations, erroneous reasoning, low-quality papers, data leaks, and unclear academic responsibilities could undermine research norms.

A deeper issue is that scientific judgment cannot be replaced by tool logic. AI excels at finding optimal solutions within existing data, but determining which questions are worth researching and which results hold scientific significance still requires human oversight.

4. How can resources be effectively integrated?

Connecting scientists, AI engineers, and industry forces to shift innovation from isolated breakthroughs to systematic acceleration.

Wu Libo, assistant president of Fudan University and chairman of the Shanghai Institute of Scientific Intelligence, states that scientific intelligence is transitioning from a “technology-centered” 1.0 era to a “scientist-centered” 2.0 era. The 2.0 era aims to make more scientists the main characters, allowing AI to truly permeate the entire research process. The Star River Qizhi Scientific Intelligence Open Platform, co-created by the Shanghai Institute of Scientific Intelligence and Fudan University, is a response to this shift.

The platform’s primary role is to lower the barriers for scientists to use AI. It builds a complete infrastructure covering data, models, computing power, experiments, intelligent agents, and collaborative communities around real research paths. Currently, the Star River Qizhi platform has gathered over 400 scientific models and tools, 22PB (petabytes) of high-value data, and 500 million literature patents, allowing scientists to conduct research without delving into technical details.

We have also launched a research intelligent agent system based on “Dasheng.” It can understand scientific problems and assist in completing the entire process from literature analysis and hypothesis generation to experimental validation. Recently, “Dasheng” introduced a customizable laboratory feature, enabling scientists to build dedicated toolchains based on their research directions.

The second role of the platform is to promote interdisciplinary, interregional, and cross-domain integration. In traditional research, data, models, and methods from different disciplines often do not communicate, making collaboration difficult. The Star River Qizhi platform facilitates sharing, reuse, and combination of results across different fields through a unified model repository and data infrastructure.

On a deeper level, the platform serves as a hub for the scientific intelligence ecosystem. It connects scientists, AI engineers, and industry forces, allowing data and methods to flow and be reused within the system, shifting innovation from isolated breakthroughs to systematic acceleration, providing sustainable institutional support for AI-driven research paradigm transformation.

5. How to build and utilize intelligent platforms effectively?

Encouraging open sharing to bridge the gap between industry and research.

Liu Tieyan, president of Beijing Zhongguancun College and chairman of the Zhongguancun Artificial Intelligence Research Institute, emphasizes that having many platforms does not equate to them being sufficient, user-friendly, or genuinely useful. Last year, Zhongguancun College surveyed over 30 materials companies in Beijing and identified 100 “bottleneck” issues. The research found that with current mainstream scientific intelligence technologies, only 20% of these problems are likely to be solved. The remaining issues are temporarily unsolvable due to low digitalization levels, data deficiencies, and insufficient algorithm accuracy in companies. This realization highlights that “AI empowering research” cannot just be a slogan; the gaps in infrastructure, technological limitations, and the industry-research divide are real.

Moreover, the open sharing of scientific intelligent agents and tools appears to be a technical issue on the surface, but at a deeper level, it is a lack of motivation to connect. Why would an institution want to open its data and platform? If this question lacks a systemic answer, “open sharing” will remain at the level of advocacy.

To break the deadlock, it is suggested to focus on three areas: first, vigorously promote industrial digitalization, allowing genuine industrial needs to guide scientific research directions. Research cannot remain in a “research first, then transform” model; industrial feedback must enter the research cycle to fill the “last mile.” Second, establish incentive mechanisms for open sharing, recognizing sharing as a contribution to research, such as making it a condition for project initiation and completion, and creating a metric system similar to citation metrics. Third, public forces should take the lead in building foundational infrastructure for cross-disciplinary collaboration. Users of scientific intelligent agents and tools are highly specialized and dispersed across disciplines. Due to insufficient market size, national strategic investment could be considered first, gradually introducing market mechanisms.

In summary, bridging the data and intelligent agent interfaces is a surface issue; restructuring incentive mechanisms is a middle-level concern; and fundamentally, aligning research with national needs and genuine industry problems is essential.

Link

The call for papers stating “the first author must be AI” has sparked heated discussions.

“The first author must be AI.” In 2025, a call for papers issued by East China Normal University stirred significant debate in the academic community. This social experiment, requiring AI to be the main author of research papers, serves as an extreme test to confront a critical question: when AI deeply engages in knowledge production, where are the ethical boundaries of AI-assisted writing, and what should be the bottom line of academic research?

“We hope to explore the public acceptance, technical feasibility, scientific quality, and academic norms of AI writing through this approach,” said Yuan Zhenguo, a lifelong professor at East China Normal University and the initiator of the experiment.

Following the announcement, controversy ensued. Supporters viewed it as a “breakthrough experiment” for academic norms in the AI era, while opponents worried it represents a “proactive retreat” of humanity in research. Zhang Zhi, director of the Intelligent Education Laboratory at East China Normal University, stated, “Currently, the penetration rate of AI in papers is high, and many students use AI to assist in writing but do not dare to disclose it. This ‘underground state’ is a greater threat to academic norms than anything else. Rather than turning a blind eye, we should respond directly.”

The experiment collected 820 research papers with “AI as the first author.” Reviewers found that AI demonstrated good capabilities in topic planning, outline generation, data analysis, literature speed reading, and logical organization. However, limitations are also significant: large models excel at “fragment reorganization and cross-domain transfer” of existing data, generating “plausible” innovative texts but lacking genuine creativity and value judgment.

“Based on this underlying logic, the reasonable application scenarios for AI in research writing should focus on non-core aspects,” Zhang Zhi stated. In paper writing, humans should assume the roles of problem proposers, tool selectors, instruction designers, and quality controllers.

“The bottom line for AI usage fundamentally concerns academic integrity and responsibility attribution. The originality bottom line cannot be breached, and the transparency bottom line must be upheld—any use of AI must be fully disclosed, specifying the tool’s name, application scope, and human review process in the paper. Moreover, the responsibility attribution bottom line cannot be ambiguous; regardless of the extent of AI involvement, human authors must bear full responsibility for the final results,” Zhang Zhi emphasized.

The significance of this experiment may not lie in drawing conclusions but in fostering a consensus: as collaboration between humans and AI in paper writing becomes a new phenomenon, only by effectively utilizing AI and upholding academic integrity can we safeguard the authentic value of academic research.

“Humans using AI to assist in paper writing does not mean relinquishing subjectivity but rather exploring a new division of labor in research, allowing AI to handle the breadth of data while humans maintain the depth of thought and the warmth of values,” said Chu Xiaobo, vice president of Peking University.