How AI Is Transforming the Scientific System: Redefining and Cultivating Talent

We cannot simply assert that judgment, questioning, and innovation will forever remain exclusive to humans. In the future, AI will also participate in these activities. Defining talent should not merely involve listing tasks that AI currently cannot perform—and then treating those as permanent human advantages. Instead, we must focus on what an individual genuinely understands within new research environments, what contributions they can make, and whether—once a problem is identified—they can actively engage in tracing its origins and correcting it. Moreover, science relies not only on the rare few who propose groundbreaking theories. Reliable experimental execution, equipment maintenance, data verification, collaborative coordination, knowledge communication, and student mentoring all underpin scientific progress. While AI will transform the nature of these tasks, evaluation systems must still recognize diverse forms of contribution—not force everyone into a single, uniform standard of achievement. V. Talent cultivation must preserve space for exploration—and also reevaluate training content. When tools change, training content must naturally evolve too. Some tasks previously requiring manual effort can now be automated; thus, retaining them solely for the sake of ‘training’ is no longer justified. Repetitive labor does not inherently possess educational value—spending considerable time on it does not guarantee the development of essential capabilities. Whether a training activity deserves retention depends on whether it equips students to tackle unfamiliar problems—not merely on how labor-intensive it is. Yet individuals still need firsthand encounters with uncertainty: attempting judgments, discovering why those judgments were mistaken, and revising their understanding accordingly. Without such experiences, students may become familiar with many answers but fail to develop effective approaches for novel challenges. What truly merits preservation are those processes through which knowledge gradually becomes embodied capability. Therefore, future cultivation systems must ask more explicitly: Is this task assigned primarily to produce a result—or to foster a specific capability? If the main goal is achieving a result, tools—including AI—can be fully leveraged. If the goal is training, educators must thoughtfully decide when and how much assistance tools should provide, ensuring students remain authentically engaged in thinking. The same task may involve different AI usage strategies across developmental stages. For example, students might first formulate their own judgments before comparing them with AI-generated analyses; or they might modify a key condition in a problem and explain whether the original method remains applicable—and where reconsideration is needed. When AI provides a proof or experimental design, students must critically examine its underlying assumptions—not merely recite its conclusions. For certain foundational competencies, opportunities for independent, tool-free completion must still be preserved. As students mature, they can progressively assume greater responsibility in designing and leading complex human–AI collaborative research. Even checking AI outputs risks becoming perfunctory. Thus, assessing cultivation effectiveness cannot hinge solely on whether students complete a given task. More crucially, when confronted with a new problem, can they still mount reasoned attempts? Can they discern why a seemingly plausible answer is actually unreliable? And do they know how to verify it next? These indicators better reflect the very capabilities we aim to cultivate. VI. Adjusting only students and supervisors is insufficient to sustain a healthy cultivation system. These adjustments may sound straightforward—but in practice, the greatest obstacles often stem from evaluation mechanisms and competitive pressures. If a system continues rewarding rapid output above all else, those investing time in thoughtful, rigorous training will bear disproportionate opportunity costs.

In the AI era, what abilities should education prioritize cultivating in children?

The Second and Third Industrial Revolutions—driven by mechanization and electrification—propelled the United States to global leadership. Now, the Fourth Industrial Revolution—centered on artificial intelligence, big data, and the Internet of Things—is shaping the contours of the next round of global competition. In essence, AI’s impact on education amounts to a fundamental reconstruction of education’s underlying logic. If the Industrial Revolution liberated human physical labor, then the digital age’s AI is liberating human cognitive capacity. Tasks once demanding human brainpower—information storage, computation, language processing—are now efficiently handled by intelligent technologies. From simple information retrieval and language translation to sophisticated natural-language semantic analysis and protein-structure sequencing, AI’s capabilities continue expanding—even surpassing human performance in certain domains. This transformation not only reshapes how knowledge is produced and disseminated, but also imposes entirely new demands on education’s goals, content selection, and pedagogical methods. The interdependent triad of education, science & technology, and talent has grown increasingly inseparable in the intelligent era. Education cultivates talent; talent drives technological advancement; and technological progress, in turn, reshapes education itself. Today, central policy documents underscore education’s foundational and pioneering role more emphatically than ever. Faced with AI’s dual promise and challenges, education must not passively adapt—it must proactively transform. At the heart of this transformation lies one pivotal question: “How do we cultivate individuals capable of competing in—and leading—the future of science and technology?” After laying out this context, Professor Shen Jiliang offers his answer: No matter how technology evolves, education’s core mission remains the cultivation of human beings. Confronting AI’s challenges, education must return to nurturing core competencies—those irreplaceable, lifelong capacities that machines cannot replicate. Shen Jiliang identifies three pillars defining students’ core competencies in the intelligent era: “a sense of meaning, imagination, and adaptability.” 1. Sense of Meaning: Awakening and Pursuing Life’s Value A sense of meaning serves as the intrinsic motivation behind students’ learning and growth—centered on answering fundamental questions such as “Why do I exist?”, “What value does my existence hold?”, and “What can I contribute to society?” Phenomena like student ‘lying-flat’ attitudes and rising youth suicide rates reflect, at their root, a crisis of meaning—where learning serves only to appease parents and teachers, university admission aims solely at securing lucrative jobs, and employment exists merely to earn money—without ever probing the deeper values underpinning these pursuits. In the intelligent era, machines will displace increasing volumes of repetitive labor, shifting human value toward the spiritual realm and societal contribution.

How should we train chemists in the age of AI?

1 day ago

Franz M. Geiger argues that excessive reliance on AI in chemical research risks preventing young scientists from developing essential independent judgment and critical thinking skills through traditional trial-and-error learning.