Two years ago, Google's Deep Thinking company announced the discovery of 2.2 million new crystal materials using deep learning technology. At the beginning of this year, Microsoft claimed that its AI model MatterGen could generate inorganic materials from scratch, which is expected to overturn the paradigm of inorganic material design. The new era of artificial intelligence (AI) driving material research seems to have begun, but criticism has also followed. Critics argue that some of the compounds envisioned by AI lack originality and practicality. Will AI completely change the field of material discovery, or will it become excessive hype? A recent report on the website of Nature in the UK pointed out that most researchers recognize the enormous potential of AI in materials science, but it requires deep collaboration with experimental chemists, while recognizing the current limitations of AI and continuously improving it in order to unleash its full potential. Before AI intervention, researchers mainly relied on the traditional computational method of "density functional theory" (DFT) to predict new materials and their properties. DFT has predicted high-quality new materials such as super magnets and superconductors. But DFT has a huge computational load, and the cost of screening millions of compounds at once is unimaginably high, highlighting the value of AI. The "Material Exploration Graph Network" (GNoME) AI system developed by DeepMind has discovered 2.2 million new crystal materials at once, covering various elements of the periodic table, including 52000 layered compounds similar to graphene and 528 lithium-ion conductors that are expected to improve the performance of rechargeable batteries. Lawrence Berkeley National Laboratory in the United States has developed the A-Lab robot system. The system has mastered the ability of formula design by studying tens of thousands of inorganic compound synthesis papers, and can synthesize compounds whose structures have been predicted by DFT but have never been prepared before. At the same time, A-Lab can control robots to perform experiments, analyze whether the products meet the standards, and adjust the formula if necessary to achieve closed-loop optimization. Shortly after the publication of GNoME and A-Lab papers, Microsoft launched the AI tool MatterGen. Compared to GNoME, MatterGen is more targeted and can directly generate materials that meet design conditions. Scientists can not only specify material types, but also set performance requirements such as mechanical, electrical, and magnetic properties, providing powerful tools for precise research and development. In addition, the basic AI team of the metaverse platform company has collaborated with the Georgia Institute of Technology to focus on "metal organic frameworks" (MOFs) porous materials, predicting over 100 MOF structures that strongly adsorb carbon dioxide, providing support for AI to accelerate the development of direct air capture carbon technology. Despite the strong exploration momentum of industry giants, the debate between originality and practicality has never ceased. Many scientists have bluntly stated that some of the compounds envisioned by AI systems lack originality and practical value. Anthony Chitam, a materials scientist at the University of California, Santa Barbara in the United States, and others browsed through the list of hypothetical crystals for deep thinking and found that the more than 18000 compounds predicted by their AI contain rare radioactive elements such as promethium and actinium, and their practical value is questionable. Robert Palgrave, a solid-state chemist at University College London, also pointed out during the verification of A-Lab research results that some of the 41 inorganic compounds synthesized in the project had incorrect material descriptions, and even had known materials that had already been synthesized. In response, A-Lab laboratory personnel confirmed through detailed reanalysis that A-Lab's description of material properties is reliable and indeed synthesized the claimed compound. A spokesperson for DeepMind stated that over 700 compounds predicted by GNoME have been independently synthesized by other researchers, and the model has also led to the discovery of several unknown cesium based compounds, which are expected to be used in the fields of optoelectronics and energy storage. Microsoft's MatterGen is also embroiled in controversy. When the team tested, they were asked to recommend a new material with a specific hardness, which synthesized a disordered compound called "tantalum chromium oxide". But a preprint paper in June this year pointed out that this material had been first prepared as early as 1972 and even included in MatterGen's training data. The collaboration project between the metaverse platform company and Georgia Institute of Technology has also been questioned. Swiss Federal Institute of Technology Lausanne computational chemist Berend Schmidt confirmed through calculations that the new material proposed in the collaborative project cannot achieve direct air capture, and the model overestimated the binding ability of the material with carbon dioxide, partly due to errors in the basic database used for training. Although there is controversy over the need to overcome multiple obstacles in practical applications, most researchers still believe that after continuous optimization, AI models will effectively promote the progress of materials science. To ensure the reliability of AI results, the Microsoft team developed the auxiliary AI system MatterSim, specifically to verify whether the structure proposed by MatterGen is stable under real temperature and pressure conditions. But even if AI assisted material discovery is proven effective, humans still face huge challenges, such as how to optimize processes according to market demand, and how to achieve large-scale manufacturing of new materials and integrate them into commercial products. Citrine's AI system is helping customers optimize existing materials and manufacturing processes. The CEO of the company, Greg Muholland, stated that each client has a customized Citrine model, which is trained based on proprietary experimental data and incorporates the "chemical intuition" of R&D personnel to enhance AI judgment. It cannot be denied that the urgent demand for new materials in society will continue to drive the exploration of AI in this field. The many major social challenges currently facing humanity are all constrained by material bottlenecks. Scientists look forward to using AI to design advanced materials that can be mass-produced and truly impact daily life, so that the value of AI in the field of materials science can truly be realized. (New Society)