About two years ago, Google's DeepMind announced the discovery of 2.2 million new crystalline materials using deep learning technology. Earlier this year, Microsoft claimed that its AI model MatterGen could generate inorganic materials from scratch, potentially revolutionizing the design paradigm for inorganic materials.

A new era of materials research driven by artificial intelligence (AI) seems to have begun, but criticism has followed. Critics argue that some AI-generated compounds lack originality and practicality. Will AI fundamentally transform the field of materials discovery, or will it become mere hype? A recent report on the website of the British journal *Nature* points out that most researchers acknowledge the enormous potential of AI in materials science, but it requires deep collaboration with experimental chemists, while also acknowledging and continuously improving upon current AI limitations to unleash its full potential.

AI-driven materials design boom

Before the introduction of AI, researchers primarily relied on the traditional computational method of density functional theory (DFT) to predict new materials and their properties. DFT has previously predicted high-quality new materials such as super-strong magnets and superconductors.

However, DFT computation is extremely demanding, and screening millions of compounds at once would be prohibitively expensive, highlighting the value of AI. DeepMind's "Graph of Materials Discovery Network" (GNoME) AI system discovered 2.2 million new crystalline materials in a single discovery, covering multiple elements from the periodic table, including 52,000 graphene-like layered compounds and 528 lithium-ion conductors that promise to improve rechargeable battery performance.

Lawrence Berkeley National Laboratory has developed the A-Lab robotic system. This system, by studying tens of thousands of papers on inorganic compound synthesis, has mastered formulation design capabilities and can synthesize compounds whose structures have been predicted by DFT but have never been prepared before. Simultaneously, A-Lab can control robots to perform experiments, analyze whether the products meet standards, and adjust the formulation to achieve closed-loop optimization when necessary.

Shortly after the publication of the GNoME and A-Lab papers, Microsoft launched the AI ​​tool MatterGen. Compared to GNoME, MatterGen is more targeted, directly generating materials that meet design requirements. Scientists can not only specify the material type but also set mechanical, electrical, and magnetic performance requirements, providing a powerful tool for precise research and development. Furthermore, the Basic AI team at Metaverse Platforms, in collaboration with Georgia Institute of Technology, focused on porous metal-organic frameworks (MOFs), predicting over 100 MOF structures with strong carbon dioxide adsorption, supporting the development of AI-accelerated direct carbon capture technology.

The debate between originality and practicality

Despite the strong momentum of exploration by industry giants, the controversy has never ceased. Many scientists have stated bluntly that some of the compounds envisioned by AI systems are neither original nor practically valuable.

After reviewing the hypothetical crystal list generated by DeepMind, materials scientists Anthony Chittam and others at the University of California, Santa Barbara, discovered that the AI ​​predicted over 18,000 compounds containing rare radioactive elements such as promethium and actinium, casting doubt on their practical value. Robert Palgrave, a solid-state chemist at University College London, also pointed out during his review of the A-Lab research results that some of the 41 inorganic compounds synthesized in the project had incorrect material descriptions, and some were even known materials that had already been synthesized.

In response, A-Lab laboratory personnel stated that further detailed analysis proved A-Lab's description of the material properties to be reliable, and that they did indeed synthesize the claimed compound. A spokesperson for DeepMind stated that the more than 700 compounds predicted by GNoME have been independently synthesized by other researchers, and that the model has also guided the discovery of several unknown cesium-based compounds, which hold promise for applications in optoelectronics and energy storage.

Microsoft's MatterGen has also been embroiled in controversy. During testing, the team asked MatterGen to recommend a new material with a specific hardness, and it synthesized a disordered compound called "tantalum chromium oxide." However, a preprint paper published in June of this year pointed out that this material was first prepared as early as 1972 and was even included in MatterGen's training data.

The collaboration between Metaverse Platforms and Georgia Institute of Technology has also been questioned. Berend Schmidt, a computational chemist at the Swiss Federal Institute of Technology in Lausanne (EPFL), confirmed through calculations that the new material proposed in the collaboration cannot achieve direct air capture. The model overestimates the material's ability to bind with carbon dioxide, partly due to errors in the underlying database used for training.

Practical application requires overcoming multiple hurdles.

Despite the controversy, most researchers still believe that with continuous optimization, AI models will powerfully drive progress in materials science.

To ensure the reliability of AI results, the Microsoft team developed MatterSim, an AI-assisted system specifically designed to verify the stability of structures proposed by MatterGen under real-world temperature and pressure conditions. However, even if AI-assisted material discovery proves effective, significant challenges remain: for example, how to optimize processes to meet market demands, and how to achieve large-scale manufacturing of new materials and integrate them into commercial products.

Citrine, an American informatics company, is helping clients optimize existing materials and manufacturing processes with its AI system. CEO Greg Mulholland stated that each client has a customized Citrine model, trained on proprietary experimental data and incorporating the "chemical intuition" of researchers to enhance AI judgment.

Undeniably, society's urgent need for new materials will continue to drive AI exploration in this field. Many of the major social challenges facing humanity today are ultimately constrained by material limitations. Scientists hope to leverage AI to design advanced materials that can be mass-produced and truly impact daily life, thus realizing the real value of AI in materials science.