L'IA peut-elle inventer de nouveaux matériaux à ajouter au tableau périodique ?

Les systèmes d'IA actuels excellent dans la modélisation de structures chimiques hypothétiques et la prédiction d'isotopes stables, mais aucun ne peut « découvrir » et nommer un nouvel élément au sens formel de l'UICPA — les éléments doivent être synthétisés dans des laboratoires d'accélérateurs et vérifiés par des observations expérimentales répétées avant d'être officiellement ajoutés au tableau périodique. Les modèles récents d'apprentissage automatique (par ex., GNoME) accélèrent l'énumération de composés inorganiques stables auparavant inconnus, mais il s'agit de matériaux étendus plutôt que de nouveaux éléments qui nécessiteraient de modifier le tableau lui-même. Ainsi, bien que l'IA augmente les pipelines de découverte, elle reste un outil d'assistance ; seule la physique nucléaire expérimentale peut étendre le tableau périodique.
SOURCE : Union internationale de chimie pure et appliquée — https://iupac.org

— Enriched May 13, 2026

Background

Current AI systems excel at modeling hypothetical chemical structures and predicting stable isotopes, yet none can “discover” and name a new element in the formal IUPAC sense—elements must be synthesized in accelerator laboratories and verified through repeated experimental observation before official addition to the periodic table (International Union of Pure and Applied Chemistry — https://iupac.org). Recent machine-learning models (e.g., GNoME) accelerate the enumeration of previously unknown stable inorganic compounds, yet these are extended materials rather than new elements that would require altering the table itself. Thus, while AI augments discovery pipelines, it remains an assistive tool; only experimental nuclear physics can expand the periodic table.

Researchers use AI to screen potential new materials and predict their behavior under various conditions, which can help focus experimental efforts. AI can assist in the discovery of new materials by predicting their properties and behavior, but it cannot independently invent new elements to add to the periodic table. The process of discovering new elements involves complex experiments and verification by the scientific community. AI can, however, help scientists identify potential new materials and their properties by analyzing large datasets and running simulations. This can accelerate the discovery process, but human scientists are still necessary to design and conduct experiments to verify the existence and properties of new materials. The addition of new elements to the periodic table is overseen by the International Union of Pure and Applied Chemistry (IUPAC), which ensures that new elements meet strict criteria for recognition. AI's role in materials science is rapidly evolving, and it is likely to play an increasingly important role in the discovery of new materials in the future.