Electrocatalysis sits at the heart of clean hydrogen production, fuel cells, and carbon dioxide conversion, yet progress toward scalable, high-performance catalysts has remained frustratingly slow. A growing body of research now suggests that artificial intelligence (AI) may be key to breaking this bottleneck—but only if it is used wisely. By reviewing three decades of AI applications in electrocatalysis, researchers reveal how the field has shifted from isolated data analysis toward end-to-end, data-driven discovery. The work highlights a critical turning point: AI is no longer just accelerating experiments, but beginning to reshape how electrocatalysts are designed, evaluated, and understood at a fundamental level.

A research team led by Pengfei Guan from Advanced Interdisciplinary Science Research (AiR) Center, Ningbo Institute of Materials Technology and Engineering (CAS), has published a topic review focusing on harnessing machine learning for metallic glasses (MGs). Through an integrated assessment of pre-Kepler-era studies, researchers have identified that the alignment gap, between simulation and experiment, currently hinders the further application of machine learning. Tailored for closing theory-experiment loop, a revolutionary AI4S (AI for Science) framework based on domain-knowledge embedding is prototyped. Actionable strategies and quantifiable metrics are also prepared to navigate the data pipelines and reshape the research paradigm from “accelerated trial-and-error” into “truly unbiased and intelligent discovery”.

Conductive gels are central to flexible electronics and wearable sensing technologies, yet their practical use is often constrained by weak mechanical strength and poor environmental tolerance. In a recent study published in Journal of Bioresources and Bioproducts, researchers report a wood-based conductive eutectogel fabricated by infiltrating metal-based deep eutectic solvents into a natural wood skeleton, followed by initiator-free ultraviolet polymerization. The resulting material exhibits a rare combination of high tensile strength, stable ionic conductivity, and broad operational temperature tolerance. By leveraging wood’s aligned micro–nano channels and metal–polymer coordination chemistry, the work demonstrates a sustainable pathway toward high-performance conductive gels capable of operating under extreme environmental conditions.

Take a dive into the streamlined workflow of DIVE (Descriptive Interpretation of Visual Expression)! DIVE combines multiple AI agents to extract images from about 4,000 scientific publications to propose new materials – all within minutes.

InstaDrive, proposed by SJTU researchers, addresses autonomous driving’s tedious annotation and long-tail data issues. It projects 3D vehicle bounding boxes and BEV map elements into 2D instance segmentation as control conditions, ensuring multi-view consistency via unified occlusion modeling and an order-invariant encoder. Tested on nuScenes, it outperforms baselines in FID and mAP, enabling precise editing of vehicles/map elements and efficient labeled data generation.