A new study shows that large language models— the same kind of AI behind chat tools— can act like a chemistry mentor, helping scientists design and test new molecules one logical step at a time. Researchers at EPFL, led by Philippe Schwaller, taught the AI to read and interpret chemical language, recognize functional groups, evaluate individual reactions, and even map out whole synthetic routes. The AI doesn’t create structures from scratch; instead, it reviews the work of traditional computer‑based chemistry programs, pointing out the most plausible pathways and flagging unlikely ones. Bigger, more advanced models performed best, while smaller versions struggled with the nuanced reasoning required. By providing a human‑like intuition for reaction mechanisms— the detailed electron‑shuffling that drives chemistry— the system can cut down on costly trial‑and‑error experiments, speed up discovery, and make the design process more transparent for researchers. The findings, published in the journal *Matter*, suggest that AI could soon become a routine partner in labs, offering step‑by‑step guidance that bridges the gap between raw data and chemical insight.
Read moreMirror World’s new “Seamless Positioning + Real‑time Twin” platform blends ultra‑precise location tracking with live digital‑twin replication, creating a single‑screen view of any physical environment. In Shenzhen’s Longgang district, the system links traffic, security and municipal data, letting officials monitor people and vehicles in real time and instantly simulate emergencies. The result? Government approval times fell 30 % and emergency‑response speed jumped 90 %, turning the district into a national smart‑city showcase. The same technology is reshaping factories. At a leading communications plant in Binjiang, the platform duplicated 60 SMT lines and 10 assembly lines, synchronising material flow, staff movement and equipment status. Production efficiency rose 42 %, energy use per unit dropped 66 %, and defect rates were cut nearly in half. These gains have been replicated across multiple heavy‑industry sectors. Beyond cities and factories, Mirror World’s engine powers “video twins” – ultra‑fast 3‑D reconstructions using NeRF 3.0 that render scenes in 0.2 seconds, while a custom AI chip delivers 128 TOPS/W for low‑power, high‑performance edge computing. Multi‑source data fusion turns raw video, IoT and GIS feeds into actionable insights, breaking data silos across tourism, cultural heritage and public safety. Together, these capabilities position Mirror World at the forefront of the digital‑twin revolution.
Read more