Materials with nearly identical atomic structures can exhibit strikingly different properties, a puzzle in glass physics. According to the researchers, from China and Denmark, the key lies not in geometry but in chemical-bonding heterogeneity. Deep-learning simulations and bond-order analysis reveal how electronic interaction variations regulate atomic mobility and β relaxation in Pd-based metallic glasses, establishing bonding heterogeneity as a fundamental driver of structure–relaxation coupling.

Inspired by Pavlov’s classical conditioning, researchers propose a bio-inspired optical neural network trained via associative learning. Using a dual-color photoresist, sequential UV and visible light exposure encodes memory directly into the material’s fluorescence response, enabling in-situ, computation-free training for pattern recognition—bypassing conventional backpropagation and offering a scalable route to low-cost, edge-compatible photonic AI hardware.

Sergei Kalinin, Weston Fulton Professor of Materials Science and Engineering in the University of Tennessee, Knoxville’s Tickle College of Engineering, has received a 2026 Southeastern Conference Faculty Achievement Award. One of the SEC’s highest honors, the award recognizes excellence in teaching, research and service.

JMIR Publications today announced the release of three pivotal articles in its News and Perspectives section, highlighting the dual-edged nature of technology in modern society

Games are often called the ‘Formula 1’ of AI: clear rules, clear winners. AlphaZero-style algorithms learn by self-play: a neural network predicts moves and guides tree search. We tested this recipe on Nim, a simple children’s matchstick game that has been mathematically solved. Because the correct move is known for every position, we can measure whether an agent plays optimally across the state space. We find a gap: learning can work on small boards, but blind spots remain and performance degrades as the board grows, with predictions approaching random. This suggests impartial games often need analytic representations, not pattern learning.

An international AI weather competition that sees teams from all over the world compete to produce the most accurate three to four-week weather forecasts has unveiled its latest results (13 March 2026).