In a new study Bar-Ilan University Prof. Ido Kanter examines how the classic idea of “more is different” applies to AI, finding a contrast between physics and artificial intelligence: from an information perspective, physics behaves as “more is the same,” while AI truly reflects “more is different.” His research shows that as AI models learn, their internal nodes become specialized, each taking on distinct roles such as recognizing patterns or language features. This specialization, combined with cooperation among nodes, allows the system to achieve capabilities greater than the sum of its parts—demonstrating that AI’s power comes not just from scale but from coordinated, emergent intelligence.

A team led by scientists at the National Graphene Institute (NGI) at The University of Manchester developed the first technique capable of capturing atomic‑resolution videos of individual gold atoms ‘dancing’ across a surface surrounded by liquid, opening a window into a hidden atomic world that has been invisible until now.

Sanket Samal and colleagues report a method for assembling a conductive polymer (CP) within embryonic and brain tissues of living zebrafish and mice, using natural blood proteins and whole blood as catalysts. The researchers show that these in vivo-assembled polymers, interfacing with neurons, can be targeted with near-infrared light to selectively and reversibly control neuronal activity in living mice. The blood-catalyzed system offers “a promising pathway to seamlessly integrate electronics into living tissues by creating functional synthetic CPs directly within biological systems,” write Samal et al. Researchers are interested in expanding the use of CPs for bioelectronic devices, because they are biocompatible and stable in biofluid environments and offer precise electronic communication needed for neuromodulation. Building these polymer interfaces inside the body from the start could increase their biocompatibility and reduce the use of residual, potentially toxic catalyst materials. Samal et al. demonstrated this approach by creating n-doped poly(benzodifurandione) (n-PBDF) with hemoprotein catalysts, resulting in stable and ionically sensitive CP networks. Guglielmo Lanzani and Maria Rosa Antognazza discuss how this technique could be optimized for neuromodulation and regenerative medicine in a related Perspective.