Dresden 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 4: Focus: Deep Learning in Electromagnetics Research

AKPIK 4.1: Hauptvortrag

Dienstag, 10. März 2026, 14:00–14:30, BEY/0127

Beyond the Hype: Can AI Truly Transform Photonics? — •Mehdi K. Hedayati — Durham University, Department of Engineering, Durham DH1 3LE, UK

AI has become a powerful enabler in photonics, offering new routes for accelerated design and materials discovery beyond conventional trial-and-error approaches. In our group, we have demonstrated several AI-driven photonic design frameworks, with selected results experimentally verified in the laboratory. These include AI-assisted metasurface design enabling direct mapping between geometry and structural colour*allowing multiple colours from a single geometry under strain*as well as machine-learning models that capture nonlinear relationships in amorphous metamaterials, leading to proof-of-concept devices such as metasurface perfect reflectors. Yet, alongside these successes, the field is increasingly shaped by exaggeration and misplaced expectations. In reality, AI rarely delivers deployable photonic technologies; instead, it proposes candidate materials, geometries, or structures that must still be fabricated and validated experimentally. Fabrication yield, scalability, and limited characterization throughput therefore remain the dominant bottlenecks, causing progress to saturate at the experimental stage. In this talk, we will review the current state of the art of AI in photonics, examine where the field is heading in the near future, identify the key limitations we are likely to face, and discuss realistic pathways forward to better align AI-driven design with experimental reality.

Keywords: AI; Photonics; AI Hype; Nano; Metamaterials