Postdoctoral researchers Dr. Andrei Ermolaev from Université Marie et Louis Pasteur, Besançon, and Dr. Mathilde Hary from Tampere University published a study in Optics Letters showing how laser light inside thin glass fibers can simulate the way artificial intelligence (AI) processes information.
Image Credit: Doug McLean/Shutterstock.com
Consider a computer that employs light instead of electronics to complete tasks more quickly and effectively. The development of ultra-fast computers is now possible thanks to a collaboration between two research teams from Finland and France, which has demonstrated a revolutionary method of processing information utilizing light and optical fibers.
Inspired by neural networks, the study has examined a specific kind of computing architecture called an Extreme Learning Machine.
“Instead of using conventional electronics and algorithms, computation is achieved by taking advantage of the nonlinear interaction between intense light pulses and the glass,” explained Hary and Ermolaev.
Conventional electronics are getting close to their limits in terms of power consumption, bandwidth, and data throughput. Electronics can only handle data so quickly, and AI models are getting bigger and more energy-hungry. On the other hand, optical fibers may amplify minute differences through intense nonlinear interactions to make them noticeable and change incoming signals at hundreds of times quicker rates.
Toward Efficient Computing
In a recent study, researchers demonstrated the working concept of an optical ELM system using femtosecond laser pulses (a billion times shorter than a camera flash) and an optical fiber that confines light in a region smaller than a fraction of a human hair. The pulses are brief enough to include various wavelengths or colors.
By sending those into the fiber with a relative delay encoded according to an image, they demonstrate that the resulting spectrum of wavelengths at the fiber's output transformed by the nonlinear interaction of light and glass contains enough information to classify handwritten digits (such as those used in the popular MNIST AI benchmark).
According to the researchers, the top systems achieved more than 91% accuracy, comparable to cutting-edge digital approaches, in less than a picosecond.
Notably, the optimal outcomes came from a careful balancing act between fiber length, dispersion (the variation in propagation speed across wavelengths), and power levels rather than from the highest level of nonlinear interaction or complexity.
Performance is not simply matter of pushing more power through the fiber. It depends on how precisely the light is initially structured, in other words how information is encoded, and how it interacts with the fiber properties.
Dr. Mathilde Hary, Postdoctoral Researcher, Tampere University
This discovery might open the door to new computing methods and explore paths to more efficient designs by utilizing light’s potential.
“Our models show how dispersion, nonlinearity and even quantum noise influence performance, providing critical knowledge for designing the next generation of hybrid optical-electronic AI systems,” continued Ermolaev.
Advancing Optical Nonlinearity Through Collaborative Research in AI And Photonics
The proficiency of both research teams in nonlinear light-matter interactions is well known worldwide. Their partnership combines cutting-edge experimental skills with theoretical knowledge to utilize optical nonlinearity for various applications.
“This work demonstrates how fundamental research in nonlinear fiber optics can drive new approaches to computation. By merging physics and machine learning, we are opening new paths toward ultrafast and energy-efficient AI hardware,” added Professors Goëry Genty from Tampere University, John Dudley and Daniel Brunner from the Université Marie et Louis Pasteur, who led the teams.
The study investigates novel forms of computing by fusing applied AI with nonlinear fiber optics. Their long-term goal is to create on-chip optical systems that can function outside the lab and in real time. Real-time signal processing, environmental monitoring, and fast AI inference are possible uses.
The study is financed by the Research Council of Finland, the French National Research Agency, and the European Research Council.
Journal Reference:
Ermolaev, A. V., et al. (2025) Limits of nonlinear and dispersive fiber propagation for an optical fiber-based extreme learning machine. Optics Letters. doi.org/10.1364/OL.562186