In a significant leap forward for computing technology, European researchers have successfully demonstrated how light, rather than electricity, can be harnessed to perform artificial intelligence computations at unprecedented speeds.
The groundbreaking research, conducted by Dr. Mathilde Hary from Finland's Tampere University and Dr. Andrei Ermolaev from France's Université Marie et Louis Pasteur, shows how intense laser pulses traveling through ultra-thin glass fibers can mimic the way AI processes information, but thousands of times faster than conventional electronic systems.
The researchers utilized a computing architecture known as an Extreme Learning Machine (ELM), which is inspired by neural networks. Their approach takes advantage of the nonlinear interaction between intense light pulses and glass to perform complex computations. When tested on the MNIST handwritten digit dataset, their optical system achieved impressive accuracy rates exceeding 91% in anomalous dispersion regimes and 93% in normal dispersion regimes.
"This work demonstrates how fundamental research in nonlinear fiber optics can drive new approaches to computation," explained Professors Goëry Genty and John M. Dudley, who supervised the research. "By merging physics and machine learning, we are opening new paths toward ultrafast and energy-efficient AI hardware."
The innovation addresses critical limitations of traditional electronics, which are approaching their physical limits in terms of bandwidth, data throughput, and power consumption. As AI models continue to grow exponentially—doubling in size approximately every 3.5 months according to OpenAI research—the energy demands for training and running these models have become increasingly unsustainable.
Potential applications for this light-based computing technology range from real-time signal processing and environmental monitoring to high-speed AI inference. The researchers aim to eventually build on-chip optical systems that can operate in real-time outside laboratory settings, potentially revolutionizing data centers, autonomous vehicles, and other AI-intensive applications.
The project, funded by the Research Council of Finland, the French National Research Agency, and the European Research Council, represents a fundamental shift in computing paradigms that could help address the growing energy crisis in AI computing while simultaneously enabling more powerful and responsive AI systems.