A new study unveils how training neural networks directly with light on photonic chips could slash energy use and speed up AI, paving the way for smarter on-device systems.
Review: Training of physical neural networks. Image Credit: sdecoret / Shutterstock
Artificial intelligence is now an integral part of our daily lives, resulting in a pressing need for larger and more complex models. However, the demand for ever-increasing power and computing capacity is rising faster than the performance traditional computers can provide.
To overcome these limitations, research is moving towards innovative technologies such as physical neural networks, analogue circuits that directly exploit the laws of physics (the properties of light beams and quantum phenomena) to process information. Their potential is at the heart of the study published by the prestigious journal Nature. It is the outcome of collaboration between several international institutes, including the Politecnico di Milano, the École Polytechnique Fédérale in Lausanne, Stanford University, the University of Cambridge, and the Max Planck Institute.
Advancing Training for Physical Neural Networks
The article entitled "Training of Physical Neural Networks" discusses the steps of research on training physical neural networks, carried out in collaboration with Francesco Morichetti, Professor at DEIB – Department of Electronics, Information and Bioengineering, and Head of the university's Photonic Devices Lab.
Photonic Chips: Politecnico di Milano's Role
Politecnico di Milano contributed to this study by developing photonic chips for creating neural networks, utilizing integrated photonic technologies. Mathematical operations, such as addition and multiplication, can now be performed through light interference mechanisms on silicon microchips that are barely a few square millimeters in size.
Energy Efficiency Through Light-Based Computation
"By eliminating the operations required for the digitisation of information, our photonic chips allow calculations to be carried out with a significant reduction in both energy consumption and processing time," says Francesco Morichetti. A step forward to make artificial intelligence (which relies on extremely energy-intensive data centres) more sustainable.
In-Situ Training with Light
The study, published in Nature, addresses the theme of training, specifically the phase in which the network learns to perform certain tasks. «With our research within the Department of Electronics, Information and Bioengineering, we have helped develop an "in-situ" training technique for photonic neural networks, i.e., without going through digital models. The procedure is carried out entirely using light signals. Hence, network training will not only be faster, but also more robust and efficient», adds Morichetti.
The use of photonic chips will enable the development of more sophisticated models for artificial intelligence, or devices capable of processing real-time data directly on-site – such as autonomous cars or intelligent sensors integrated into portable devices – without requiring remote processing.
Source:
Journal reference:
- Momeni, A., Rahmani, B., Scellier, B., Wright, L. G., McMahon, P. L., Wanjura, C. C., Li, Y., Skalli, A., Berloff, N. G., Onodera, T., Oguz, I., Morichetti, F., Del Hougne, P., Le Gallo, M., Sebastian, A., Mirhoseini, A., Zhang, C., Marković, D., Brunner, D., . . . Fleury, R. (2025). Training of physical neural networks. Nature, 645(8079), 53-61. DOI: 10.1038/s41586-025-09384-2, https://www.nature.com/articles/s41586-025-09384-2
AI Text Generation Evolves Beyond Templates to Transform Communication