Mohammedia – A new scientific breakthrough is pushing artificial intelligence closer to operating at the speed of light. Two recent studies by international research teams have demonstrated that key AI calculations can be performed using light waves instead of electronic circuits, opening the door to faster and more energy-efficient computing.
Researchers published a major paper in the journal Nature Photonics explaining how they created a method that allows light itself to carry out complex mathematical operations.
These operations, known as tensor calculations, are the foundation of modern AI systems.
Instead of relying on traditional electronic chips that process one step after another, the new approach uses coherent light to process large blocks of data all at once as it moves through an optical setup.
The team showed that this process can match the accuracy of GPU-based calculations while using far less energy.
This could reshape how AI systems are built. The method enables the occurrence of several matrix operations in one single pass of light, removing common bottlenecks that slow down neural networks.
Because the physics of light allows large-scale parallel processing, the system does not suffer from heating issues or memory limits, which are common issues with today’s hardware.
Read Also: Google Introduces Gemini 3, Its Most Advanced AI Model Yet
The researchers tested the method on real neural network tasks involving both real and complex numbers and found that the results were consistent with traditional computing.
AI experts say this development could mark the beginning of a new generation of optical computers.
These machines would not replace electronic systems overnight, but could be used to power increasingly demanding AI models that require massive processing power.
Aalto University researchers, who published related work, said that practical optical computing hardware could appear within three to five years if development continues at the current pace.
They noted that optical systems could be especially useful for real-time applications like robotics, video analysis, and large language models that need extremely fast data flows.
The technology remains in an early stage and is still limited to laboratory conditions. Scaling the optical setup, improving stability, and making it compatible with commercial chips will require further engineering work.
Even so, the results suggest that computing based on light could help overcome the growing energy and speed limitations faced by the AI industry.
The two studies offer a glimpse of what future AI hardware might look like. Instead of electrons moving through silicon, intelligence could one day run on beams of light that race through compact optical processors.
