Scientists at the University of the Witwatersrand (Wits) have developed a system they call a "brain on light." It recognises hidden information in complex light patterns and could accelerate future communication systems.
The study was published in Nature Communications. The work was led by master's student Hadrian Bezuidenhout, with contributions from Kade Peters and Rama Nandan Kumar. The project was supervised by Dr Isaac Nape and Professor Andrew Forbes of the Wits Structured Light Laboratory.
The system uses principles of artificial intelligence and physics. Light carries data in special forms resembling letters of the alphabet. The problem is reading these patterns quickly, especially when they are distorted in transmission.
Instead of capturing and processing thousands of points in detail, the light is passed through special layers that direct it to detectors. Where the light lands reveals the required information. The scientists compare this to counting the twists in a coiled ribbon, even if it is bent in different ways: what matters is the topology, not every fold.
After training on just 15 examples, the system correctly identified 81 different light structures, including ones it had not encountered before. It continued to work even when noise was added to the detectors.
"Creating light patterns for communications is only useful when we can read them as quickly as we create them," Nape notes. How exactly does the light-based system cope with distortions and noise without complex computations?
Professor Benjamin Rosman, director of the MIND Institute, emphasises the value of an interdisciplinary approach: combining ideas from AI and fundamental physics opens up new ways of processing information. Professor Forbes adds that the work helps solve a key challenge for future quantum and classical networks — the practical readout of topological structures of light.
Nape sees prospects beyond communications as well: in quantum technologies, where extracting data from individual photons is especially difficult. The study shows how optical intelligence can help in these areas.
