South African researchers have developed an artificial intelligence system that uses light itself to recognise information, a breakthrough that could contribute to faster and more efficient communications technologies.
The research was conducted by scientists at the University of the Witwatersrand (Wits University) in Johannesburg and demonstrates how specially structured light can be analysed without first capturing a detailed image.
The development, described by the university as a light-powered brain, combines principles from artificial intelligence and physics to identify information encoded in complex optical patterns.
The research was published in the journal Nature Communications in October 2026.
The team was led by master’s student Hadrian Bezuidenhout, under the supervision of Dr Isaac Nape and Professor Andrew Forbes, with contributions from Cade Peters and Ram Nandan Kumar.
The technology addresses a challenge in optical communications, where information can be carried through specially designed patterns of light. Although these patterns can encode complex information, reading them accurately becomes difficult when they change during transmission.
Conventional approaches often require detailed images of the light patterns, followed by extensive computational analysis.
The Wits system takes a different approach by directing light through specially designed optical layers towards detectors. The position where the light lands reveals information encoded in the pattern.
The researchers focus on a property known as topology, which describes characteristics that remain unchanged even when a pattern is distorted.
The principle is similar to a twisted ribbon: its shape can change when bent or folded, but the number of twists remains the same.
By identifying these persistent features, the system can recognise information without analysing every detail of the light pattern.
According to the university’s research announcement, the system was trained using only 15 examples and successfully recognised 81 different light structures within the family studied.
The findings suggest that optical systems could eventually process certain forms of information with less dependence on conventional image capture and digital computation.
Potential applications include optical communications, sensing, imaging and quantum communications.
For example, future communication systems could use structured light to carry information across optical links between buildings or satellites. The ability to identify encoded information despite distortion could be particularly useful in environments where atmospheric turbulence affects light transmission.
The researchers also demonstrated the recovery of an image encoded using a 15-letter topological alphabet.
However, the technology remains at the laboratory research stage. Further development and testing will be required before it can be deployed in practical communication networks.
Improving optical efficiency and integrating the system into reliable hardware will be important steps towards commercial applications.
The research also highlights the potential of collaboration between artificial intelligence specialists and physicists in developing new approaches to information processing.
For South Africa, advances in optical technology could strengthen local scientific research and create opportunities in telecommunications, advanced computing and quantum technology.
The findings demonstrate the country’s contribution to emerging technologies while highlighting the importance of sustained investment in research, scientific skills and innovation.
If further developed successfully, light-based neural networks could become part of a new generation of communication systems designed to process information more efficiently.
