Spanish
The project, called BRIGHT, replaces traditional transistors with tiny light-emitting diodes (LEDs) and seeks to transform the way AI systems process information in data centers. This initiative is led by the Technical University of Braunschweig, in collaboration with other academic and research institutions in Germany.
Unlike conventional computers that use silicon transistors, the developed system employs LEDs as computational elements within a neuromorphic architecture that mimics the parallel communication between neurons in the human brain. This approach allows for more efficient signal processing, with energy consumption far lower than that of traditional circuits.
Researchers have achieved a functional demonstration of the LED-based design and plan to optimize the system over the next five years, increasing the number of optical connections, improving the LED components, and perfecting the hybrid integration of different chip technologies. Innovation combines two branches of microelectronics that until now functioned separately: silicon-based CMOS circuits for logic and control, and gallium nitride devices known for their efficiency in light emission. The relevance of this project lies in the growing energy demand of data centers worldwide, driven by the training and operation of artificial intelligence systems. The neuromorphic approach with LEDs forms dense parallel networks capable of processing information with a minimal use of energy, which represents a promising and more sustainable alternative for the future development of AI. The team of scientists hopes that this breakthrough marks the beginning of a new generation of technologies capable of facing energy challenges in the digital age, with applications that could transform computing and intelligent systems over the next decade.
