Because artificial intelligence has aroused widespread interest, researchers are focusing on understanding how the brain achieves cognition so that they can build artificial systems with intelligence comparable to human intelligence.
Many have responded to this challenge by using conventional silicon microelectronics in conjunction with light. However, the fabrication of silicon chips with electronic and photonic circuit elements is difficult for many physical and practical reasons related to the materials used in the components.
In Applied physics letters, Researchers at AIP Publishing, the National Institute of Standards and Technology, propose an approach to large-scale artificial intelligence that focuses on integrating photon components with superconducting electronics instead of semiconductor electronics.
“We argue that by operating at low temperatures and using superconducting electronic circuits, single-photon detectors and silicon light sources, we are opening the way to rich computational functionality and scalable manufacturing,” said author Jeffrey Shainline.
The use of light for data transmission, together with complex electronic circuits in computing, could enable artificial cognitive scale and functional systems beyond what can be achieved with either light or electronics alone.
“What surprised me the most was that optoelectronic integration can be much easier when working at low temperatures and using superconductors than when working at room temperature and using semiconductors,” Shainline said.
Superconducting photon detectors allow the detection of a single photon, while semiconductor photon detectors require about 1000 photons. So silicon light sources not only work with 4 kelvins, but they can also be 1000 times less bright than room temperatures and still communicate effectively.
Some applications, such as mobile phone chips, require working at room temperature, but the proposed technology would still have wide applicability to advanced computing systems.
The researchers plan to study more complex integration with other superconducting electronic circuits and to introduce all components that contain artificial cognitive systems, including synapses and neurons.
It is also important to show that the hardware can be manufactured in a scalable way, so large systems can be implemented at a reasonable cost. Superconducting optoelectronic integration could also help create scalable quantum technologies based on superconducting or photonic crystals. Such quantum-nerve hybrid systems may also lead to new ways to exploit the strengths of quantum binding with spike nerve cells.