As artificial intelligence has drawn attention, researchers are focused on understanding how the brain completes cognition so they can create artificial systems with general intelligence similar to humans’ intelligence.
Many have approached this challenge by using a typical combination of silicon microelectronics and light. However, the fabrication of silicon chips with photonic and electronic circuit components is complicated for various practical and physical reasons related to the elements used for the components.
In Applied Physics Letters, by AIP Publishing, scientists at the National Institute of Standards and Technology suggest a proposal to large-scale artificial intelligence that concentrates on combining photonic elements with superconducting electronics rather than semiconducting electronics.
“We confirm that by operating at low temperature and employing silicon light sources, superconducting electronic circuits, and single-photon detectors, we will discover a way toward rich scalable computational functionality and fabrication,” said author Jeffrey Shainline.
Using light for communication with multiple electronic circuits for computation could let artificial cognitive systems of scale and functionality beyond what can be accomplished with either electronics or light alone.
“What amazed me most was that optoelectronic integration may be much simpler when operating at low temperatures and using superconductors than when operating at room temperatures and using semiconductors,” said Shainline.
A single photon can be exposed by the Superconducting photon, while 1,000 photons are required by semiconducting photon detectors. So not only silicon light sources are 1,000 times less bright than their room temperature but they also work at 4 kelvins and still communicate efficiently.
Some applications, such as PCBs in smartphones, require working at room temperature, but the advanced technology would still have wide-reaching applicability for advanced computing systems.
The scientists plan to examine more complex integration with other superconducting electronic circuits as well as show all the elements that consist of artificial cognitive systems, including neurons and synapses.
Showing that the device can be produced in a scalable way, so large systems can be available at a reasonable cost, will also be necessary. Superconducting optoelectronic integration could also assist to design scalable quantum technologies based on photonic qubits or superconducting. Such quantum-neural hybrid systems may also guide to new methods of leveraging the strengths of quantum entanglement with spiking neurons.
“Optoelectronic intelligence” by Jeffrey M. Shainline, 20 April 2021, Applied Physics Letters.