Optoelectronic Integration at low temperatures with superconductors might be more efficient than at room temperature using semiconductors.
Artificial intelligence is a topic of great interest. Researchers are now trying to understand how the brain works so that they can create artificial systems that have intelligence comparable to humans.
This challenge has been tackled by many using traditional silicon microelectronics with light. Stimulating silicon chips with electronic or photonic circuit elements can be difficult due to the material used.
Researchers at the National Institute of Standards and Technology have proposed a large-scale approach to artificial intelligence in Applied Physics letters. This involves integrating photonic components with semiconducting electronics rather than semi-conducting.
Jeffrey Shainline, the author, stated that operating at low temperatures and using single-photon detectors, superconducting electronic devices, and silicon light sources will lead to rich computational functionality and scalable manufacturing.
Combining light with complex electronic circuits to compute could allow artificial cognitive systems with scale and functionality that are not possible with electronics or light alone.
Shainline stated, “What surprised me the most was that optoelectronic Integration may be easier when working with superconductors and low temperatures than when working with semiconductors and at room temperature.”
Semiconducting detectors can detect one photon, while superconducting sensors can see about 1,000. Silicon light sources can operate at 4 kelvins and be 1,000 times brighter than their room-temperature counterparts while still communicating effectively.
While some applications, like chips in cell phones, need to be operated at room temperature, the technology proposed would still have broad applicability for an advanced computing system.
Researchers plan to investigate more complex integrations with other superconducting electronic circuits and demonstrate all components of artificial cognitive systems, including synapses.
It is essential to show that hardware can be produced in a scalable fashion, so large systems are possible at a low cost. The superconducting optoelectronic Integration could be used to create scalable quantum technologies. These technologies would use photonic qubits or superconducting. These quantum-neural hybrid systems could also leverage the strengths of quantum connectivity with spiking neurons.
