Google Quantum AI Labs fellows Masoud Mohseni and Hartmut Neven have filed a patent for producing chips including both quantum and classical processors using the same process.
A brief primer on quantum computing can be found at ExtremeTech.com.
Why It Matters
The authors speculate that such a chip would be valuable for enabling Artificial Intelligence problem-solving. AI tasks can typically be translated into machine learning optimisation problems. Some tasks can be carried out by classical digital processors; however, many optimisation problems can be performed better using quantum hardware. by including both quantum and classical processors on one chip, Mohseni and Neven postulate substantial improvements in computational speed for certain types of problems.
According to the patent, on such a chip the output of a classical algorithm run by the classical processor can be used to seed a quantum algorithm run by the quantum processor. Then the output would be passed back to a classical processor for post-processing and potentially back and forth between classical and quantum processors until a solution is reached.
The patent’s quantum computing element depends on entanglement between electrons, a situation which typically lasts for small fractions of seconds until they un-entangle or ‘decohere’. Supercooling can increase the time the electrons remain entangled as this 2017 experiment demonstrated. So, the patent proposes a cooling unit around both processor types. By also supercooling the classical processor, it enables the semiconductors involved to function considerably more energy-efficiently. The patent appears to claim that, even after the energy for a supercooling refrigeration unit is accounted for, the processor would use significantly less power than operating at room temperature. 6GWorld has reached out to Google to clarify that this is the case, but no response has been forthcoming at this time.
Placing classical and quantum processors on the same chip minimises the distance between the processors, reducing processing time. According to Mohseni and Neven, the occurrence of decoherence in the quantum processor may also be reduced: “This passive error-avoidance architecture is significant as there is not yet any known active error-correction strategy for such systems.”
