Nvidia and British quantum computing company Orca have teamed up to find a way to speed up and improve the machine learning process. The two companies have coreleased that making use of both quantum processors and standard GPUs can improve the output quality and decrease the training time. This is seen as a boon for machine learning models which require a great deal of computing power to perform complex tasks.
The quantum processor from Orca Computing helps the GPUs in the machine learning process in much the same way the GPU would do it itself. Splitting up the task between the two allows for a significant performance increase because the processor is able to deliver the best placement of each individual pixel being generated by the GPU with much needed speed and accuracy. This hybrid approach is also being used in generative modelling by Orca, which is able to apply its quantum machine learning to generate high quality images from text prompts at a much faster rate than before.
William Clements, head of machine learning at Orca Computing, speaks of the hybrid system as a way to not only improve the machine learning process, but also a way to scale the process up to larger systems. The team has already seen the results of combining a photonic QPU with eight GPUs, and they’re preforming beyond expectations.
Nvidia has been a pioneer in developing what enables this hybrid application. Its open source CUDA Quantum toolkit, and its quantum simulation tools are proving essential to connect the classical world with the quantum world. For now, these applications are the closest we can get to pure quantum computing — which will be the way of the future — until then, companies like Orca are taking advantage of the hybrid solution to achieve quantum advantage today.
