Using quantum computers to augment AI processing mostly in the training stages. Quantum AI runs on traditional GPUs but uses the quantum processor to process certain algorithms much faster.
AI training is a massive set of numbers in the model's neural network that are computed with each other over and over and adjusted in the backward passes using gradient descent, the standard algorithm for model training today (see
AI backpropagation). Because optimization is a major application of quantum computing, experts believe quantum computers used in conjunction with classical computers (regular CPUs and GPUs) will speed up training by orders of magnitude.
Pundits hypothesize that AI will even discover better algorithms, which would propel using both classical and quantum computers for AI training. See
Google Willow,
technology singularity,
AI and
quantum computing.