Quantum for AI, AI for Quantum

Nvidia have just released a paper on AI for quantum computing. Hot off their heals, Google have released AlphaQubit, an AI-based decoder for quantum error correction (QEC), using neural networks to tackle one of the most challenging problems in scalable quantum computing.

Fault-tolerant quantum computing (FTQC) is a quantum computing approach that ensures reliable operation despite the presence of errors in qubits and quantum gate operations.

We’ll be following with keen interest the technical developments in the areas highlighted by the Nvidia paper:

AI for quantum pre-processing

Quantum pre-processing involves preparing quantum algorithms for execution on quantum devices. AI plays a crucial role in several pre-processing tasks, such as unitary synthesis, circuit reduction, and state preparation.

AI for the development and design of quantum hardware

Characterising quantum devices is vital for making informed control, tuning, and optimisation decisions. Machine learning (ML) methods have been instrumental in learning device characteristics that are otherwise challenging to access experimentally.

AI for quantum device control and optimisation

AI techniques, particularly ML algorithms, have been applied to tune, characterise, and optimise quantum devices. For instance, classifiers and neural networks (NNs) have been utilised in semiconductor quantum dot devices identify charge transitions in large parameter spaces and to detect Pauli Spin blockade.

AI for quantum error correction

QEC protocols make joint measurements on syndrome qubits to infer which physical data qubits have most likely experienced errors. FTQC requires scalable and adaptable QEC, which is hard to realise in practice when using a classical decoding algorithm. AI-based decoders, particularly convolutional neural networks (CNNs), have been employed to decode high-dimensional QEC codes.

AI for quantum post-processing

Post-processing entails extracting meaningful insights from quantum measurements and optimising the measurement process. AI techniques can reduce the data points required to estimate observables, enabling state reconstruction with fewer samples. NNs are used to classify multi-qubit states while mitigating qubit readout crosstalk, and in quantum state tomography for state reconstruction.

Broader Applications of AI in quantum computing

Beyond the aforementioned applications, AI copilots can provide user-friendly hybrid programming workflows to facilitate the integration of quantum processors within AI supercomputers. Other promising areas include the use of AI to generate novel quantum algorithms, and the development of quantum-specific foundational models.

 

For further information, please contact Tom Woodhouse.