Over the past year, there has been increasing focus on how quantum computers fit into and link to classic computing architectures. Quantum computers could act as an accelerator to perform complex calculations for certain tasks that are beyond the capabilities of even classical supercomputers. The classical computers or servers are used for preprocessing in the development of quantum algorithms and circuits and for postprocessing to manage the errors, improve the results, and complete the processing task. As is evident from the growing number of AI use cases, AI can enhance classical computing capabilities. So, it stands to reason that AI could also enhance quantum computing capabilities and several companies are working towards achieving this goal.
Even though many people and companies are starting to combine quantum and AI into a single term, the two are very distinct technologies. AI is the training and use of neural network models developed and run on classical computing platforms powered by CPUs, GPUs, NPUs, DSPs, FPGAs, and other traditional binary-processing logic elements. Quantum computers use alternative compute architectures, such as superconducting transmon qubits, to solve very complex problems using quantum physics. While the two require different hardware, software, and support systems, the integration of the two is moving forward, especially for the benefit of quantum computing. IBM is one of the companies paving the way for AI to complement quantum computing development.
IBM is considered the leader in the quantum computing segment with continued advancements in hardware, software, and systems technologies, and with development quantum computers already deployed around the world. IBM is also a leader in AI technology through its watsonx platform, which has logged many advances beginning with its Jeopardy game show win in 2011. Since then, watsonx has evolved to a scalable enterprise platform with the AI studio, data, governance, and assistant solutions. Now IBM is bringing the two technologies together to enhance quantum computing and accelerate its adoption.
In a recent discussion with IBM, the company outlined how it is integrating its AI technology into the Qiskit software to improve the ease of use of the SDK tools and OpenQASM3 (open quantum assembly language). IBM is using its watsonx generative AI platform, leveraging the company’s Granite AI model, to generate digital agents capable of providing developer support and quantum code assistance.
In addition, IBM is researching and developing new AI models to improve other critical aspects such as circuit optimization, resource management, and improved error suppression, mitigation, and correction.
As part of its commitment to integrating AI into quantum computing, IBM is also introducing the Qiskit Code Assistant service with a Visual Studio Extension and plans to offer two quantum chatbots – one to assist developers and the other to general users of IBM Quantum services.
In terms of circuit optimization, AI models can be embedded as plugins to the Qiskit SDK through a transpiler service or be combined with heuristic methods. According to IBM, the transpiler service provides better mapping of abstract circuits to quantum ISA circuits resulting in up to a 40% improvement in circuit size, better quality, and a 2x to 5x improvement in processing speed.
For resource management, IBM is developing AI solutions to better estimate the quantum runtime, flag workloads that are likely to fail, and partition circuits for parallel processing to better utilize both the classical and quantum resources. This includes leveraging AI supercomputers.
Combined with IBM’s aggressive roadmap to reach 100 million gates by the end of the decade and 1 billion gates around 2033, quantum computing is rapidly moving toward the deployment of practical quantum applications over the next few years. As a result, we may begin to see heterogeneous data centers that combine the performance of the latest CPUs, AI accelerators, and QPUs (quantum processing units) by the end of the decade.
