To see a glimpse of how quantum computers will be programmed in the future, do the following:
- Open ChatGPT on your computer
- Ask ChatGPT: Write a Qiskit program that learns a 20-bit string encoded in an oracle
- Repeat, substituting your preferred Quantum SDK language (CIRQ, TKET, Q#, PennyLane, etc.) for Qiskit.
A big prediction for the classical world is that most of the classical computer code written in a few years will be created by AI. This was recently expressed by Microsoft CTO Kevin Scott, Facebook CEO Mark Zuckerberg, Anthropic CEO Dario Amodei, and many others.
If you think this will be important for coding classical computing programs, it will be even more critical for quantum computing programs. It was once said that if anyone created a million qubit quantum computer and put it online next week, no one would be able to immediately write a program that takes full advantage of it. With AI created quantum programs, that will no longer be true. There are far fewer people today who can write a quantum program versus classical programs and the industry has been worried about whether there would be enough quantum programmers trained fast enough for quantum to reach its full potential. AI-based coding assistants will help to lessen that problem.
So, in the future, the language that may be most used for creating a quantum computer program won’t be Qiskit, CIRQ, Q#, Pennylane, TKET, or others, it might be English. A subject matter expert will describe the problem with a text description and perhaps an Excel spreadsheet containing the input data and the Quantum AI SDK will do the rest. They won’t need to program anything at the Hadamard, Cnot, or other gate level. The user won’t even need to decide which algorithm to employ. The Quantum AI SDK will do it for them.
Like most AI models, a quantum computing programming AI will need to be trained on all the quantum algorithms that are available. A good start might be the Quantum Algorithm Zoo. But the providers of these AI quantum code generator tools will likely supplement this with examples they have collected from their own internal development efforts, contributions from users and other algorithm libraries. In addition, besides requiring the Quantum AI code generator to understand the user’s application and selecting the best algorithm from the ones it has been trained on, there will need to be a backend phase to use AI to optimize the generated circuits so that can run most efficiently on the target quantum computer.
So, will there be any need for quantum software developers? Certainly yes! But the job’s responsibilities will change. Rather than writing code that runs a user’s problem using existing algorithms, there will still be a need for creating completely new algorithms and improving the efficiencies of existing algorithms through new approaches. An Quantum AI code generator won’t be able to do this on its own. A human quantum software developer will be needed to discover the new algorithms and then add them to the quantum algorithm libraries for future use.
There are a number of companies that are working on technology to do this. Here are some examples.
- Microsoft is integrating Copilot into its Azure Quantum platform to provide code assistance quantum programming. Copilot in Azure Quantum acts as a generative AI assistant, helping users learn and explore quantum computing concepts. It can generate and explain Q# code, which is Microsoft’s programming language specifically designed for quantum algorithms.
- IBM is developing a Qiskit Code Assistant service with a Visual Studio code extension that will help customer generate code for their Qiskit programs. This tool is a generative AI assistant that has been trained on millions of text tokens of Qiskit code examples created over the years. It uses IBM’s Granite Code large language model (LLM) and IBM’s watsonx AI platform to help automate the code process so that users can generate high-quality Qiskit code with more productivity, at a more abstract interface level, and obtain more efficient code.
- AWS has a code assistant offering called Claude-3. This software can be used by both beginners and advanced users building quantum workloads. With the code assistant, customers can ask quantum computing questions and each question provides a working code snippet with detailed comments in a matter of seconds.
- Quantum software startup Strangeworks is also actively working on this. You can see a demonstration of their vision on how this might work here. (the relevant part starts at time 9:50) in the video.
So when can we start seeing this? The example we showed with ChatGPT at the beginning of the article is still quite primitive and the time estimates vary. But our belief is that this approach will become very common for end users creating quantum programs for use in their production applications within the next few years.
Of course, code generation is not the only place where AI and Quantum will work together. GQI will be issuing shortly a 75 page report titled Quantum Meets AI that covers:
- How AI is used for quantum
- How quantum is used for AI
- How Quantum + AI will work in a combined fashion to process hybrid algorithms
To give you a preview, GQI’s framework below shows how a Quantum AI SDK would fit into the overall AI for Quantum stack. The AI code assistant is shown below in the dark green square, but you can see there are uses up and down the stack where AI can be applied in a quantum system.
For those interested in finding out more about AI and Quantum technologies will be working together, contact GQI at [email protected] to find out how you can obtain an early copy of this report. Also available for GQI clients is a GQI AI Analyst that has been trained on a decade of QCR articles and GQI reports to answer many of the quantums question people may have. Ask for a demo of that too.
April 9, 2025
Leave A Comment