We can catch a glimpse of the emerging multitool quantum AI industry by considering a pioneering vendor in this regard. Xanadu’s PennyLane is an open-source development and training framework for AI, executing over hybrid quantum/classical platforms.
Launched in November 2018, PennyLane is a cross-platform Python library for quantum ML, automatic differentiation, and optimization of hybrid quantum-classical computing platforms. PennyLane enables rapid prototyping and optimization of quantum circuits using existing AI tools, including TensorFlow, PyTorch, and NumPy. It is device-independent, enabling the same quantum circuit model to be run on different software and hardware back ends, including Strawberry Fields, IBM Q, Google Cirq, Rigetti Forest SDK, Microsoft QDK, and ProjectQ.
Lack of a substantial and skilled developer ecosystem
As killer apps and open source frameworks mature, they are sure to catalyze a robust ecosystem of skilled quantum-AI developers who are doing innovative work driving this technology into everyday applications.
Increasingly, we’re seeing the growth of a developer ecosystem for quantum AI. Each of the major quantum AI cloud vendors (Google, Microsoft, Amazon Web Services, and IBM) is investing heavily in enlarging the developer community. Vendor initiatives in this regard include the following:
- Microsoft plans to integrate its QDK with development tools and Visual Studio so they can be used to build quantum programs for quantum hardware platforms from Honeywell, IonQ, QCI, and others, and also to simulate program performance on these and other platforms.
- AWS’ offering enables scientists, researchers, and developers to begin experimenting with computers from quantum hardware providers (including D-Wave, IonQ, and Rigetti) in a single place. It allows users to explore, evaluate, and experiment with quantum computing hardware to gain in-house experience as they plan for the future.
- IBM recently announced the expansion of Q Network, its three-year-old quantum developer ecosystem, under which more than 200,000 users are running hundreds of billions of executions on IBM’s quantum systems and simulators through IBM Quantum Experience. Participants in the network have access to Qiskit, to IBM’s quantum expertise and resources, and to cloud-based access to the IBM Quantum Computation Center. Many of the workloads being run include AI, as well as real-time simulations of quantum computing architectures.
In addition, enterprise quantum computing industry vendors such as D-Wave, Baidu, AmberFlux, CogniFrame, and Honeywell generally have consulting offerings geared at building the development ecosystem of partners and customers.
In the development of a tool- and platform-agnostic quantum AI developer ecosystem, Creative Destruction Lab is a key catalyst. Its Quantum Incubator Stream brings together entrepreneurs, investors, scientists in quantum technologies, and quantum hardware vendors to build ventures in the nascent domain of quantum computing, ML, optimization, sensing and other applications of quantum technologies. It provides quantum computing resources from D-Wave Systems (access to the latest D-Wave system and software libraries), IBM (access and hands-on technical support for the public IBM Q Experience systems and Qiskit tool), Rigetti (Rigetti Forest programming environment, with access to cloud-connected superconducting quantum processors and Quantum Virtual Machine), and Xanadu (Strawberry Fields, an open source library for photonic quantum computing, with a suite of simulators for execution on CPU/GPU, and access to Xanadu’s cloud-based quantum photonic chips).
The quantum AI market remains far from enterprise prime time deployment, but it has started to climb that maturity curve.
At the very least, the quantum AI industry will need to attain the milestones highlighted above to be considered fully mature: a consensus compelling app, a widely adopted open source development environment, and a broad development ecosystem. These maturity milestones have already been attained by leading AI tools that support modeling and training on purely classical computing architectures. We expect to see the market for hybrid quantum/classical AI mature to this point within the next three to five years.
The quantum AI market’s immaturity should not deter data scientists and other developers from exploring the technology today for proofs of concept, pilot projects, and even some production deployments. In this regard, we provide the following strategic recommendations.
To get ahead of the curve on quantum AI, application developers and data scientists should adopt solutions that leverage hybrid quantum/classical computing platforms. They should deploy quantum platforms as coprocessors not as outright replacements to handle specific AI workloads, such as autoencoders, GANs, and reinforcement learning agents. In addition, they should integrate investments in quantum-enhanced AI tools with legacy AI modeling and training platforms. They should also apply quantum AI tools to neuromorphic cognitive models, adaptive machine learning, training parallelization, and other advanced projects to identify workloads on which these solutions offer a clear advantage over classical computing platforms.
To position themselves for this growing opportunity, IT solution providers should expand their professional services offerings and partnerships in order to train the next-generation development ecosystem for quantum AI. They should integrate their quantum AI development environments with the widely adopted open source AI frameworks, most notably TensorFlow (especially the new TensorFlow Quantum) and PyTorch. Also, they should build more automated ML features into their quantum AI tools to simplify and accelerate the data preparation, model development, training, and deployment of quantum AI applications. They should align their quantum AI libraries, software, and services with leading data-science pipeline management, devops, and multicloud environments in order to pave the way for future production deployments of quantum-enhanced AI applications.
Market investors should place their bets with any providers of quantum-enhanced AI solutions who are building the tools for broad enterprise deployment of these capabilities over the next several years. Specifically, the priority should be on funding startups that follow Xanadu’s lead in providing framework-agnostic Python libraries for rapid prototyping of quantum AI applications to run on diverse software and hardware back ends.