Quantum Computing, Challenges & Opportunities

QCaaS, Investment, Key Industry Players & Startups, Potential Applications, Hardware, and Software

Andi Sama CIO, Sinergi Wahana Gemilang

TL;DR;
Starting with IBM in 2016, more and more big industry players and startups have been providing cloud-access to quantum computers (Quantum Computing as a Service), along with related quantum software development kits, whether developed in-house or through strategic partnerships. Among others are Alibaba, D-Wave, Rigetti, AWS, Microsoft, and Xanadu.
VCs have been selectively increasing their investment in quantum computing startups. At the same time, they can not be sure who will be the winner(s), especially on the foundation hardware technologies to build the basic quantum computation unit, the qubit.

ioneering by IBM, the world’s quantum enthusiasts can now have access to the future-computing platform today. IBM launched the IBM Quantum Experience in May 2016, accessible through IBM Cloud.

Providers of QCaaS, Quantum Computing as a Service, accessible on Cloud. IBM started as a pioneer in May 2016 with its IBM Quantum Experience launching, followed by several others until 2020 (source: Christopher Barnatt, 2020 and others).

The author quickly explored the IBM quantum experience platform for just a few weeks when it was still based on QASM in 2016, the Quantum Assembly language, leaving it hanging untouched for some years until revisited in early 2020 (Andi Sama, 2021b). This time, Qiskit, the higher-level quantum library, has improved so much as now it supports Python programming language.

Qiskit started to be available as open-source since its introduction in 2017 by supporting IBM quantum computer hardware (superconducting qubit). Also, qiskit has supported other quantum hardware technology like AQT (Alpine Quantum Technologies), based on trapped-ion qubits (IBM, 2019).

For exploration purposes, to some extent, using the IBM Quantum Experience on IBM Cloud is free of charge, from 1-qubit to 15-qubits (as of March 2021).

To be eligible to access higher capacity quantum computers with more qubits or better job priority in the queue when processing our submitted quantum circuits, we can arrange to have a special agreement through IBM Q Network at a certain annual cost. By 2020, IBM has a quantum computer with 65 qubits and intends to launch 1,121 qubits by 2023 and millions of qubits later on.

QCaaS, Quantum Computing as a Service

QCaaS is the term that we can use for Quantum Computing service that is accessible from the Cloud. We typically access the quantum computer from a classical computer, such as laptops or smartphones. Thus, the term Hybrid Classical-Quantum Computing.

When announced, the IBM quantum experience platform was still based on QASM, the Quantum Assembly language.

Qiskit, the higher-level quantum library, has been accessible through Python programming language as open-source since its introduction in 2017 by supporting IBM quantum computer hardware (superconducting qubit). Since then, Qiskit has supported other quantum hardware technology like AQT (Alpine Quantum Technologies), based on trapped-ion qubits (IBM, 2019).

For exploration purposes, to some extent, using the IBM Quantum Experience on IBM Cloud is free of charge, from 1-qubit to 15-qubits (as of early 2021).

To be eligible to access higher capacity quantum computers with more qubits or better job priority in the queue when processing our submitted quantum circuits, we can arrange to have a special agreement through IBM Q Network at a certain annual cost. By 2020, IBM has a quantum computer with 65 qubits, intending to launch 1,121 qubits by 2023 and millions of qubits later on.

Venture Capitals and Investment in Quantum Computing

During one of the Quantum for Business (Q2B, 2020) sessions in December 2020, “Investment Trends in Quantum Computing,” VCs’ discussion revealed that VCs have been investing in multiple types of quantum startups, from hardware to software, valued at about USD 500M. Besides, governments and big industry players have played a key role in developing quantum computing with investment commitments in the range of billions of US dollars.

The discussion also revealed that it is now too early to know that among competing quantum technologies that build the qubits, who will be the winner. Aggressive VCs may spread the investments in a select few promising technologies, hoping that a few will come as winners in a not-so-distant future. Others may wait and see or place their investment in quantum computing through syndication to share the risks. Others may invest in safer areas like the software framework that can utilize multiple future scalable quantum hardware technologies that do not exist yet.

Following the moderator’s question on “whether is it now the right time for the VCs to invest in quantum computing?”, in general, all the panelists agreed that “it is now the right time to invest.” Although the stakes are high, like binary betting with just two successful outcomes: yes or no, the successful ones’ rewarding potential is also high. It remains to be seen when a scalable fault-tolerant quantum computer can be available, whether it is based on superconducting or promising trapped-ion technologies or others. We have been starting to see a few bright lights for a potential successful quantum computing journey in years to come.

Quantum Applications

The commercialization of quantum use cases can be broadly categorized into 3 areas (IBM, 2020): Machine Learning, Simulation, and Optimization.

Machine Learning creates models to predict the outcome, given sufficient historical data. In the case of deep learning as part of the subfield of Machine Learning, the training process to create the model is adjusting the weights in neural network layers, given a loss function (i.e., minimizing the average sum of errors between actual data/ground truth and predicted results) and constraints in hyperparameters. Quantum Machine Learning (QML) is possible, e.g., by doing transfer learning in which part of the trained model is done classically, and part of it is then optimized with a quantum computer.

One example of a potential quantum application in Artificial Intelligence is through hybrid classical-quantum Machine Learning. It trains deep learning neural network layers by a hybrid classical-quantum means through transfer learning. Part of it — usually the last few trained (classically) neural network layers — is retrained by a quantum computer). An example use-case is the classic Image Classification, a binary classifier using a pre-trained ResNet neural network architecture for generating a neural network model to predict whether a person is wearing a mask or not (Andi Sama, 2021b).

Quantum Machine Learning. A generic neural network model trained on the ImageNet dataset is used as the base for Transfer Learning on the Image Classification task (based on ResNet18). This pre-trained model’s last layer is modified by quantum means through a quantum machine learning framework: Pennylane.ai. The framework provides convenient access to multiple quantum simulators and real quantum computer backend, including IBM real Quantum Computer on IBM Quantum Computing Experience (IBM Cloud) through an open-source Qiskit API (Application Programming Interface) accessible by Python programming language.

Simulation

Quantum chemistry simulation that performs quantum simulation on finding the right catalyst (that does not require high temperature) for chemical reactions can significantly improve new material discovery in material science (Cem Dilmegani, 2021).

Likewise, molecular biology and healthcare research (e.g., drug-discovery) include a process similar to chemistry research. In this case, quantum simulations can replace laboratory experiments.

Optimization aims to minimize or maximize a given objective function, given constraints. This can minimize travel distance in a logistic network or maximize capacity and system throughput to avoid conflicts and backlogs.

Recent potential applications of quantum computers in various fields are shown below (TheQuantumDaily, 2020).

The commercial application potentials of quantum computers.

Quantum Chemistry (e.g., for drug discovery) stays on the first rank in the TQD survey, followed by Security. A Quantum Internet, for example, relies mostly on Quantum Security.

As we are now in the era of Noisy Intermediate-Scale Quantum (NISQ), we do expect to see Fault-Tolerant-Quantum Computer (FTQC) closed to the year 2035 (Antonio Manzalini, 2020).

In addition to quantum computing applications, other use potential use cases like quantum sensing (e.g., for medical imaging and GPS-free navigation starting in 2025) and quantum communication (e.g., quantum internet towards 2035).

Gate-based Approach Quantum Technology

One of the two most adopted methods in developing quantum computer hardware is gate-based-approach. Another method is called the analog-based-approach, like the one implemented by D-Wave with its quantum annealer.

According to (Kristel Michiels et al., 2017), “A gate-based quantum computer is a device that takes input data and transforms this input data according to a unitary operation, specified as a sequence of gate operations and measurements (i.e., the algorithm) and conveniently represented by a quantum circuit.” An example of unitary operation is the Hadamard gate that transforms quantum state |1> to |+> and quantum state |0> to |->, a condition that puts a qubit in a superposition. Other available gates are X-gate, CNOT-gate, and Z-gate, to mention a few.

Available technologies to build qubits exist and being pursued by various industry players (Antonio Manzalini, 2020):

  • Superconducting qubits (superposition of current flowing in superconductors) — IBM, Rigetti, Google, Alibaba.
  • Spin qubits (qubits encoded in the spin of electrons) — Intel.
  • Topological qubits (quasi-particles like Majorana particles) — Microsoft.
  • Ion trap qubits (ion trapped in electric fields) — IonQ, Honeywell, IQT.
  • Neutral atom qubits (atoms trapped in magnetic or optical fields) — Cloud Quanta, Atom Computing.
  • Photonics qubits (qubits are encoded in states of photons) — Psi Quantum, Xanadu, ORCA.
An IBM 65-qubits Quantum Computer (ibmq_manhattan, online and operational) is based on superconducting qubit technology. Accessed through IBM Cloud as of March 14, 2021.
IBM Quantum Computers with superconducting qubit technology (IBM, 2021b).

Challenges Ahead

For most of us who are not physicists by formal education, it is tough to explore and learn about quantum computing. It is expected that it will eventually come to be implemented in practical applications to solve some of the world’s most challenging problems that are tough to solve within a reasonable time, even with the most sophisticated classical supercomputers.

Becoming exposed to and having an early hands-on practical experience in emerging technologies such as quantum computing or related new open-source frameworks would be invaluable in being relevant in the future.

In addition to understanding fundamental quantum computing concepts like superposition, entanglement, and interference, mastering a few leading developing software frameworks and software tools is also essential.

Quantum Software, Frameworks and Libraries

A software framework encapsulates many lower-level hardware details so we can focus on the higher-level problems to solve. Some startups are building software frameworks to work with any underlying hardware when they become available in the future. Zapata Computing for quantum workflow and Xanadu for quantum artificial intelligence, for instance.

The framework for Quantum Machine Learning or QML, like the one provided by Xanadu’s Pennylane.ai, combines Artificial Intelligence with Quantum Computer capability for developing a Hybrid Classical-Quantum Deep Learning model. See an example of doing image classification using quantum transfer learning (Andi Sama, 2020b).

Another framework by Zapata Computing called Orquestra provides an orchestrated integrated workflow to work in a hybrid classical-quantum environment.

Qiskit Open Source by IBM consists of rich quantum libraries for building a variety of quantum applications. Including Qiskit aqua that provides higher-level quantum libraries enables developers to focus more on building vertical applications, capitalizing on the underlying quantum hardware platform.

Algorithms and Tools

Generic tools to master include programming languages such as Python and C++. Programming languages are indispensable in working with quantum algorithms such as the two famous Grover’s and Shor’s algorithms. Grover’s quantum algorithm is used for searching unsorted data with quadratic speedup. Shor’s quantum algorithm is used for factoring large prime numbers with polynomial speedup.

More inventions of novel quantum algorithms with polynomial or exponential speedup would be something we may see in the coming years.

Quantum Hardware

On the hardware side, scalable quantum computers’ construction towards thousands and millions of qubits is still in active research either in Universities or Industries. We may expect to see substantial advancements towards 2030 and beyond. We are now in the “Early Industrial Era for Quantum Computing,” according to Prof. John Preskill, an American theoretical physicist and the Richard P. Feynman Professor of Theoretical Physics at the California Institute of Technology.

In NISQ-era — IBM, with its 1,000+ logical qubits’ plan, expects that 2023 could be the inflection point for having practical applications of quantum computing. Google, Xanadu, IonQ, and a few others have also been racing towards FTQC. The big challenge now is that a logical qubit needs to be built from many physical qubits to be fault-tolerant. As many as 1,000 physical qubits are needed to make one logical qubit.

In the future, there may be a time that a breakthrough in material science will enable the creation of a high-fidelity logical qubit without much effort to build it from hundreds of physical qubits, compensating for quantum error correction. When that time comes, the business impact of creating millions of logical qubits may accelerate many developments of commercial applications.

IBM, Google, Alibaba, and Intel have been experimenting with Superconducting qubits, aiming for universal quantum computers. Likewise, Xanadu with Photonic qubits. Intel, in fact, is looking for another promising technology: Spin qubits.

D-Wave has been persistently focusing on quantum annealer, targeting specific areas around optimization and simulation problems.

By relying on topological quantum qubits, Microsoft has not shown any significant results so far in the last 15 years. Recent news (Matt Swayne, 2021) could be a huge setback for Microsoft (or an opportunity for a new challenge?) as the underlying quantum hardware technology that it depends on may not be visible for practical implementation.

References