The race to quantum AI
Quantum computing promises a radical transition to computers that are infinitely faster, solving problems that would require unattainable amounts of computing power today.
Last week, New York-based company IBM announced that it's forming 'IBM Q,' a commercial quantum-computing service that will allow people to make use of quantum hardware via the Internet. IBM has been working on quantum computing nearly as long as it’s been an area of study, but the new business unit is a bold indication that it believes its research could be commercialized in the near future.
Whereas classical computers encode information as bits which represent either a 1 or a 0, qubits (quantum bits) can be in a strange state called 'superposition' of both at once. This, together with qubits’ ability to share a quantum state called entanglement, should enable the quantum computer to essentially perform many calculations at once, rather than in sequence like a traditional machine. And the number of such calculations should, in theory, double for each additional qubit, leading to an exponential speed-up.
As it stands, IBM’s largest quantum computer has only five qubits. By contrast the average laptop has hundreds of millions of bits in its processors, although the two types of computers are not directly comparable. IBM hopes to continue its research with the aim of building a universal quantum computer with a 50- to 100-qubit processor. That still doesn't seem very much, but you couldn’t build a large enough classical computer to simulate a quantum system of just that size.
Until now, the only commercially available quantum computers, such as the one that Google and NASA have been testing since 2013, are build by a Canadian company called D-Wave. In January it announced it's latest quantum machine, which boasts a whopping 2,000 qubits. However, the D-Wave systems use a different approach, known as quantum annealing, than IBM's more traditional way of quantum computing. D-wave’s qubits are much easier to build than their equivalent in the so-called 'universal' quantum computers, but their quantum states are also more fragile, and their manipulation less precise.
D-Wave has been able to develop their quantum machines a lot faster but they are suitable only for solving certain tasks, known as optimization problems. These are problems that involve finding the best solution from a large number of possible solutions. Take, for example, calculating the shortest route that a baker should take to deliver all orders to his customers. Because the quantum system sifts every possible answer at once, it could in theory resolve this type of issues in no time at all.
D-Wave’s computers don’t necessarily provide the most efficient answers to an optimization problem—or even a correct one. Instead, the idea is to provide solutions that are probably good, if not perfect solutions, and to do it very quickly. That narrows the D-Wave machines’ usefulness to optimization problems that need to be solved fast but don’t need to be perfect, which could include many artificial intelligence applications.
To help developers program D-Wave machines without needing a background in quantum physics, the company has designed a new software tool and released it as open source, meaning anyone will be able to freely share and modify it. The software, named 'Qbsolv,' joins a small but growing pool of tools aimed at easing the burden of writing code for D-Wave machines by freeing developers from having to worry about the underlying hardware. The goal is to kickstart a quantum computing software tools ecosystem and foster a community of developers working on quantum computing problems.
The field of quantum computing will soon achieve a historic milestone — quantum supremacy. This means that a quantum computer is actually capable of performing a task that is beyond even the most powerful 'classical' supercomputer. It explains why various institutions, including NASA, and companies like Google and Microsoft have already entred the race to develop the new, more efficient algorithms needed for quantum computing.
One of the most anticipated applications is 'quantum simulation.' Instead of spending years, and hundreds of millions of dollars, modelling a handful of materials and chemical reactions, researchers could study millions of candidates in virtual space. Whether the aim is stronger polymers for aeroplanes, more-efficient materials for solar cells, better pharmaceuticals or more-breathable fabrics, faster discovery pipelines could be of great benefit.
Another very promising application of quantum computing is the development of quantum neural networks. Artificial neural network models are widely used in machine learning for important tasks such as pattern recognition, translation and speech. The difficulty to train those classical networks, especially in big data applications, has motivated scientists to try to use the advantages of quantum information in order to design a quantum neural network.
The technological implementation of quantum computers is still in its infancy, and to date, quantum neural network models are mostly theoretical proposals that await their full implementation in physical experiments or research into artificial intelligence. But who knows, with the increasing power of quantum computing, machine learning and neural networks combined, an AI might one day create a quantum computer on its own.