The Rise of Quantum Computing: How It Will Revolutionize AI and Data Processing
One hundred years ago, physicists started finding out about strange properties of matter at the atomic level – for instance, superposition and entanglement – and this gave birth to quantum computing.
This includes the development of new materials and drugs, optimising factory floors or global supply chains, spotting fraud or risk patterns in financial transactions and so on.
Acceleration
This is measured in terms of bits, which are either on or off – a natural way to represent computers. In quantum computing, ‘quantum bits’, or qubits, can be in more than one state at a time, which increases the machine’s computational potential exponentially. And then there’s another core quantum property, entanglement. Entanglement binds qubits so that their states influence each other, independently of the distance between them. Firms typically look to quantum computers to deliver orders of magnitude speedups over classical ones, which could mean that machine-learning training times for models shrink, even as more experimentation over longer periods becomes possible. Quantum computing might solve many of the problems that artificially constrain AI today; for instance, making it possible for pharmaceutical companies to simulate chemical reactions among moving molecules faster than could ever be achieved by today’s best supercomputers, and thus accelerating the discovery of new drugs. One of the things that quantum computing does really well is produce random numbers – something currently achieved in classical computers through the use of prosthetic quantum phenomena such as radioactive decay. By fractionally preparing quantum versions of its system states, a quantum computer could therefore be used to enhance optimisation procedures at large scales, such as global shipping and airline timetabling – with savings to be reaped in terms of both cost and environmental sustainability.
Big Data
Large datasets lend themselves to quantum computers because a calculus of uncertainty allows a quantum computer to search an immensely large permutation space of possible input combinations – instead of checking one element at a time – for a specific point. An independent review in Nature (2020) reveals the usefulness of QC testing for business owners wanting to see where inefficiencies could be found during production or where there may be a market opportunity, as well as helping companies with supply chain optimisation, logistics management and improved scheduling of staff. Pharmaceutical companies are harnessing the power of QC to discover drugs faster than ever before. Today’s supercomputers are programmed to run a simulation in which participants interact according to specific rules. But this approach quickly becomes illogical, Williamson says, when trying to model a single molecule, on a supercomputer, to understand its molecular interaction: ‘You go down to the minimum distance, and there’s no two points of a molecule that are exactly the same.’ QC kicks in at this point and, instead of waiting for the computer to process two points and then three – the way a classical computer works – ‘both points are moving down at the same time’. If one obtains a negative result, its counterpart moves on. On today’s computers, interaction between molecules can take weeks for the equations to play out. But on a quantum computer, it might only take a matter of hours to explore all possible ways that it interacts with other molecules. QC can help a company of any sort produce more efficiently and deliver more effectively as it scales – and it can also help minimise wasteful environmental impacts. Last month, the German car manufacturer Daimler announced that it is using QC to accelerate developments in electric vehicle batteries and improve the chemistry of their cells.
Machine Learning
With AI, a quantum computer’s ability to run larger data sets at speed will accelerate feature extraction and pattern matching too, raising the quality of inputs to machine learning algorithms, and making them more effective. One example of this is the ‘Traveling Salesman Problem’, in which the optimal route between various cities must be established. A classical supercomputer could potentially solve around 50 cities in 10 20 years, take a breath, and then move on to trying to solve just 500 cities. An efficient quantum computer, by contrast, could solve 500 cities in under a second. Daimler is looking at how quantum computing (QC) might shorten the time needed to manufacture batteries, optimise logistical operations of fleets and coordinate congestion on traffic-heavy routes suited to autonomous vehicles. Optimisation problems such as these are often too complicated for classical computing, which could benefit from those provided by a QC-powered quantum neural network. Autonomous systems, improved natural language processing, computer vision, enhanced feature extraction and time series analysis that might benefit from access to a quantum neural network stand to see transformational leaps forward.
Optimization
Quantum computers have the potential to be central to solving critical computational problems such as DNA research, drug discovery, financial modelling and encryption, as well as numerous others. When error correction and fault tolerance are fully realised, hybrid quantum-classical data centres will be powerful engines for tackling the large issues such as these. A further core application of QC lies in the realm of optimisation, also important for AI and machine learning applications aimed at accelerating various business efficiency improvements. For example, imagine a group of medical boards meeting in order to decide upon optimal treatments for rare diseases, but this process may be too slow and there are possible pitfalls that could make it unlikely to succeed. Quantum computers can theoretically calculate all possible solutions to a problem almost instantaneously, enabling faster decision-making in fields such as finance and logistics. For example, a quantum algorithm might be able to read and simulate several financial situations in a shorter time than its classical counterparts – helping stockbrokers make smarter investment decisions while also improving current models for assessing risk, thereby creating a better return and reducing logistical costs and delivery times.