Most of the time, when we talk about functional quantum computers in the world, we’re talking about D-Wave. That company has taken the lead on quantum computing and provides its own 2048-bit quantum annealers for solving certain types of problems. IBM, however, has its own five-qubit system it’s been working on. Last year, IBM announced it would make quantum computing free in the cloud to those who were interested in test-driving software. Today, the company announced the availability of a new SDK and API meant to allow for easier quantum programming without needing, as IBM puts it, “a deep background in quantum physics.”
This is an important step forward compared with where IBM was just under a year ago. Back in May 2016, IBM announced it would make its five-qubit universal quantum computer available to developers who applied for permission to use the system. An API and software development kit should make developing new quantum computing software easier. IBM is also offering a new quantum simulator that allows users to simulate a model with up to 20 qubits.
IBM has thrown shade at D-Wave on multiple occasions, stating “The consensus of the scientific community is a quantum annealer has no known advantages over conventional computing.” But how can IBM possibly hope to compete against D-Wave, when D-Wave can flex 2048 qubits of processing power and IBM’s system has only five? It has to do with each system’s fundamental approach to quantum computing.
If you treat “quantum computing” as the act of taking advantage of quantum state phenomena to perform calculations, there are, broadly, four top-level approaches to quantum computing so far: quantum gates, one-way quantum computers, adiabatic quantum computers (also called quantum annealing), and topological quantum computers. They all take advantage of quantum superpositions — the state of being neither 0 nor 1, but both — to do their calculations. D-Wave’s quantum computer uses quantum annealing, which is powerful in optimization problems like the traveling salesman problem, when you’re trying to calculate a maximum or minimum. It uses magnetic fields to nudge a qubit from its superposition into the state that best solves the problem.
IBM’s “IBM Q” is a universal quantum computer, which is a different beast. A universal quantum computer is capable of doing more kinds of calculations than the relatively domain-limited quantum annealing machine. There’s an analogy to a RAID 5 array in how the IBM system works: of its five qubits, four do calculations and one is reserved for error checking the others. The error checking bit is called a “syndrome bit.” IBM expects their Quantum Experience will soon be solving problems in drug discovery and computational chemistry, materials science, AI, and even logistics and supply chains.
IBM is gearing up to sell machines with 50-100 qubits in the next few years, which would be much more capable of solving practical problems; this API release aims to get developers working on software now in order to be ready for when the next wave hits.