IBM公布最新量子计算软件开发计划

IBM has unveiled an ambitious plan to speed widespread adoption of quantum computing by introducing simpler programming tools within the next five years.

The new software development plan, announced Thursday, is a complement to a quantum computing hardware “road map” the company announced in September.

IBM is among a number of companies racing to develop quantum computers, machines that harness the bizarre properties of quantum physics to perform calculations. This enables them to, at least in theory, perform calculations that even today’s largest conventional supercomputers can’t solve in a reasonable time period, as well as to significantly lessen the amount of time it takes to perform other hard computing tasks.

So far, most of today’s quantum computers are just on the cusp of being able to do things that are commercially useful. A number of companies, including IBM, Google, Honeywell, D-Wave Systems, and California-based startup Rigetti Computing offer access to quantum computers through cloud-computing networks. And, already, several large companies have been experimenting with using these machines to speed up processes, such as assessing the risk of financial portfolios or optimizing delivery routes.

A slew of other companies are developing algorithms based on the ideas of quantum computing, some of which can be used to improve performance even when run on traditional computers.

But programming quantum computers is not straightforward. In a traditional computer, all information is represented in a binary format, called a bit, as either a 0 or 1. In a quantum computer, information is processed using qubits, which can represent both a 0 and 1 at the same time. In a traditional computer, the status of each bit is supposed to be independent from every other bit. But in a quantum computer, the status of qubits is interdependent. In a traditional computer, you get the same answer every time you run a calculation. In a quantum computer, that isn’t the case, and instead the same calculation must be run hundreds or even thousands of times, and the probability distribution of possible answers analyzed, with the most probable answer usually being the best solution.

What’s more, right now, different quantum computers form these qubits in different ways—some use superconductors, some use photons fired from a laser, some use trapped ions—and even different quantum chips from the same company may have different properties depending on how many qubits each processing unit contains.

IBM hopes to eventually free programmers from having to worry about all this complexity. As a first step in its new software road map, the company Thursday released a toolkit for developers, which it calls Qiskit. It’s designed to make it easier for programmers to configure circuits on an IBM quantum computer accessed through the company’s cloud-computing network. It also makes it easier to run and store those programs.

IBM said this new system involves using traditional computer servers located alongside a quantum computer in the same data center to handle some of the calculation. The company said it hoped that executing the entire program in the cloud, with some workload handled on classical machines, could reduce the time it takes to run some computations from months to just hours.

Next year, IBM says it will offer ways that make it easier for programmers to create software that incorporates both quantum computing and traditional computing elements in the same program.

Then, starting in 2023, IBM says it will offer libraries of pre-built quantum circuits for its customers, as well as systems for controlling the very large quantum computers it has promised to make available. The company has said it will have a machine with 1,000 qubits by 2023, versus the 65-qubit computers that are its largest commercially available models today.

IBM says that in 2023 it will also begin offering pre-built quantum algorithms that programmers will be able to use via a simple cloud-based application programming interface, or API, a tiny snippet of code that allows a coder to call on the functions of another piece of software. This will enable developers to write software for a quantum computer in programming languages that they already know, without having to learn a new coding language, IBM said in a blog post announcing the new software road map.

The company said it wants software programmers working in corporations to be able “to explore quantum computing models on their own without having to think about the quantum physics.” By 2025, IBM said, it hoped to offer what it called “frictionless” quantum computing, where programmers no longer had to consider what kind of quantum computer they were using, or even what portion of a program was being executed on a quantum computer and what part was being run on a classical computer.