IBM quantum computers' usefulness in sight -- using binoculars

IBM moved its quantum computing initiative forward late last month with new systems, including its first 53-qubit model, along with opening its IBM Quantum Computation Center in upstate New York.

Along with the new systems, IBM announced Quantum Volume, a new metric to better determine the overall power of a quantum computing system above and beyond its speed and performance. The metric will play a key role in accurately assessing progress toward achieving quantum advantage, or that point when quantum computers will surpass classical systems in practical use cases.

Bob Sutor, IBM's vice president of Q strategy and ecosystem with IBM Research, sat down to discuss the new systems, the Quantum Volume metric of IBM quantum computers and offered some insights into recent breakthroughs in quantum computing science.

What will the arrival of the 53-qubit system mean for quantum developers?

Bob Sutor: There are several ways to measure how powerful a quantum computer is. If you are trying to develop an application, the number of qubits matters. There are some problems that simply require that. But what you really need are enough good qubits and that is where our idea of Quantum Volume comes in because it measures all the characteristics of your system to ensure you have enough computing capacity at the right levels of accuracy.

Is there any significance to a 100-qubit system in terms of a breakthrough for meaningful application development?

Bob Sutor

Sutor: People haven't talked about 100 so much as they have 50 as a breakthrough. You can run a simulator on your phone that can run a 20-qubit system. But once you get up close to 50, the amount of memory you need to run a simulator gets into the petabytes [of memory]. Over 50 and there is no way you can pretend to be a quantum computer. At that point you really need a true quantum computer.

So how long before developers can produce meaningful applications for IBM quantum computers?

Sutor: It takes roughly 1,000 physical qubits to get one logical qubit. We now have 53 [physical qubits] so we will need 1,000 just to get 1 logical qubit. It is these logical qubits that enable application development in the long run. For some applications you might need 500 or 1,000 logical qubits and that is likely to come within the next decade. That is when we can achieve quantum advantage. Meanwhile, there are some things [developers] can do with the qubits we have. We plan to double the quantum volume number every year.

What are some of the things you can do better with an IBM quantum computer versus a classical system over the short term?

Sutor: An area like financial services involving things like risk analysis. If you are making in investment in stocks and bonds, or [thinking] about where you should build a new factory, there are many different factors and potential factors that go into that. Also, there are many probabilities, such as the quality of the labor, the cost of materials and many other interrelated factors that accurately calculate the risk. People tend to simplify these factors dramatically and trade off accuracy for the sake of computability. Quantum computing can contain this sort of exponential blowup.

Google recently claimed to achieve quantum supremacy in lab testing conducted in concert with NASA, although some of the testing methodologies were not fully disclosed. Do you see this as an important breakthrough?

Sutor: We always applaud good research wherever it happens, but in this case it's too hard to tell [if they achieved quantum supremacy]. Once all the results are published and can be peer reviewed and people can check their calculations and the assumptions built into it we will know better.

What is likely, in a few years, once we can reproduce their testing on the type of hardware they used, we can say whether they did it or not. But [quantum supremacy] is not our focus, we aren't trying to do lab-based experiments as benchmarks. We are focused on quantum advantage and putting out our 53-qubit system this month and making it available users of our Q Network. [Google's] benchmarks have value for extending the science, but that is different than getting people to use this stuff to solve real-world problems.

You are opening the IBM Quantum Computation Center this week. Do you think this will spur greater interest among IT shops in developing data center applications?

Sutor: Yes. But it doesn't mean enterprise developers can start doing quantum computing for their everyday business. We do have companies like Exxon, JP Morgan Chase, Daimler [AG] and the U.S. Air Force now part of the Q Network because they want access to the latest and greatest. So, this [Quantum Computation Center] is proof to them and others these machines are getting bigger and better with time.

We think we can also satisfy the needs of people doing active research, including other industrial researchers. For [IT people] on the sidelines thinking about jumping in, this can help them make decisions about when they should start forming a strategy.

With some principles of quantum mechanics and quantum physics closely associated with quantum computing, how important is it for IT professionals to have some understanding of the former?

Sutor: If you use [IBM's development toolkit] Qiskit, you won't have to learn quantum physics and mechanics. We can teach you how to code quantum computers without a physics background; everyday IT people won't have to deal with that.

Over the short term, what role does AI or machine learning play in quantum computing?

We can teach you how to code quantum computers without a physics background; everyday IT people won't have to deal with that.

Sutor: I think of it in terms of small, medium and large applications. For small applications, if you have an AI system that pulls in lots of data routines, but somewhere you have a calculation that is too big or slow or inaccurate for a classical system, you can replace that single computation with something that can be done on a quantum system. For medium-sized apps, if you have an AI process with lots of subcomponents, you can replace one or two of them with quantum alternatives that are more efficient. For larger applications, if you are working on a quantum system where multiple qubits are connected in complex, interesting ways, then you can see patterns in the data that would never pop out using a classical system, which can make data analysis more precise.

Where do you think the next major breakthrough will happen with quantum computing? Some suggest it will involve major improvements in error correction.

Sutor: My prediction is we aren't going to wake up one morning and go from zero error correction to 100% error correction. In the next few years we will start bootstrapping error correction into machines, but we need many more qubits to start looking at that solution. Eventually, we will get to the point of error correction fault tolerance. But even now, we are effectively decreasing the noise by implementing some error mitigation techniques and the apps that use them will follow along.