April 15, 2024

Where AI and quantum computing meet

For many IT leaders, quantum computers seem closer to science fiction than something that can be implemented in their data centers. But it’s on its way; Last month, IBM introduced System Two, the first quantum computer that connects three processors to work together.

Last year’s small steps on the quantum roadmap are turning into this year’s biggest advances. IBM tasked Scott Crowder, vice president of quantum adoption, with helping customers discover new uses for quantum computing, as well as developing the software to accomplish those tasks. We asked Crowder to provide CIOs with a progress report on where quantum computing technology has advanced and what it will take to bring it to the enterprise.

For those who have heard of quantum computing but don’t fully grasp it, how is it different from classical computing that powers our laptops, phones and desktop computers?

Scott Crowder: It basically uses a different data science. It’s not classical in the sense that we were doing classical information science before we invented digital computers. This is different in the way it performs the calculation. Therefore, it is better at certain types of mathematics than today’s computers and vice versa.

In theory, you could run anything on a universal quantum computer, but you wouldn’t want to. You just want to quantum analyze things that classical computers are not good at and that quantum computers have been proven to be good at. They take advantage of quantum mechanics, so it’s like science fiction come to life. It can perform certain types of calculations that we may never be able to perform in a classroom.

IBM Quantum System Two debuted last month. It is the company’s first modular quantum computer that connects multiple quantum processors.

When will we see more widespread adoption of quantum computers and what will that look like?

Crowder: Before this year, you could argue that anything you could do with a quantum computer could be simulated classically. There was no point in doing the calculation on a computer other than learning about its data science. But that has changed. This year, for the first time, something could be run on a quantum computer that cannot be run on a classical simulator. It doesn’t mean you can run anything on a quantum computer. They are the first two types of calculations where you can actually get value from a quantum computer rather than trying to simulate it.

Over the next few years, utility or usefulness will continue to expand. At this time, there are limitations to the magnitude of the problem we can address due to the quality of the systems. But we’re past the point where there’s value in running a quantum computer. It doesn’t mean there is still business value, because problems tend to get bigger, they need to be integrated into your workflows, etc. But we don’t think it will take until 2033 for other people to realize commercial value.

In the 1940s, we didn’t carry classic computers in our pockets or do what we do on our phones. They were the initial use cases in programming. I think that’s going to be true in this decade. [for quantum computers]. In the next decade, as systems get bigger and better, and better, we will see more and more use cases.

What will be the first use cases?

There are three types of mathematics that quantum computers are getting better at.

One of them is to simulate nature. Materials, properties, physics, chemistry – think about all things industrial chemistry, as well as health and life sciences.

The second type of math that quantum computers will be better at is certain types of complex structure in the data. The most famous algorithm, Shor’s algorithm, which all nation-states are interested in, is that kind of mathematics. It does factorization: A times B equals C. A times B; Regular old computers are good at giving you C. But, given C, your computer isn’t good at figuring out what A and B were. Classical computers aren’t good at that kind of math, which is a good thing. If we don’t have cryptography, we don’t have digital economy.

This is part of the discussion about quantum. If it falls into the hands of bad actors, we will be in serious trouble. But this type of math is also used in machine learning, in things like classification. It can help find fraud, better testing sites for clinical trials, and better treatments when data from a patient’s medical history is provided. There is a lot of interest in the industry to take advantage of quantum computers in the short term for these types of problems.

The last type of mathematics, which is also interesting (but for the second phase of the journey at the end of this decade or in the next), has to do with optimization. What takes me N tries on a regular old computer, will take the square root of N tries on a quantum computer. So N is equal to 100, the square is equal to a factor of 10. There could be advancements in that space as well. Some examples could be portfolio optimization in financial services, risk management and logistics, a lot of things that today people struggle with regular computers to document.

Quantum computers operate at a temperature close to zero Kelvin. How are we going to solve the freezer problem? Put them in space?

Crowder: Unfortunately, space isn’t cold enough.

We need to isolate the computing part from the rest of the universe because you are programmatically entangling these qubits with each other in a specific way. It can’t be perfectly isolated (at absolute zero) because if it’s perfectly isolated, we can’t get them to do anything. It needs to be connected enough to the rest of the universe that you can program it, but the rest of the universe can’t alter it. That’s why you either need to keep it very, very, very, very cold, like we do with our technology, or you need to shoot lasers at it. [using a light-based approach] to eliminate entropy. It’s messy and it’s not at room temperature no matter how you do it.

The good news is that there are commercial refrigeration techniques that are stable. They are low cost and consume little power compared to regular old computers, as well as compared to a shelf of electronic devices. These things seem extremely efficient. The cooling action is not a big problem. There are other problems with scaling them up and reducing the cost, but the underlying technology is there.

Do you think quantum computers will ever find their way into the average enterprise data center? Or will it be reserved for specialized use that only large companies will be able to afford?

Crowder: The infrastructure around quantum computers, I know, seems strange and different right now. But we have deployed them at the Cleveland Clinic; we have implemented them in Germany, Japan and Canada. We have large data centers. I think in the short term, like the next few years, technology is advancing so quickly that it probably doesn’t make sense to put it in enterprise data centers, because you’ll want the latest technology.

Cloud delivery has definite advantages because the software stack is evolving rapidly, allowing us to deliver new capabilities to everyone at the same time, and because the underlying hardware is improving year after year. There will be quantum computers in enterprise data centers, whether [via] cloud or on-premises provider. It will happen. It just doesn’t make sense in the next few years.

Explain how quantum computing will intersect with AI. We’ve heard that quantum technology is no match for generative AI.

Crowder: It’s a mix. People often use the word AI to refer to the latest trend in AI.

Think about AI in a broader sense [than just generative AI]Yes, there is a direct connection in terms of finding data patterns and complex structure problems, through machine learning or other means. Quantum will automatically do a better job of classification, for example. That’s not generative AI.

Generative AI is the last stage of AI, and that is now the definition of AI for the next year or two until we come up with something else: the next definition of AI. Generative AI has only a tenuous connection to quantum computing. There are people who are doing research and looking to leverage quantum technology in neural networks instead of deep neural networks. I don’t think anything has shown that quantum technology is going to be better in that space. But some researchers believe that could be the case. In the coming years we will find the answer. But up to this point I haven’t seen any data that definitively says “yes.” But I also haven’t seen any data that definitively says “no.”

Don Fluckinger covers digital experience management, end-user computing, CPUs, and other topics for TechTarget Editorial. Do you have advice? Send him an email here.

Leave a Reply

Your email address will not be published. Required fields are marked *