IBM, one of the world’s oldest tech companies, is building a refrigerator. That, in itself, is not unprecedented. Other tech companies have built fridges before. LG sells the impressive Wi-Fi-connected LG InstaView Door-in-Door Smart Fridge. Samsung, another global device maker, makes the excellent RF23J9011SR 4-Door Flex with Power Cool feature.
But IBM’s fridge (still in development) is different. Very different, in fact. It will be enormous for one thing: 10 feet tall and 6 feet wide. It will also be unimaginably cold, around 15 millikelvin, or -459 Fahrenheit, which is colder than outer space. It’s also named after a James Bond movie, Goldeneye.
However, the biggest difference between it and your run-of-the-mill kitchen fridge is its planned contents. Don’t expect a built-in egg holder, vegetable drawers, and space for your seasonal eggnog. Instead, it’s going to be home to the world’s first 1-million-qubit quantum computer — once that’s built as well.
“For the quantum effects to emerge, [quantum computers] need to be cooled down to extremely low temperatures,” Jerry Chow, director of Quantum Hardware System Development at IBM, told Digital Trends. “In fact, all the infrastructure that goes around even just the processor itself requires a fair amount of cooling, especially as you scale it up, right?”
It was this scaling-up process that led Chow and his team to the inescapable conclusion that IBM really needed to get into the refrigeration business — at least when it comes to its own quantum computers. For one thing, there’s a limit to current cooling capacity. Then there are problems with things like maintaining vacuum integrity and balancing the weight of the various components needed for chilling. The computer scientist Alan Kay once said that the company serious about software should also build its own hardware. Perhaps the quantum equivalent of this should be that the company serious about quantum computing should not only build its own quantum computer, but its own fridge to house it.
“If we just do some back-of-the-envelope scaling, you start to see that, at some point, what you can get from the commercial vendors falls short,” Chow said. “You have to start thinking about how do you push beyond [that]?”
IBM’s super fridge is, on some level, a red herring. It’s a bit like building a fancy new garage for the Tesla you’re having delivered. Sure, that fancy remote control garage door you’ve installed is exciting — but it’s not the exciting bit. In this analogy, the new Tesla Model S or Cybertruck is IBM’s planned one-million qubit quantum. And, provided IBM can build it as planned, it’ll be a doozy, more than worthy of the world’s most sophisticated refrigerator.
Quantum computers were first proposed in the 1980s by the American physicist Paul Benioff, although the quantum mechanics upon which they are based date back to the 1920s, when physicists began to notice that certain experiments didn’t produce the results they had predicted using their current understanding of physics. Richard Feynman, David Deutsch, Yuri Manin, and others seized upon the idea of a quantum mechanical model of a Turing machine, suggesting that a quantum computer could be used to simulate things that simply cannot be simulated through a classical computer using classical physics. In 1994, Dan Simon showed that a quantum computer could be exponentially faster than a classical computer.
One of the big differences with quantum is the concept of superposition. A classical computer can be either a state of A or B (or, in binary terms, one or zero). A quantum computer can be a mixture of the two. (That’s the Schrödinger’s cat thought experiment in which a cat in a box could be either alive, dead, or both alive and dead simultaneously.) Then there are other concepts such as collapse, uncertainty, and entanglement, which make quantum computers very different from the ones you and I grew up on.
In the same way that a classical computer operates on bits, quantum computers operate on what are referred to as qubits. At present, IBM’s current largest quantum computer has 65 qubits. By 2023, it wants to build one with 1,000 qubits. And sometime after that — a date the company will not commit to, but which is certainly on its road map — it will build a 1-million-qubit machine.
Jumping from 65 qubits to a million qubits is quite the leap. But computing, even classical computing, turns out to be pretty good when it comes to exponential leaps. Moore’s Law states that the number of transistors that can fit onto a circuit board doubles approximately every two years. The closest thing quantum has to Moore’s Law is what is referred to as Rose’s Law, formulated by Geordie Rose in 2002. Rose’s Law states that the number of qubits in a quantum computer doubles every couple of years.
Compared to Moore’s Law, the implications of Rose’s Law are arguably even more profound because, as Peter Diamandis and Steven Kotler observe in their book The Future Is Faster Than You Think: How Converging Technologies Are Transforming Business, Industries, and Our Lives, qubits in superposition have far more power than the binary bits in transistors.
Because “more” doesn’t always equal “better,” one of IBM’s conceptual tweaks to this notion is based around the more nuanced concept of what IBM calls quantum volume. “It’s not just about scaling the physical number of qubits,” Chow said. “In the end, it’s about both the number of qubits and how well they perform; how large of a circuit can you actually run on that hardware before the qubits decohere and your quantum information disappears. Quantum volume is such a metric.”
“Everything we call real,” said Niels Bohr, one of the founding figures of quantum mechanics, “is made of things that cannot be regarded as real.” Given the premise of quantum superposition, it is perhaps appropriate that quantum computers today exist in a strange twilight world of here and not here. IBM is just one of the companies to have built functioning quantum computers (Google, Baidu, are Amazon are some of the other big names.) There are quantum algorithms, too — in some cases, ones that cannot yet be run effectively on the quantum computers people have built.
And yet, for all the proof of concepts and causes for excitement, it’s fair to say that the world has not yet begun to come close to tapping the enormous power of quantum computing. “What [quantum computing] entails in terms of actual applications is still not fully known,” Chow said.
“This holy trinity of future technologies is made up of quantum computing, artificial intelligence, and the cloud.”
Some of the most exciting potential use cases — whether it’s computational chemistry, financial modeling, cybersecurity and cryptocurrency, or advanced forecasting — remain ghosts in the quantum machine. For now, at least.
Why is IBM focused on quantum computation? “Our focus is on how we deliver the future of computation,” said Chow. Quantum is an unavoidable part of that future.
Quantum computing is one of IBM’s three big bets for the future. This holy trinity of future technologies is made up of quantum computing, artificial intelligence, and the cloud. But these are not individual bets as would be the case if you were to invest your savings in three promising startups, believing that one of the three has the chance of becoming a unicorn that will more than offset any losses incurred by the other two.
Quantum, for example, could be a game-changer for A.I. There is no doubt that artificial intelligence — and, most specifically, machine learning — has enjoyed astonishing advances using classical computing architecture. But quantum promises to speed things up even more. Quantum versions of current machine learning algorithms (or, more likely, entirely new, much faster alternatives) will be able to carry out enormous data-driven A.I. calculations at a significantly faster rate. They will be able to handle the mind-boggling number of dimensions that arise from data and map them in the large quantum feature space. Quantum entanglement could be used to discover fresh patterns that can’t be discovered with traditional classical computing.
The cloud is also a fundamental part of IBM’s quantum bet. Broadly speaking, the popular progression of classical computing was a transition from mainframes to minicomputers to personal computers. In the 1950s, people only had access to enormous computers in large, air-conditioned rooms. By the late 1970s and ’80s, people had computers in their homes. By the 1990s, people had laptop computers they could carry in their bags. Today, we have computers in the form of smartphones that we carry in our pockets.
It seems unlikely that quantum computers will experience the same shift in form factor due to the requirements (such as extreme cooling) for a quantum computer.
“In terms of [having a physical quantum computer] on your desk, I may be wrong, but it’s not clear to me that that will be the case,” said Chow. “Most of the systems that you build which require this level of quantum coherence, be it a superconducting system or trapped ions, all require a fair bit of infrastructure for you to maintain them — and especially as you scale up.”
But this is where the disruption of cloud computing enters the picture. Cloud computing means that users have access to supercomputer capabilities regardless of whether they are in the same physical vicinity. Compute power or storage is no longer limited to the hardware that is available on your desk the way that it was 20 years ago.
“So much today is done over the cloud [and] people don’t even notice,” Chow said. “How many times do people realize that something isn’t processed on their own laptops or on their own phones, but somewhere else? That’s how quantum over the cloud is going to work.”
It is, to a degree, how quantum computing is already working. In May 2016, IBM launched its Quantum Experience, a five-qubit quantum processor and connected matching simulator that lets users carry out experiments on a quantum computer system. To date, IBM Quantum has deployed 32 quantum processors on the cloud, with more than 280,000 users worldwide collectively running upward of 1 billion quantum circuits daily. As more powerful quantum computers are made available, these too will be accessible to users via the cloud.
“You’re going to have problems that are naturally solved using the best techniques that we know in traditional computers,” Chow said. “But then there are also parts of these problems which are too complex to solve [with even high-performance computing systems] today that might be suitable for quantum computers.”
No, you won’t be running your Excel spreadsheet on a quantum computer any time soon (if ever). Classical computers can run Excel just fine. But parts of applications could certainly harness quantum capabilities, whether for things like encryption or better machine learning. There could even be some more fascinatingly frivolous examples. For example, James Wootton, another IBM engineer, is using quantum computing to do random terrain generation within computer games. Ever dreamed of a game that could totally reconfigure itself each time you play to an unimaginable degree? Quantum’s your answer.
“This is what we mean by the hybrid cloud computational model,” Chow said. “You will have your problem workload that’s fed into a computer and the right parts go to a classical computer and the other parts go to a quantum computer. Then out comes a solution. That’s the picture that you can imagine in the future. [Quantum is] is not a replacement [for classical computers], but they will certainly work hand in hand.”
IBM won’t commit to when exactly it will deliver its million qubit computer — or, for that matter, when its Goldeneye fridge will be finished. But it’s pretty clear about its belief that quantum computing is going to be a game-changer.
In a post written for IBM’s blog earlier this year, Jay Gambetta, IBM fellow and vice president of quantum computing, likened the next generation of IBM quantum computers to the Apollo missions that resulted in the moon landing. That’s quite the comparison. It may also be accurate.
Here in 2020, with the prospect of a new moon landing tantalizingly closer than it’s been in decades, that sounds like a far more upbeat comparison than it might have even just a few years ago. It should be well worth the wait.
- Why now is the worst time to build a PC in nearly 8 years
- Stop worrying so much about benchmarks when buying a new GPU
- Microsoft quits its creepy, emotion-reading A.I.
- Metaverse giants form new standards to address lack of interoperability
- The Paper Laptop could be the e-paper device you didn’t know you needed