DARPA Hopes To Help Settle Quantum Computing Wild West | AFCEA International

Physicists debate whether quantum computer technology will transform the world as we know it or prove utterly worthless. Either way, researchers at the Defense Advanced Research Projects Agency (DARPA) want to settle the matter.

Quantum computers may offer unbreakable encryption, meaning the country that wins the quantum race could protect its own systems while easily hacking those belonging to other countries. They might also be useful for solving complex logistics problems, developing new materials and pharmaceuticals and analyzing genomes.

Joe Altepeter, DARPA program manager for the Underexplored Systems for the Utility-Scale Quantum Computing (US2QC) program, asserts that asking 10 different physicists what quantum computers will be transformative for will likely result in 11 different answers. “Really, it’s the Wild West of quantum computer science theory. If I think of the 10 smartest physicists that I know, anywhere in the world, probably five of them are convinced that quantum computers are going to be the most transformative technology of the 21st century, and it’s going to change every aspect of everything we do. And the other five are convinced that even if you can build one of these machines—which you definitely will not be able to do because it’s too hard—quantum computers won’t do anything better than your laptop will do it, and so they will be totally useless.”

The state of the art in quantum computing is essentially experimental. Quantum computers use qubits as the basic unit of information rather than conventional computing bits, which are often represented as 1s and 0s.

In simplistic terms, a qubit can be a 1, a 0, or both simultaneously. But the reality is much more complex, Altepeter explained. “A quantum bit can exist as a 0 and a 1, both at the same time. That’s weird enough. But one level deeper, it turns out that it’s a continuum. Not only can you have it be mostly 0 and a little bit 1, or mostly 1 and a little bit 0, but it can be half and half, or really anywhere on that continuum,” he said. “And that’s not actually as complicated as it gets. There’s a whole extra dimension, which we call a phase.”

Scientists use a sphere known as a Bloch sphere to illustrate the concept. If the north pole of the sphere is a 1, and the south pole a 0, qubits can exist as any combination of points anywhere along the inside or outside surface of the sphere. “They’re within some constrained region, but there’s really an infinite number of those things where the classical bit is just the north pole itself,” Altepeter elaborated.

The sheer array of approaches to quantum computing invites the Wild West comparison. Some companies use ions suspended in space by ionic fields. Others have neutral atoms with no ionic charge at all. Other options include quantum dots; single photons in perpetual motion at the speed of light; and tiny, superconducting loops built into a wafer inside a super-frigid freezer at 10 Millikelvins, or about -459 degrees Fahrenheit.

“It’s not like you have competing tech companies all trying to create different versions of the same killer app on your cellphone. They have wildly different physics underpinning their approach to a quantum computer,” Altepeter said. “There’s probably 12 different approaches to build a quantum computer, which are credible, that have serious genius-level people pursuing them and trying to make them work.”

Additionally, it can be hard to tell which solutions are viable. “It’s very easy for hype to completely get out of control. Most of the articles that I read about quantum computing are just completely over the top, overhyped, and are claiming things that I know are definitely not true,” Altepeter reported. “But it’s really difficult to get ground truth on these technologies because it’s so out there.”

With the US2QC program, DARPA hopes to determine whether quantum computers are likely to be the most disruptive technology in the history of mankind, a waste of time and resources, or a helpful solution somewhere along that continuum. “If this technology is somewhere in the zone of completely transformative to totally useless, that sounds like a recipe for, one way or another, us being surprised. And DARPA’s job is to prevent technological surprise,” Altepeter declared.

The program webpage explains that two separate factors bring into question quantum computing’s potential. They are fault tolerance, or the ability to continue working properly despite interference, and a lack of rigorous comparison between quantum algorithms and applications to conventional computing. Resolving both challenges is central to building a quantum computer that is “utility scale,” meaning that its computational value exceeds its costs.

The lack of comparison is simply because quantum and conventional computers behave and solve problems in very different ways. “It’s because in many cases, you’re trying to invent a new way to solve a problem. It’s easy to test a quantum algorithm against a classical computer in the way that a classical computer solves the problem,” he said.

But to illustrate the challenge, he cites an example of using both types of computers to conduct a chemical simulation but asking the quantum computer to do it the same way a classical computer would. “It’s not going to beat a classical computer at its own game. It’s not a computer that’s 1000 times faster but does everything else the same as a classical computer. It’s going to be able to solve the problem in a completely different way that gets the answer in a really surprising pathway.”

Fault tolerance is a challenge because qubits are highly sensitive to electromagnetic signals that “we swim in every day,” and the more electromagnetic interference or “noise” affecting the quantum system, the more likely it is to produce errors. Conventional computers have built-in immunity to that interference, but solving the same problem for quantum computers poses a much larger challenge. Any noise that mucks with the qubits only creates more errors. “The bigger your machine and the longer you want to run it, the more likely those errors just take over and whatever calculation you’re trying to do doesn’t work and just devolves into a whole bunch of noisy static,” Altepeter declared.

One potential solution involves “finding a very special recipe” to get many noisy qubits—tens, hundreds or even thousands—to work together as one. “Then collectively, you’re able to make them all function as one, like, meta-qubit, one computational grade, logical qubit. That is this huge engineering chasm you have to try to get over,” he added.

The US2QC program asks whether novel, or underexplored, solutions might be the answer. The DARPA team metaphorically shot down all but two approaches. Microsoft Corporation, Redmond, Washington, and PsiQuantum Corporation, Palo Alto, California, are facing off in a quantum showdown.

Microsoft is building an industrial-scale quantum system based on a “topological qubit architecture,” which the company theorizes would enable their machine to be small enough to fit in a closet, fast enough to solve problems in a practical timeframe and have the capability to control more than 1 million qubits. PsiQuantum is using “silicon-based photonics to create an error-corrected quantum computer based on a lattice-like fabric of photonic qubits.”

The Microsoft approach begins with improved building blocks in the form of less noisy qubits. “They basically have invented qubits that use the mathematics of topology. If it works, and it’s a huge ‘if’, then they should be able to create qubits that are much higher quality than competing, noisy qubits from other systems, which means they need far fewer of them to get to those computational-grade systems. They should be able to scale up to really big quantum computers faster,” the program manager said.

PsiQuantum, on the other hand, is using thousands of photons working together and entirely avoiding noisy qubits. “It just uses a huge number of photons, but we’re going to make them do a dance in a particular way. You measure them in just the right way and have them interact and interfere with each other in just the right way on an integrated photonic circuit, basically, like a silicon racetrack for photons,” Altepeter spelled out. “If you get the right kind of measurements, and everything does its job, then collectively they act as these big computational qubits.”

The program manager acknowledged the challenge of understanding and explaining how it all works. “As near I can tell, there’s not a good, easy, straightforward way to answer why that happens. It actually takes just a ton of math and a team of engineers to convince yourself that it really works. This lattice of photonic qubits is very difficult to understand without a whiteboard, and ideally, a Ph.D. in quantum photonics—and we’re probably not going to be convinced until we really see them build it and show us that it works,” he said.

The program started with phase zero, proving the solutions offer “a plausible path to the no-kidding, utility-scale computer.” During phase one, the teams will establish milestones and critical metrics for gauging success. Phase two will be dedicated to building the components and subsystems for a fault-tolerant prototype. Phase three will be to build the fault-tolerant prototype or multiple prototypes. The program could go into a fourth phase. The DARPA team left the total number of phases open-ended so that they could “meet the companies where they’re at” and add another phase if necessary.

Both vendors survived phase zero and are now in phase one. “That doesn’t mean we’ve endorsed their approach. But it means we tried really hard to kill the concepts, and we weren’t able to, so they’re still alive,” Altepeter reported.