Honeywell Takes Tiny Steps Toward a Quantum Leap in Computing
LISTEN TO ARTICLE
SHARE THIS ARTICLE
A decade ago, two scientists atHoneywell International Inc. pitched their bosses a most improbable idea: The century-old industrial company should build a quantum computer. The idea was met with some skepticism. For one thing,International Business Machines Corp. andGoogle were already moving ahead on quantum computing, which promises to be much faster, for certain tasks, than even the speediest transistor-based chips. And Honeywell was better known for making cockpit controls, security systems, and even work boots.
Yet management signed off, and scientists set up a laboratory in a one-story brick building outside Boulder, Colo. Now, Honeywell is pursuing its own path to producing a practical quantum computer that can be used for everyday business. On Oct. 29 it unveiled what it dubbed the H1, its second-generation quantum computer.JPMorgan Chase & Co.,BP Plc, and other companies are paying for subscriptions to write algorithms on the machine and test those results with classical supercomputers.
These deals are very much experiments—Honeywell’s clients aren’t using its quantum machine to make business decisions, and any commercial quantum computer is years away. And Honeywell’s achievements are less impressive than Google’s boast last year that it had achieved “quantum supremacy” after its machine performed a calculation that chokes even the largest supercomputer. IBM first made a quantum computer available to the public in 2016 and has rolled out regular upgrades.
Honeywell makes a claim that’s impressing industry observers: Its machine, though slower, appears to produce more accurate results. In this early stage all quantum computers make errors, or noise, as scientists call it. Honeywell has been able to minimize them. “Quality is really the holy grail,” says Denise Ruffner, who previously worked at IBM’s quantum unit and just started as a vice president at IonQ, a startup quantum computer maker. “What any user really wants is a quantum computer with the least amount of error or noise. So far, the technology that Honeywell is using has been the best at doing that.”
Quantum computing is based on the mind-bending concept that atomic particles can be in multiple states at once. Conventional computers depend on electric currents being switched on or off—states that are represented by a “1” and a “0.” Quantum computers use units known as quantum bits, or qubits, which can represent both a “0” and a “1” at the same time. Generally, quantum computers look at all the different combinations of data simultaneously, while classical computers work sequentially, making them much slower when dealing with large sets of variables.
But errors are prevalent. Honeywell decided early on to use existing atoms as the basis of its computer, a technology known as trapped ion, to minimize errors. Google, IBM, and others are assembling tiny devices that simulate particles using electric currents at extremely cold temperatures, a method called superconducting. This technology makes for a faster computer that can be scaled up in power using the industry’s existing supply chain. But it’s also more error-prone. “Any design element is looking at a trade-off,” says Ilyas Khan, chief executive officer of Cambridge Quantum Computing, a software startup that counts Honeywell and IBM as investors. “If you wanted a speedster, you’d be looking at your superconducting quantum machine. And if you’re more looking at accuracy and you didn’t care about speed, you’d be looking at the ion trap.”
Honeywell CEO Darius Adamczyk predicts that within 10 years its quantum computing business will have sales of $1 billion (the company had about $37 billion in revenue last year). “Over time, the market is going to be worth billions,” he says.
One skeptic is Bob Sutor, vice president for quantum solutions at IBM, who says Honeywell’s trapped-ion technology will have trouble scaling up. “We believe we can get through the scaling problem,” he says. “The ion trap people will face that. They will hit it, and they will hit it pretty hard in the not-too-far-off future.”
JPMorgan Chase is running software programs on Honeywell’s computer with small sets of data that can be checked with a standard computer, helping boost confidence on the accuracy, says Marco Pistoia, chief of the bank’s Future Lab for Applied Research and Engineering. He says he’s eager to run more complex algorithms as the computers become more powerful so JPMorgan can be ready when the machines routinely solve problems that a standard computer can’t.
Both approaches are far from finished. “Superconducting qubits have come a long way and today serve as a proof point of the potential that quantum computing promises,” says James Sanders, an analyst at S&P Global Market Intelligence’s 451 Research. “Proponents of trapped ions tout increased accuracy in comparison, but these are presently slower to operate. Both need further development to eclipse the capabilities of classical computers.”
Read next: 9-Year-Old Coders Show How India Is Pioneering Online Education
Source: Read Full Article