![\"Embedded](\"https:\/\/science.sciencemag.org\/sites\/default\/files\/highwire\/sci\/365\/6460\/1364\/embed\/graphic-1.gif\")
<\/span><\/div>\nGoogle researchers in Santa Barbara, California, say their advance may lead to near-term applications of quantum computers.<\/p>\n
<\/p>\n <\/p>\n According to the\u00a0Financial Times<\/em>, which broke the story, the paper appeared last week on the website of NASA’s Ames Research Center in Moffett Field, California; some of the researchers there are paper authors. Readers downloaded the manuscript before it vanished, and it is circulating online. John Martinis, the physicist who leads Google’s quantum computing effort in Santa Barbara, California, declined to comment on the paper, but others in the field think it is legitimate.<\/p>\n A quantum computer aims to exploit the strange aspects of quantum mechanics to perform types of calculations that would swamp a classical computer. Whereas a typical computer depends on \u201cbits\u201d of information that can be set to either zero or one, a quantum computer employs qubits that can be set to zero, one, or\u2014thanks to quantum mechanics\u2014zero and one at the same time. That enables a quantum computer to process a multitude of inputs simultaneously. For example, a 10-qubit quantum computer could process 210<\/sup>, or 1024, possible inputs at once instead of analyzing them one at a time.<\/p>\n But such a computer’s real power comes from other quantum phenomena. For certain computational problems, all potential solutions can be thought of as quantum waves simultaneously sloshing among the qubits. Set things up right and those waves interfere with one another so that incorrect answers cancel one another and the right answer pops out. Such interference should enable a full-fledged quantum computer to hack current internet encryption schemes by factoring the huge numbers that underlie them (Science<\/em>, 23 August, p.\u00a0730<\/a>).<\/p>\n That feat would require thousands of qubits, so Martinis and colleagues conceived a problem on which a quantum computer with just dozens of qubits could best any conventional rival. The 53 qubits in their device consist of superconducting metal circuits that can be in a low-energy state to denote zero, a high-energy state to denote one, or both at the same time\u2014at least until measured, when the state collapses one way or the other. The researchers then made pairs of qubits interact in various ways through a fixed but random set of operations.<\/p>\n Taken as a group, the qubits will output any number between zero and 253<\/sup>. Thanks to quantum interference caused by the operations, some numbers should show up more often than others. And as the number of qubits grows, calculating that uneven distribution of outputs would become overwhelmingly difficult for an ordinary computer. So, if experimenters see the telltale unequal pattern of numbers, they have evidence their quantum device calculated something a normal computer cannot.<\/p>\n As with any quantum computing effort, the key was to preserve the qubits’ delicate quantum states throughout the process. If they fuzz out then all outputs become equally likely. But the Google team reports that it managed to see the telltale pattern in the generated numbers. To prove the pattern wasn’t just noise, researchers compared the results for smaller trials and subgroups of the qubits with supercomputer simulations. They couldn’t do that for the biggest instances of the problem, however. What the quantum computer could do in a little over 3 minutes would take a supercomputer 10,000 years to reproduce, they estimate.<\/p>\n Some researchers say the demonstration isn’t so much a computation as an effort to cook up a quantum state that’s hard to simulate. \u201cQuantum computers are not \u2018supreme\u2019 against classical computers because of a laboratory experiment designed to essentially \u2026 implement one very specific quantum sampling procedure with no practical applications,\u201d says Dario Gil, director of IBM Research in Yorktown Heights, New York, which is also developing machines with superconducting qubits.<\/p>\n The Google computer also lacks the ability to correct errors, which may be key to making a full-fledged quantum computer. That requires encoding a single, more stable \u201clogical\u201d qubit in several less reliable \u201cphysical\u201d ones, to enable the machine to maintain quantum states much longer, Kuperberg explains. However, Rigetti notes that Google’s achievement may put the company in an ideal position to demonstrate such error correction, too.<\/p>\n Gil voices another worry held by many the field: that after all the hype surrounding quantum supremacy, quantum computing may experience a letdown like the one that plagued artificial intelligence from the 1970s until the current decade, when technology finally caught up with aspirations. In the leaked paper, however, the 76 authors conclude: \u201cWe are only one creative algorithm away from valuable near-term applications.\u201d<\/p>\n <\/p>\n <\/p>\nPHOTO: JHVEPHOTO\/ISTOCK.COM<\/q><\/div>\n<\/div>\n