简体中文
Google showed off its work on D-Wave’s 2X quantum annealer computer at NASA’s Ames Research Center on Dec. 8, 2015.Alistair Barr for The Wall Street Journal
(Wall Street Journal) A computer that’s millions of times faster than the most powerful machine available today could improve everything from climate and disease research to understanding the contents of every YouTube video.
That’s the promise of another ambitious, long-term Google project that the Internet giant opened up about this week. Like other so-called moonshots, this one could take a decade or more to produce anything of tangible value, according to experts in the field. But Google, a division of
Alphabet
is hopeful.
“We think this kind of investment in fundamental research in computing is very, very important,” said John Giannandrea, who leads Google’s artificial-intelligence work. He said Google often runs up against problems that conventional computers can’t solve, “and this is Google which has a lot of computers.”
Much more powerful computers could one day help Google’s search engine learn and improve in real time as people pose questions and click on specific answers and web links. Google now has thousands of engineers analyzing such behavior and tweaking its search algorithm. Such fast learning is part of the promise of this research, a Google spokesman said.
The computer Google is working on is made by D-Wave Systems Inc., a Canadian startup, some of whose claims have been questioned by experts. The machine is installed in the super-computing lab at NASA’s Ames Research Center near Google’s headquarters in Mountain View, Calif.
The D-Wave 2X is a primitivetype of “quantum” computer that uses radically different technology than classical computers. Existing machines rely on bits representing ones and zeros to complete tasks. In contrast, a quantum bit, or “qubit,” pushes the boundaries of physics by representing a one and a zero at the same time. Many qubits working together at the same time would in theory create a computer that performs some calculations exponentially faster.
This week, Google
unveiled the results
of carefully designed test in which the D-Wave 2Xsolved a problem 100 million times faster than a traditional computer.
The tests “confirm there’s quantum-ness in the D-Wave machine and that’s significant,” said Massachusetts Institute of Technology professor Edward Farhi, who studies the use of quantum mechanics in computing.
Matthias Troyer, an expert in this type of quantum computer at Swiss university ETH Zurich who has questioned D-Wave claims in the past, said Google chose the test to favor its computer over a classic computer, and not even a classic supercomputer. “If it had failed then we’d know it was useless. So now we know it’s not useless and it might be useful in the future,” the professor added.
In a blog post, Google also acknowledged that some classic computing techniques can solve the problem in its test “faster than the current quantum hardware,” but said it’s working on improvements.
Google also described the limitations of the 2Xat press conference earlier this week. It can only be designed to perform a single, specific task; it needs near-absolute zero temperatures to work and was at risk of blowing up when it had to be cleaned recently.
However, Google’s quantum researchers — and other experts — think this type of quantum computer could solve complex problems with many variables that conventional machines can’t handle. Dr. Farhi said it could help airlines cut delays by picking the best locations for aircraft and crew.
Google thinks this type of quantum computing will first be useful in machine learning, a field of artificial intelligence that uses data to teach computers to be smarter.
Traditional computers need to be shown a lot of well-organized and labeled information to learn, which takes a long time. More powerful quantum computers could learn much quicker on smaller amounts of “dirty,” or un-labeled data, said Hartmut Neven, the head of Google’s quantum research group.
Other technology companies, such as
International Business Machines
, are working on “universal” quantum computers capable of handling many different tasks. This is the holy grail, and another group of researchers at Google, led by John Martinis, is pursuing this. That will take another 15 to 20 years, Dr. Troyer said.
