Topic: Google and NASA ride D-Wave to a quantum future

[devnullius]

FROM: http://www.newscientist.com/article/mg22429983.300-google-and-nasa-ride-dwave-to-a-quantum-future.html?full=true



03 December 2014 by Jacob Aron, Magazine issue 2998.

Short summary by Devvie: not even the D-Wave creators know if they're actually selling quantum based computers  That's quantums for ya' 

Quantum Chill

To find out if Google's D-Wave is the real quantum deal, Hartmut Neven and his colleagues gave it a simple problem: an energy landscape of two valleys, one lower than the other. To find the right solution, the computer must reach the bottom of the lower valley, without getting trapped in the other, giving a false solution (see diagram).

There are two ways the D-Wave could be solving this problem. A computer that employs quantum mechanics should be able to use quantum tunnelling to pass through the hill separating the two valleys, making it more likely to succeed. This works best at low temperatures, when quantum effects are strongest. By contrast, an ordinary machine has a better chance of reaching the right answer at higher temperatures, where it has enough energy to jump over the hill.

Google's data shows the D-Wave had a 75 per cent chance of success operating at temperatures of 15 millikelvin, dropping gradually to around 65 per cent at 35 millikelvin. "The apparent trend between temperature and success probability revealed by these experiments is consistent only with quantum models," write the team.

A New Scientist investigation reveals Google's grand plans for its quantum computer, as well as the first hints about what's really going on under its hood

THEY could be the most powerful computers in the world – so perhaps it's no surprise that the biggest internet company on the planet is testing one out.

Last year Google purchased a quantum computer from D-Wave Systems in British Columbia, Canada, currently the only firm claiming to sell chips powered by exotic physics. However, this claim is controversial; some say D-Wave has yet to fully demonstrate its chips' quantum capabilities. Now a New Scientist investigation reveals Google's future plans, as well as the results of its recent tests to address the quantumness controversy.

In theory, quantum computers offer a huge advantage over ordinary PCs. Regular computers code information in binary bits that are either on or off – 0 or 1. But a quantum "qubit" can be both at the same time. This could let quantum machines crunch through certain problems, like searching a database, at blistering speeds even compared to a supercomputer. Such zippy calculation is an attraction for companies like Google that deal with large volumes of data.

Google certainly isn't alone in its quantum aspirations: its D-Wave Two machine is housed at NASA's Ames Research Center in California and maintained by the Universities Space Research Association (USRA). New Scientist's freedom of information request to see the contract signed between the parties reveals they are pursuing a range of applications.

Aiming high

Their joint aims are easy to state, if difficult to achieve: "The goal is to develop quantum AI [artificial intelligence] algorithms, test them on real world problems and quantify the gains over classical computing machinery."

Then there are the parties' individual goals. Google wants "revolutionary new powerful quantum algorithms" for its core operations. These include ranking search results, personal assistants, ad placement and spam filtering.

Personal assistants may conjure images of a quantum-powered version of Siri, Apple's digital assistant, able to chat and joke with you. But there is no guarantee a quantum computer will be better at this than a regular computer. Google wants to be the first to find out either way.

NASA is after better algorithms for air traffic control, planning rover missions on other planets, analysing data and more. The agency has said it wants to use quantum computers to help in the search for exoplanets.

There are also plans to hook the D-Wave computer up to the NASA Ames supercomputer to develop hybrid quantum-classical AI algorithms.

At least, that's the long-term plan. However, an important stepping stone is to prove the D-Wave really can solve problems faster than ordinary PCs. Now Google researchers, working with others at NASA and D-Wave, say they've found the first evidence that it employs quantum effects to perform computation.

So far, labs around the world have been able to build machines with just a few qubits, which can only handle problems that wouldn't trouble a pocket calculator. But since 2007, when D-Wave Systems first revealed a 16-qubit chip it claimed could solve Sudoku problems by exploiting quantum mechanics, it has repeatedly ramped up the number of qubits in its computers. The D-Wave Two has around 500 qubits and the firm plans to release a 1152-qubit version next year.

So how is D-Wave seemingly able to out-qubit everyone else? The company uses a different approach to most others in the game, called quantum annealing. Rather than shunting qubits through the quantum analogue of the logic gates found in regular computers, it translates problems into a landscape of hills and valleys. D-Wave's qubits explore this landscape to settle on the lowest energy state, which corresponds to the solution. For this to work, the qubits must be cooled as close to absolute zero as possible – the chips are housed in a custom fridge the size of a small room.

However, critics said it wasn't clear the energy-landscape approach would provide an advantage, and had doubts that D-Wave's computers are properly quantum.

Champing at the qubit

One way to prove quantumness is showing that your qubits have a property called entanglement. This can't be measured directly while the D-Wave is operating, so it has to be inferred by other means. The firm has published a number of studies to try and demonstrate such properties, but the crucial question of whether they were actually involved in computation remained open.

Now a team led by Google's Hartmut Neven has revealed what they think is going on under the hood (arxiv.org/1411.4036). They found the computer performed better at lower temperatures – which suggests it was harnessing quantum effects during computation (see "Quantum chill").

Although hopeful, the results don't demonstrate the explosive quantum speed-up promised by theory. But the team stresses that the progress made so far is "a big step" towards proving a speed-up compared to a version of D-Wave that had no quantum properties.

"If the device had failed that test then we'd be sure that there is no chance for quantum speed-up in these types of problems," says Matthias Troyer of ETH Zurich in Switzerland, who has studied D-Wave computers in detail. "They've seen the machine uses quantum mechanics to solve a problem."

But Troyer says Google is likely to be disappointed by these results. "They wanted to write something on the tests of the past year, it's not the real great breakthrough that they had hoped for."

The variation in performance with temperature also means cooling the D-Wave computer is vitally important. But New Scientist obtained reports for the period July 2013 to July 2014 that reveal struggles with the water supply at the NASA Ames facility, with a potential impact on the computer's cooling. "This doesn't mean the system won't work, it just means that it has a higher risk of unplanned down time," says D-Wave's Colin Williams.

"We are working with D-Wave and USRA to address the water supply requirements and closely monitoring the situation to ensure this does not affect the system's performance," says Rupak Biswas at NASA Ames.

That might not be the only problem. Google may be finding it difficult to work with government restrictions enforced by NASA on interacting with Iranian citizens. The July 2013 report says one Google employee, who has dual citizenship with Iran and Canada, was denied an account on the D-Wave computer by NASA security. Google did not respond to requests for comment, and Biswas says NASA is adhering to US government policy.

Google has been upping its own quantum computing operations, separate from D-Wave, by hiring John Martinis of the University of California, Santa Barbara, to build its own quantum annealer.

Whatever the firm's aims, Troyer says new hardware is the only way to achieve quantum speed-up. "I'm very confident that one can build a quantum annealer that works better than a classical annealer for some specific well-chosen problem," he says, but whether those translate into commercially important problems remains to be seen.

Williams says Google's decision to build a quantum annealer shows that D-Wave's approach is the best option. "As we continue our technical interactions with Google I am quite sure we will learn from each other's efforts," he says. "As we have seen many times before, by the time a critic comments on our technology, their assumptions regarding it are invariably already out of date."

This article appeared in print under the headline "Google rides D-Wave"