Are powerful quantum computers finally here? Google is claiming that its D-Wave quantum computer can solve certain problems , a result that it says could lead to huge improvements in artificial intelligence. But researchers contacted by 快猫短视频 say these claims are overblown, and show little-to-no performance increase over a regular PC. Before you get too excited about a new era of computing, read on.
What has Google done?
In 2013, Google purchased a quantum computer from D-Wave of Burnaby, Canada, the only company that sells such devices. D-Wave鈥檚 machines have been controversial amongst academics, and previous claims of faster computation have been called into question. Now Google has published its strongest results yet, showing a 100 million times speed advantage.
Advertisement
How does the D-Wave computer work?
D-Wave鈥檚 computer is a specialised device called a quantum annealer, which works by exploring an energy landscape of hills and valleys that corresponds to the problem it is trying to solve. The goal is to reach the lowest point on this landscape, which corresponds to the best solution. A property called quantum tunnelling allows the D-Wave to traverse this landscape more quickly by 鈥渢unnelling鈥� through the hills, thus theoretically letting it reach an answer faster.
How did Google achieve such a massive speed boost?
The Google team compared the D-Wave to an ordinary computer running simulated annealing software. As the name suggests, this software attempts to simulate the energy landscape exploration, but unlike the D-Wave, it cannot take shortcuts by tunnelling, making it slower. Crucially, Google tested both computers with large problems, which should reveal differences in performance.
鈥淲e found that for problem instances involving nearly 1000 binary variables, quantum annealing significantly outperforms its classical counterpart, simulated annealing,鈥� wrote Google鈥檚 Hartmut Neven in a . 鈥淚t is more than 108 times faster than simulated annealing running on a single core.鈥�
What do others say?
鈥淵ou need to read the fine print,鈥� says of the Swiss Federal Institute of Technology in Zurich. 鈥淭his is 108 times faster than some specific classical algorithm on problems designed to be very hard for that algorithm but easy for D-Wave.鈥� In other words, the D-Wave had a massive home advantage.
Better versions of the simulated annealing algorithm can reduce this advantage to just 100 times faster, says Troyer, while other more complex algorithms running on an ordinary PC can beat D-Wave entirely. 鈥淎 claim of 鈥�108 speedup鈥� is thus very misleading,鈥� he says.
You should also keep in mind that the D-Wave is a specialised piece of hardware that costs around $10-15 million. As the Google team admit in a , a similarly specialised but non-quantum device might be able to match the current D-Wave computer, though they think it isn鈥檛 worth exploring this avenue as they believe larger quantum computers will soon exceed these capabilities.
So is Google trying to pull the wool over our eyes?
Not quite. Neven admits at the end of his blog post that other algorithms can beat the D-Wave, but the Google team think this advantage will disappear as quantum computers get larger.
Others are less sure. 鈥淭his is certainly the most impressive demonstration so far of the D-Wave machine鈥檚 capabilities,鈥� says of the Massachusetts Institute of Technology. 鈥淎nd yet, it remains totally unclear whether you can get to what I鈥檇 consider 鈥榯rue quantum speedup鈥� using D-Wave鈥檚 architecture.鈥�
Will we ever have definitive proof of speedy quantum computers?
Maybe 鈥� but it is likely to require new hardware. D-Wave鈥檚 goal has always been to get quantum computers to market as fast as possible, but Aaronson thinks their 鈥渜uantum coherence鈥� 鈥� a measure of the fragile quantum states necessary for computation 鈥� isn鈥檛 as good as that found in quantum chips built by teams who are taking a more slow and steady approach.
Google is hedging its bets 鈥� it has hired external researchers to build its own quantum chip. And IBM recently received US government funding to develop its own version.
Both of these seem more promising, says Aaronson. 鈥淭hings are happening a good deal faster than I鈥檇 expected,鈥� he says. 鈥淏ut it remains as important as it ever was to separate out the marketing buzz from what鈥檚 actually been shown.鈥�
Image credits, from top: D wave; NASA
