backtop


Print


  (Source: Warner Bros.)
$10M USD quantum computer outdoes entire data centers, according to Google

Google Inc. (GOOG) owns a dominant position in the search and search advertising markets.  It owns an even more dominant position in the mobile operating system space and it makes one of the most used email services.  And its Linux PC operating system -- Chrome OS -- has seen an explosion of sales and interest.

I. Google Enters the Quantum Age

So when Google in 2009 announced that it was teaming up with Canadian firm D-Wave Systems, Inc. -- the first commercial seller of quantum computing hardware -- and The U.S. National Aeronautics and Space Administration (NASA) -- which verified D-Wave had a true quantum computer -- many took note.

But since then Google's been relatively silent on the collaboration, leaving some to wonder if it'd given up on quantum computing.  But in May the trio announced the Google Quantum Artificial Intelligence Lab.  And now a new short film set to debut at the Imagine Science Films Festival at Google New York is ready to deliver yet another blow to skepticism on the efforts.



A recent August tour of the Quantum AI Lab gave us our first glimpse of the actual D-Wave system in action.  


A black box, otherwise unassuming but for the brightly lit "D-Wave" text emblazoned on the side, the quantum computer operates at extremely cold temperatures thanks to helium cooling.  The video shows a typical operating temperature of 13.8 millikelvin (mK).  By contrast the coldest naturally occurring temperature in space is around 1 K, while the average space temperature is around 2.7 K -- almost 200 times the temperature inside the quantum computer.

Inside is a computer based on D-Wave's proprietary quantum chips.  Here's a video detailing D-Wave's 128 qubit hardware technology:



A D-Wave 2 quantum computer like Google's has 4 128s qubit ("quantum bit") chips, for 512 total qubits of algorithmic capacity.  

II. Optimizing Android

Some of Google's work uses quantum adiabatic algorithms discovered by Edward Farhi and collaborators at Massachusetts Institute of Technology (MIT).  Google showed applications of those algorithms off in 2009 when it used the D-Wave computer to learn visual descriptions of objects locating them with record speed in images previous unseen to the quantum computer.

D-Wave chips
D-Wave is the first company to commercially produce quantum computing hardware.
[Image Source: D-Wave]

The brains at Google, NASA, and D-Wave admit that the hardest challenge is knowing what to ask, in part due to the fact that we're still learning what quantum computers can do.

Quantum computers were thought to be inferior to traditional computers at solving binary or arbitrary node trees (sometimes known as game trees) until Northeastern Univ. Professor Sam Gutmann and MIT's Professor Farhi showed otherwise.  Now quantum algorithms are fast becoming the best tool to solve this class of problems, having also shown themselves amenable to solving binary trees (e.g. the set of all possible moves from a position, the set of all possible moves from those positions and so on) in O(N^(1/2)) time and some more complex cases O(N^(1/2)*log N) time -- much better than the best traditional algorithm (O(N^.753) in the binary case).
Binary game tree
Quantum computers can now solve binary trees faster than traditional ones. [Image Source: Scott Aaronsen]

Comments NASA's Eleanor Rieffel in the new short film, "We don't know what the best questions are to ask that computer.  That's exactly what we're trying to understand."

D-Wave 2
The D-Wave hardware reportedly set Google back an estimate $10M USD. [Image Source: D-Wave]

The clues dropped by Google in the film and elsewhere hint that the most promising application of the Quantum computers may be in algorithm development and in filtering data sets that are too noisy for normal computers to handle.  The company wrote in May:

We’ve already developed some quantum machine learning algorithms. One produces very compact, efficient recognizers -- very useful when you’re short on power, as on a mobile device. Another can handle highly polluted training data, where a high percentage of the examples are mislabeled, as they often are in the real world. And we’ve learned some useful principles: e.g., you get the best results not with pure quantum computing, but by mixing quantum and classical computing.


Google has created optimized power-efficient algorithms for Android, using its quantum computer.

A new test of the Quantum AI lab that Google is working on is to develop algorithms to differentiate between intentional winks and natural blinking, so that blinks can be used to navigate the Glass Explorer wearable computer -- a glasses-form factor wearable that runs a variant of Google's Android operating system.

In April a redittor found snippets of code in Glass Explorer's open source, which hinted at an algorithm to detect blinks.  Mike DiGiovanni the lead developer at Roundarch Isobar in May showed off a hack that allowed developers to use the wink API in their apps -- however it proved relatively inaccurate, requiring exaggerated winks, and often mistaking winks for blinks.  Google is leaving the algorithm off, but it hopes to soon be able to drop in a replacement that the quantum computer's been devising.

Glass Explorer
Google has its eye on Glass Explorer blink detection optimization. [Image Source: TechRadar]

Google says it spent $10M USD in total on the exotic setup -- but that money appears well worth it, as Google says that the quantum computer has generated algorithms faster than entire datacenters it had previously put to the task.  In the future Google wants to use its quantum computers not only to optimize its algorithms, but as a complement to its traditional processing backend, to accelerate some special kinds of searches.

Oh, and one last thing to mention -- Google says it has further evidence that D-Wave's computers are operating on true quantum entanglement, not some lesser equivalent as some doubters speculated -- so take that, haters. 

Sources: Google on YouTube, The Verge





"Intel is investing heavily (think gazillions of dollars and bazillions of engineering man hours) in resources to create an Intel host controllers spec in order to speed time to market of the USB 3.0 technology." -- Intel blogger Nick Knupffer







Latest Blog Posts
The Best Android Apps
Saimin Nidarson - May 20, 2017, 6:16 AM






botimage
Copyright 2017 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki