GoogleServe 2013: Giving back on a global scale

Every year in June comes a week where Googlers around the world stop reviewing code, ignore their inboxes and leave their cubicles behind to participate in GoogleServe, our global week of service.

This year, more than 8,500 Googlers from 75+ offices participated in 500 projects. Not only was this our largest GoogleServe to date, but it was also one of the more unique, as many projects were designed to expand the notion of what it means to give back to the community. Here’s a glimpse at some of what we were up to this year:

• In Thimphu, Bhutan, Googlers led a workshop about media literacy at the Bhutan Centre for Media and Democracy helping youth prepare to participate in shaping the future of this young democracy.
• Googlers in Mountain View, Calif., created a bone marrow donation drive and partnered with the Asian American Donor Program to raise awareness about the need for more donors from diverse racial and ethnic backgrounds.
• Googlers from our Hyderabad, India office volunteered at Sri Vidhya's Centre for the Special Children, helping children who suffer from a wide range of cognitive disabilities to learn how to identify colors, write their own names, and prepare meals for themselves.
• A team of Googlers walked the New York, N.Y., streets gathering information to improve AXS Map, a crowd-sourced platform for mapping wheelchair accessibility which is populated with data from Google Maps and Google Places APIs.
• In Lagos, Nigeria, Googlers mentored entrepreneurs at Generation Enterprise, a small business incubator that equips at-risk youth to start sustainable businesses in slum communities.
• In Randwick, Australia, Googlers taught computer and Internet skills with the Australian Red Cross Young Parents Program which aims to develop the capacities of young parents to live independently and to parent successfully.
• A group of gourmet Googlers cooked a meal for families with children undergoing cancer treatment with Ronald McDonald House in London, U.K.
• Googlers tutored and mentored youth in Kuala Lumpur, Malaysia, with the Dignity For Children Foundation.
• Googlers partnered with Un Techo Para Mi País to help build a new house for a family living below the poverty line in Bogota, Colombia.
• In Dublin, Ireland, Google engineers taught youth how to program interactive stories and games with Scratch in partnership with Coder Dojo.

Click for more photos from this year's GoogleServe

Over the past six years, GoogleServe has transformed from a single week of service into a week of celebration and inspiration for ongoing giving. Googlers also give back year-round through our GooglersGive programs which include 20 hours of work time annually to volunteer with an approved charitable organization. If you’re inspired to join us, please check out All for Good or VolunteerMatch for opportunities to give back in your community.

Posted by Zanoon Nissar, on behalf of the GoogleServe Global Leadership Team

Connecting across continents

There’s only so much students can learn about the world from the static pages of a textbook. Meeting people from other countries face-to-face provides unique insight into the world’s varied cultures, and the Internet is making this possible in unprecedented ways. To increase global connections, we’re working with First Lady Michelle Obama, the State Department and the Global Nomads Group, to connect students across continents over Google+ Hangouts.

As a keystone event in The White House’s Africa Tour, the First Lady will host a Google+ Hangout On Air from the SciBono Discovery Center in Johannesburg this Saturday at 9:30 a.m. EDT. After Mrs. Obama shares her thoughts on the importance of education, students in Johannesburg, L.A., Houston, New York, and Kansas City will get the chance to talk with one another directly, sharing ideas about education in their countries face-to-face-to-face—it’s a 21st-Century pen pal program, hosted by the First Lady. (RSVP to watch.)

The discussion won’t stop there. This Hangout On Air kick-starts a series of global exchanges on Google+, organized by the State Department and the Global Nomads Group, a nonprofit organization that facilitates cultural exchanges, launching early in the new school year. During the summer, students are encouraged to join the Global Nomads Group’s Google+ Community, “Connecting Continents,” to discover and connect with peers around the world. We look forward to announcing the next hangouts in the near future—stay tuned to the Global Nomads Community for details.

Posted by Will Houghteling, Lead, Google+ Education Partnerships

Meet 15 Finalists and Science in Action Winner for the 2013 Google Science Fair

Creating a world-class science project is no easy task, but this year thousands of 13-18 year olds from more than 120 countries submitted their project to the third annual Google Science Fair. After further judging and deliberation, today we’re announcing the 15 finalists from our top 90 regional finalists, as well as the winner of the Scientific American Science in Action Award.

From the creation of an exoskeletal glove to support the human hand to managing the impact of infrastructure projects on endangered species to an early-warning system for emergency vehicles, the caliber, ingenuity and diversity of this year’s projects is a testament to the fact that young minds really can produce world-changing ideas.

The 15 finalists will join us at our Mountain View headquarters on September 23 to present their projects to an international panel of esteemed scientists for the final round of judging. The Grand Prize winner will receive a 10-day trip to the Galapagos Islands with National Geographic Expeditions, $50,000 in scholarship funding and more.

Congratulations to our finalists:

Age 13-14
Alex Spiride (USA): Squid-Jet: Bio-Inspired Propulsion System for Underwater Vehicles
Venkat Sankar (USA): Ecology or Economy: Managing the Impact of Infrastructure Projects on Endangered Species
Kavita Selva (USA): Superconductor Tapes: A Solution to the Rare Earth Shortage Crisis
Liza Sosnova and Tina Kabir (Russia): Lyytinen - Universal hydrostatic densitometer
Viney Kumar (Australia): The PART (Police and Ambulances Regulating Traffic) Program

Age 15-16
Elif Bilgin (Turkey): Going Bananas!-Using Banana Peels in the Production of Bio-Plastic As A Replacement of the Traditional Petroleum Based Plastic
Ann Makosinski (Canada): The Hollow Flashlight
Yi Xi Kang, Kwok Ling Yi and Tricia Lim (Singapore): Efficacy of Estrogens and Progesterone in Hepatic Fibrosuppression
Valerie Ding (USA): Rapid Quantum Dot Solar Cell Optimization: Integrating Quantum Mechanical Modeling and Novel Solar Absorption Algorithm
Shrishti Asthana (India): Solar Light Assisted nanoZnO Photo Catalytic Mineralization - The Green Technique for the Degradation of Detergents

Age 17-18
Charalampos Ioannou (Greece): An Exoskeleton Glove which Enhances and Supports the Movement of the Human Palm
Esha Maiti (USA): Stochastic Monte Carlo Simulations to Determine Breast Cancer Metastasis Rates from Patient Survival Data
Elizabeth Zhao (USA): A Novel Implementation of Image Processing and Machine Learning for Early Diagnosis of Melanoma
Eric Chen (USA): Computer-aided Discovery of Novel Influenza Endonuclease Inhibitors to Combat Flu Pandemic
Vinay Iyengar (USA): Efficient Characteristic 3 Galois Field Operations for Elliptic Curve Cryptographic Applications

We’re also announcing the winner of the Scientific American Science in Action Award, which honors a project that makes a practical difference by addressing an environmental, health or resources challenge. An independent panel has selected Elif Bilgin from Turkey for this award for her work using banana peels to produce bioplastics. Congratulations to Elif, who will receive $50,000 and and a year’s worth of mentoring from Scientific American to help develop her project. Elif’s project is also one of the 15 finalists, and she is still in the running for the Grand Prize Award.

Which of the 15 finalist projects do you think has the potential to change the world? While the official judges will decide the 2013 Grand Prize Winner, in August you’ll be able to participate in this year’s competition by voting for the Voter's Choice Award. Visit the Google Science Fair website August 1-30 to vote for the project you think has the greatest potential to change the world.

Check back for more details, and tune in live to see the finalist gala on September 23, which will be broadcast on our website, Google+ page and YouTube channel. Congratulations to all our finalists. We look forward to meeting in Mountain View!

Posted by Clare Conway, Google Science Fair team

Update July 30: Updated the name of the Voter's Choice Award (previously the Inspired Idea Award).

Fast, Accurate Detection of 100,000 Object Classes on a Single Machine

Posted by Tom Dean, Google Research

Humans can distinguish among approximately 10,000 relatively high-level visual categories, but we can discriminate among a much larger set of visual stimuli referred to as features. These features might correspond to object parts, animal limbs, architectural details, landmarks, and other visual patterns we don’t have names for, and it is this larger collection of features we use as a basis with which to reconstruct and explain our day-to-day visual experience. Such features provide the components for more complicated visual stimuli and establish a context essential for us to resolve ambiguous scenes.

Contrary to current practice in computer vision, the explanatory context required to resolve a visual detail may not be entirely local. A flash of red bobbing along the ground might be a child’s toy in the context of a playground or a rooster in the context of a farmyard. It would be useful to have a large number of feature detectors capable of signaling the presence of such features, including detectors for sandboxes, swings, slides, cows, chickens, sheep and farm machinery necessary to establish the context for distinguishing between these two possibilities.

This year’s winner of the CVPR Best Paper Award, co-authored by Googlers Tom Dean, Mark Ruzon, Mark Segal, Jonathon Shlens, Sudheendra Vijayanarasimhan and Jay Yagnik, describes technology that will enable computer vision systems to extract the sort of semantically rich contextual information required to recognize visual categories even when a close examination of the pixels spanning the object in question might not be sufficient for identification in the absence of such contextual clues. Specifically, we consider a basic operation in computer vision that involves determining for each location in an image the degree to which a particular feature is likely to be present in the image at that particular location.

This so-called convolution operator is one of the key operations used in computer vision and, more broadly, all of signal processing. Unfortunately, it is computationally expensive and hence researchers use it sparingly or employ exotic SIMD hardware like GPUs and FPGAs to mitigate the computational cost. We turn things on their head by showing how one can use fast table lookup — a method called hashing — to trade time for space, replacing the computationally-expensive inner loop of the convolution operator — a sequence of multiplications and additions — required for performing millions of convolutions with a single table lookup.

We demonstrate the advantages of our approach by scaling object detection from the current state of the art involving several hundred or at most a few thousand of object categories to 100,000 categories requiring what would amount to more than a million convolutions. Moreover, our demonstration was carried out on a single commodity computer requiring only a few seconds for each image. The basic technology is used in several pieces of Google infrastructure and can be applied to problems outside of computer vision such as auditory signal processing.

On Wednesday, June 26, the Google engineers responsible for the research were awarded Best Paper at a ceremony at the IEEE Conference on Computer Vision and Pattern Recognition held in Portland Oregon. The full paper can be found here.

Securing your WiFi network

This post is part of a regular series of privacy and security tips to help you and your family stay safe and secure online. Privacy and security are important topics—they matter to us, and they matter to you. Building on our Good to Know site with advice for safe and savvy Internet use, we hope this information helps you understand the choices and control that you have over your online information. -Ed.

More than a quarter of Internet users worldwide use WiFi at home to connect to the web, but many aren't sure how to protect their home network, or why it is important to do so. The best way to think of your home WiFi network is to think of it like your front door: you want a strong lock on both to ensure your safety and security.

When data is in transit over an unsecured WiFi network, the information you’re sending or receiving could be intercepted by someone nearby. Your neighbors might also be able to use the network for their own Internet activities, which might slow down your connection. Securing your network can help keep your information safe when you’re connecting wirelessly, and can also help protect the devices that are connected to your network.

If you’re interested in improving your home WiFi security, the steps below can help make your home network safer.

1. Check to see what kind of home WiFi security you already have.
Do your friends need to enter a password to get on your network when they visit your house for the first time and ask to use your WiFi? If they don’t, your network isn’t as secure as it could be. Even if they do need to enter a password, there are a few different methods of securing your network, and some are better than others. Check what kind of security you have for your network at home by looking at your WiFi settings. Your network will likely either be unsecured, or secured with WEP, WPA or WPA2. WEP is the oldest wireless security protocol, and it’s pretty weak. WPA is better than WEP, but WPA2 is best.

2. Change your network security settings to WPA2.
Your wireless router is the machine that creates the WiFi network. If you don’t have your home network secured with WPA2, you’ll need to access your router’s settings page to make the change. You can check your router’s user manual to figure out how to access this page, or look for instructions online for your specific router. Any device with a WiFi trademark sold since 2006 is required to support WPA2. If you have a router that was made before then, we suggest upgrading to a new router that does offer WPA2. It’s safer and can be much faster.

3. Create a strong password for your WiFi network.
To secure your network with WPA2, you’ll need to create a password. It’s important that you choose a unique password, with a long mix of numbers, letters and symbols so others can’t easily guess it. If you’re in a private space such as your home, it’s OK to write this password down so you can remember it, and keep it somewhere safe so you don’t lose it. You might also need it handy in case your friends come to visit and want to connect to the Internet via your network. Just like you wouldn’t give a stranger a key to your house, you should only give your WiFi password to people you trust.

4. Secure your router too, so nobody can change your settings.
Your router needs its own password, separate from the password you use to secure your network. Routers come without a password, or if they do have one, it’s a simple default password that many online criminals may already know. If you don’t reset your router password, criminals anywhere in the world have an easy way to launch an attack on your network, the data shared on it and the computers connected to your network. For many routers, you can reset the password from the router settings page. Keep this password to yourself, and make it different from the one you use to connect to the WiFi network (as described in step 3). If you make these passwords the same, then anyone who has the password to connect to your network will also be able to change your wireless router settings.

5. If you need help, look up the instructions.
If you’ve misplaced your router’s manual, type the model number of your base station or router into a search engine—in many cases the info is available online. Otherwise, contact the company that manufactured the router or your Internet Service Provider for assistance.

Please check out the video below to learn more about the simple but important steps you can take to improve the security of your Internet browsing.



For more advice on how to protect yourself and your family online, visit our Good to Know site, and stay tuned for more posts in our security series.

Posted by John Munoz, Technical Program Manager

Only clear skies on Google Maps and Earth

To celebrate the sunny days of summer (in the northern hemisphere at least), today we're launching new satellite imagery for Google’s mapping products. This stunning global view is virtually cloud-free and includes refreshed imagery in more locations—giving you an even more accurate and comprehensive view of our planet's landscape.

The new, even more beautiful global view in Maps and Earth.

Our satellite imagery is usually created like a quilt: we stitch together imagery of different parts of the world. Using a process similar to how we produced the global time-lapse imagery of the Earth, we took hundreds of terabytes of data from the USGS's and NASA’s Landsat 7 satellite—sometimes dozens of photos of a single spot in the world—and analyzed the photos to compute a clear view of every place, even in tropical regions that are always at least partly cloudy.

The result is a single, beautiful 800,000 megapixel image of the world, which can be viewed in Earth and Maps when you're zoomed out to a global view. This global image is so big, if you wanted to print it at a standard resolution of 300 dots per inch you’d need a piece of paper the size of a city block! This image is then blended into our highest resolution imagery, giving a beautiful cloud-free global view and detailed images in the same seamless map.

Central Papua, Indonesia: before and after.

This update also includes refreshed imagery in many regions of the world, especially in areas where high-resolution imagery is not available, including parts of Russia, Indonesia and central Africa.

Saudi Arabia: before and after, showing increased agricultural expansion

You can see the new satellite imagery by going to Google Maps and turning on satellite view, or by opening Google Earth, and zooming out. And to read more about what went into creating this imagery, check out our detailed post on the Lat Long blog. Have fun exploring!

Posted by Matt Hancher, Tech Lead, Google Earth Engine

Transparency Report: Making the web a safer place

Two of the biggest threats online are malicious software (known as malware) that can take control of your computer, and phishing scams that try to trick you into sharing passwords or other private information.

So in 2006 we started a Safe Browsing program to find and flag suspect websites. This means that when you are surfing the web, we can now warn you when a site is unsafe. We're currently flagging up to 10,000 sites a day—and because we share this technology with other browsers there are about 1 billion users we can help keep safe.

But we're always looking for new ways to protect users' security. So today we're launching a new section on our Transparency Report that will shed more light on the sources of malware and phishing attacks. You can now learn how many people see Safe Browsing warnings each week, where malicious sites are hosted around the world, how quickly websites become reinfected after their owners clean malware from their sites, and other tidbits we’ve surfaced.

Sharing this information also aligns well with our Transparency Report, which already gives information about government requests for user data, government requests to remove content, and current disruptions to our services.

To learn more, explore the new Safe Browsing information on this page. Webmasters and network administrators can find recommendations for dealing with malware infections, including resources like Google Webmaster Tools and Safe Browsing Alerts for Network Administrators.

Posted by Lucas Ballard, Software Engineer

From Sutton Hoo to the soccer pitch: culture with a click

Museums, libraries and galleries are a tourist staple of the summer holiday season. Often they’re the first place we head to when visiting a new city or town in order to learn about the heritage of that country. Though only a lucky few have the chance to travel to see these treasures first-hand, the Internet is helping to bring access to culture even when you can’t visit in person.

At the Google Cultural Institute, we’ve been busy working with our partners to add a range of new online exhibitions to our existing collection. With more than 6 million photos, videos and documents, the diversity and range of subject matter is large—a reflection of the fact that culture means different things to different people. What the exhibitions have in common is that they tell stories; objects are one thing but it’s the people and places they link to that make them fascinating.

The British Museum is the U.K.’s most popular visitor attraction and the 4th most visited museum in the world. It’s well known for housing one of the most spectacular archaeological discoveries ever made—the 1,400 year old Anglo-Saxon burial from Sutton Hoo, untouched until its discovery in 1939. Their online exhibition “Sutton Hoo: Anglo-Saxon ship burial” explores the discovery of the ship, featuring videos of the excavation and photos of the iconic helmet and a solid gold belt buckle. All this tells the story of how the burial and its contents changed our understanding of what Anglo-Saxon society was like.

From archaeology we take you to sport, which is integral to the culture of many nations, including Brazil. In the lead-up to Brazil's hosting of the 2014 World Cup, the Museu do Futebol has told the story of how the “beautiful game” came to Brazil. The photos, videos and posters in “The Game and the People” track the social impact of the sport and its transition from a past time for the wealthy (with their pleated pants and satin belts) to the modern game.

Science remains a perennially fascinating topic and the Museo Galileo in Italy has put together a series of three exhibitions looking at the link between art and science. The Medici Collections, the Lorraine Collections and the Library Collections examine the beginnings of science and technology 500 years ago and chart developments from the discovery of the sun dial to the Google Maps of today. As well as being informative, the exhibitions include beautiful objects such as the Jovilabe, which was used to calculate the periods of Jupiter’s moons.

So if broadening your cultural horizons through travel isn’t in the cards this summer, settle down in your armchair and browse through through some of the world’s heritage and history online. Keep up to date with new material on the Cultural Institute Google+ page.

Posted by James Davis, program manager, Google Cultural Institute

Experience stunning new heights with Street View in Dubai

What does it feel like to stand on top of the tallest building in the world? To give you a better sense of how that may feel, we took Street View to the Burj Khalifa in Dubai, our first-ever collection in the Arab World. Described as a “vertical city,” the Burj Khalifa is the world’s tallest manmade structure, towering over the Dubai skyline at 828 meters (2,717 ft).

This is the first time we’ve captured a skyscraper on Street View—making Google Maps even more comprehensive and useful for you. The imagery was collected over three days using the Street View Trekker and Trolley, capturing high-resolution 360-degree panoramic imagery of several indoor and outdoor locations of the building.

In addition to the breathtaking views from the world’s tallest observation deck on the 124th floor, you can also see what it feels like to hang off one of the building’s maintenance units on the 80th floor, normally used for cleaning windows!



Visit the highest occupied floor in the world on the 163rd floor, experience being in the fastest-moving elevators in the world (at 22 mph) and check out the highest swimming pool in the world on the 76th floor.

Even if you’re afraid of heights, we hope you enjoy the view from the top! To see highlights from the Burj Khalifa Street View collection, visit www.google.ae/streetview.

Posted by Tarek Abdalla, Head of Marketing - Middle East & North Africa, Google

“You’ve come a long way, Baby”: remembering the world’s first stored program computer

Sixty-five years ago today, the Manchester Small Scale Experimental Machine—nicknamed “Baby”—became the earliest computer in the world to run a program electronically stored in its memory. This was a flagship moment: the first implementation of the stored program concept that underpins modern computing.

Earlier computers had their instructions hardwired into their physical design or held externally on punched paper tape or cards. Reprogramming them to do a different task entailed internal rewiring or altering the physical storage media. The Baby marked a new computing era, described by some as the “birth of software,” in which swapping programs was far simpler—requiring only an update to the electronic memory. Both instructions and data were held in the Baby’s memory and the contents could be altered automatically at electronic speeds during the course of computation.



Developed at Manchester University by “Freddie” Williams, Tom Kilburn and Geoff Tootill, in size the Baby was anything but: more than 5m long and weighing a tonne (PDF). Its moniker was due to its role as a testbed for the experimental Williams-Kilburn tube, a means of storing binary digits (“bits”) using a cathode ray tube. This was a big deal because up until this point, computers had no cost-effective means of storing and flexibly accessing information in electronic form.

In technical terms, the Williams-Kilburn tube was the earliest form of random access memory, or RAM. The Baby’s memory consisted of one of these tubes, able to store up to 1,024 bits—equivalent to just 128 bytes. In contrast, the average computer today has RAM in multiples of gigabytes, more than a billion times bigger.

The Baby was only ever intended to be a proof-of-concept rather than to serve as a useful calculation tool. So once it had shown the new memory was reliable, attention shifted to building a more powerful and practical machine using the same concepts. This resulted in the Manchester Mark 1, which in turn was the model for the Ferranti Mark 1, the world’s first computer to be sold commercially, in February 1951.

While today nothing remains of the original Baby, a working replica is on display at the Museum of Science and Industry (MOSI) in Manchester. It’s well worth a visit to reflect on just how far computing has come.

Posted by Lynette Webb, Senior Manager, External Relations

America’s businesses are growing. The web is helping.

Michael Edlavitch was a middle school math teacher in Minnesota when he started a website with free math games to engage his students. With free online tools, a passion for math and an initial investment of just $10 to register his domain, www.hoodamath.com was born. Eventually Michael’s website became popular with more than just his students. So Michael gave Google AdSense a try as a way to earn money by placing ads next to his content. As word spread and traffic grew, the revenue generated from his site allowed Michael to devote himself full time to Hooda Math. Today, www.hoodamath.com has more than 350 educational games and has had more than 100 million unique visitors to the site. Beyond building a business for himself, Michael is helping students everywhere learn math while having fun.

Over in New York, Roberto Gil designs and builds children’s furniture—loft beds, bunk beds and entire custom rooms. Casa Kids’ furniture is custom designed for the family to grow along with the child. Roberto works out of his Brooklyn workshop and doesn’t sell to large furniture stores, which means the Casa Kids website is an essential tool for him to connect with potential customers. To grow even further, Roberto began using AdWords in 2010. In the first few months traffic to his site went up 30 percent. Today, two-thirds of his new customers come from Google. Meet Roberto and learn more about how he’s making the web work for Casa Kids:



These are just two examples of how the web is working for American businesses. According to a McKinsey study, small businesses that make use of the web are growing twice as fast as those that aren’t on the web. That’s because the web is where we go for information and inspiration—from math games to practice over the summer to someone to design and build that perfect bunk bed for your kids. Ninety-seven percent of American Internet users look online for local products and services. Whether we’re on our smartphones, tablets or computers, the web helps us find what we’re looking for.

Here at Google, we see firsthand how the web is helping American businesses grow and thrive. Through our search and advertising programs, businesses like Casa Kids find customers, publishers like Hooda Math earn money from their content, and nonprofits solicit donations and volunteers. These tools are how we make money, and they’re how millions of other U.S. businesses do, too.

In 2012, Google's search and advertising tools helped provide $94 billion of economic activity for more than 1.9 million American businesses—advertisers, publishers and nonprofits. This represents a 17 percent increase from 2011. Check out the impact made in each state, along with stories of local businesses using the web to grow.

Whether it’s building skills or building furniture, Google helps to build businesses. We’re thrilled to be part of such a vibrant industry and are committed to continuing to help make the web work for people and businesses everywhere.

Posted by Allan Thygesen, Vice President, Global SMB Sales

Google scholarships recognize 84 computer science scholars in Europe, Middle East, and Africa

We’d like to recognize and congratulate the 84 recipients and finalists of the Google Anita Borg Memorial Scholarship and Google Scholarship for Students with Disabilities in Europe, the Middle East and Africa. The full list of the 2013 scholars and finalists and the universities they attend can be found in this PDF.

Both scholarships aim to encourage underrepresented students to enter the computing field. The Google Anita Borg Memorial Scholarship honours the memory of Dr. Anita Borg who devoted her life to encouraging the presence of women in computing; we recently announced the U.S. recipients of this scholarship. The Google Europe Scholarship for Students with Disabilities aims to help dismantle barriers for students with disabilities as well as encourage them to excel in their studies and become active role models and leaders in creating technology.

All of the students receiving the scholarships are pursuing degrees in computer science or related fields at universities across Europe, the Middle East and Africa. This summer, they’ll attend the annual Google EMEA Scholarships Retreat in Zurich, where they’ll have the opportunity to attend tech talks on Google products, participate in developmental sessions, network with Googlers and attend social activities. Notable speakers at the 2013 retreat include Alan Eustace, SVP of Knowledge, Megan Smith, VP of Google [x], and Caroline Casey, Founder of Kanchi.org.

Applications for the scholarships will be open again in just a few short months. Learn more about how the scholarships impacted the lives of previous recipients:



For more information on all of our scholarships and programs, please visit the Google Students site.

Posted by Efrat Aghassy, EMEA scholarships program manager

Some Innovative MOOCs

Posted by Maggie Johnson, Director of Education and University Relations

Last summer, we jumped into the world of MOOCs (Massive Open Online Courses) with our own course on search skills, Power Searching. Soon after, we open sourced the platform that we developed to present the course -- Course Builder. A large number of courses have been hosted on Course Builder since, with many more coming soon. As can be seen from the list of courses, our goal is to provide the capability for anyone to create a MOOC. We’ve been pleasantly surprised by the variety of courses and the creativity of the instructors building on Course Builder.

For example, GivingWithPurpose is an innovative MOOC presented by Learning By Giving, one of Doris Buffett’s foundations for “giving it all away.” Instructor Rebecca Riccio, who teaches philanthropy at both Northeastern and Brandeis Universities, feels that reaching thousands of people in a discussion about how to allocate scarce resources to address the needs of our communities has huge potential. “For all the social, cultural, and economic value we derive from the nonprofit sector, we do shockingly little to educate people about why it is so important and what we can do to help it thrive. So while I believe GivingWithPurpose will be successful in its primary goal of teaching students how to give more effectively, in ways that both satisfy their own motivations for giving and support high-performing nonprofit organizations, my second, perhaps more ambitious goal is to promote more informed civic engagement.”

We’ve also hosted MOOCs on evaluating and selecting soccer players, how to search for a job, and how to develop digital learning opportunities for students in public schools. We have many university courses such as Game Theory from Stanford and Information Visualization from Indiana University.

Course Builder’s support of both traditional and non-traditional education opportunities is core to its mission. We’ll continue to build features that help university professors, K12 teachers and anyone else who has something important to teach.

Celebrating 10 years of shared success

Ten years ago we launched AdSense to help publishers earn money by placing relevant ads on their websites. I can still remember the excitement and anticipation as AdSense went live that first day. Our small team huddled together in a cramped conference room, and right away we saw that publishers were as excited about AdSense as we were.

Fast-forward 10 years, and AdSense has become a core part of Google’s advertising business. The AdSense community has grown to include more than 2 million publishers, and last year alone, publishers earned more than $7 billion from AdSense. AdSense is a community that thrives because of all the content creators we are so fortunate to partner with. Their stories inspire us to do our part to make AdSense great.

On this occasion, it’s especially inspiring to hear the stories of partners who have been with us since the very beginning—like a retiree in New Zealand who was able to pursue her dream of writing about her garden, a tech support expert in Colorado who can spend more time with his kids, and a theme park reviewer who now sends employees around the world to test and review rides—all thanks to money earned from AdSense.

As part of our 10th anniversary celebration, we hope you’ll tune into our live Hangout on Air today at 10 a.m. PDT (5 p.m. GMT) on the AdSense Google+ page. I look forward to joining several of our partners to share stories from the early days of AdSense, talk about how we’ve all grown since then, and discuss the future for publishers and online advertising. And if you want even more 10th anniversary celebration, just visit our AdSense 10th anniversary page at any time.

Posted by Susan Wojcicki, SVP, Ads and Commerce

Chromebooks: coming to more stores near you

In Northern California where I live, summer is here, which means family vacations, kids’ camps, BBQs and hopefully some relaxation. But it also means back-to-school shopping is just around the corner. So in case you’re on the hunt for a laptop in addition to pens, paper, and stylish new outfits, your search just got a whole lot easier. Chromebooks—a fast, simple, secure laptop that won't break the bank—will now be carried in over 3 times more stores than before, or more than 6,600 stores around the world.

In addition to Best Buy and Amazon.com, we’re excited to welcome several new retailers to the family. Starting today, Walmart will be making the newest Acer Chromebook, which has a 16GB Solid State Drive (SSD), available in approximately 2,800 stores across the U.S., for just $199. Look for Chromebooks coming to the laptop sections of a Walmart near you this summer.

And beginning this weekend, Staples will bring a mix of Chromebooks from Acer, HP and Samsung to every store in the U.S.—more than 1,500 in total. You can also purchase via Staples online, while businesses can purchase through the Staples Advantage B2B program. In the coming months select Office Depot, OfficeMax, and regional chains Fry’s and TigerDirect locations will begin selling Chromebooks.

In the 10 other markets worldwide where Chromebooks are sold, availability in national retailers continues to expand. In addition to Dixons in the UK, now 116 Tesco stores are selling Chromebooks, as well as all Media Markt and Saturn stores in the Netherlands, FNAC stores in France and Elgiganten stores in Sweden. In Australia, all JB Hi-Fi and Harvey Norman stores will be carrying Chromebooks for their customers as well. We’re working hard to bring Chromebooks to even more countries later this year.

Chromebooks make great computers for everyone in the family—and now you shouldn’t have to look very far to find one. Happy summer!

Posted by David Shapiro, Director of Chromebook Marketing

Happy Small Business Week.

Our first AdWords customer was a small business selling live mail-order lobsters. It’s been a long time since then, but a majority of our customers are still small businesses, who play a vital role not only for Google, but for the American economy. More than 60 percent of new jobs each year come from small businesses.



This Small Business Week, we want to celebrate you. We’re grateful to you for everything you do for us and our communities. Whether you fix people’s cars, offer music lessons to aspiring musicians, or make the world’s best homemade ice cream—when you do what you love, our lives get better.

As part of the celebration, we’ll be highlighting some amazing small businesses across the country, so keep an eye on the Google+ Your Business page. And in the meantime, check out some of the Google tools that are designed to help you take care of business.

Happy Small Business Week.

Posted by Lisa Gevelber, VP Marketing, Americas

Our continued commitment to combating child exploitation online

The Internet has been a tremendous force for good—increasing access to information, improving people’s ability to communicate and driving economic growth. But like the physical world, there are dark corners on the web where criminal behavior exists.

In 2011, the National Center for Missing & Exploited Children’s (NCMEC’s) Cybertipline Child Victim Identification Program reviewed 17.3 million images and videos of suspected child sexual abuse. This is four times more than what their Exploited Children's Division (ECD) saw in 2007. And the number is still growing. Behind these images are real, vulnerable kids who are sexually victimized and victimized further through the distribution of their images.

It is critical that we take action as a community—as concerned parents, guardians, teachers and companies—to help combat this problem.

Child sexual exploitation is a global problem that needs a global solution. More than half of the images and videos sent to NCMEC for analysis are found to have been uploaded to U.S. servers from outside the country. With this in mind, we need to sustain and encourage borderless communication between organizations fighting this problem on the ground. For example, NCMEC’s CyberTipline is able to refer reports regarding online child sexual exploitation to 66 countries, helping local law enforcement agencies effectively execute their investigations.

Google has been working on fighting child exploitation since as early as 2006 when we joined the Technology Coalition, teaming up with other tech industry companies to develop technical solutions. Since then, we’ve been providing software and hardware to helping organizations all around the world to fight child abuse images on the web and help locate missing children.

There is much more that can be done, and Google is taking our commitment another step further through a $5 million effort to eradicate child abuse imagery online. Part of this commitment will go to global child protection partners like the National Center for Missing & Exploited Children and the Internet Watch Foundation. We’re providing additional support to similar heroic organizations in the U.S., Canada, Europe, Australia and Latin America.

Since 2008, we’ve used “hashing” technology to tag known child sexual abuse images, allowing us to identify duplicate images which may exist elsewhere. Each offending image in effect gets a unique ID that our computers can recognize without humans having to view them again. Recently, we’ve started working to incorporate encrypted “fingerprints” of child sexual abuse images into a cross-industry database. This will enable companies, law enforcement and charities to better collaborate on detecting and removing these images, and to take action against the criminals. Today we’ve also announced a $2 million Child Protection Technology Fund to encourage the development of ever more effective tools.

We’re in the business of making information widely available, but there’s certain “information” that should never be created or found. We can do a lot to ensure it’s not available online—and that when people try to share this disgusting content they are caught and prosecuted.

Update June 17: Clarified language around NCMEC's Child Victim Identification Program and CyberTipline.

Posted by Jacquelline Fuller, Director, Google Giving

Introducing Project Loon: Balloon-powered Internet access

The Internet is one of the most transformative technologies of our lifetimes. But for 2 out of every 3 people on earth, a fast, affordable Internet connection is still out of reach. And this is far from being a solved problem.

There are many terrestrial challenges to Internet connectivity—jungles, archipelagos, mountains. There are also major cost challenges. Right now, for example, in most of the countries in the southern hemisphere, the cost of an Internet connection is more than a month’s income.

Solving these problems isn’t simply a question of time: it requires looking at the problem of access from new angles. So today we’re unveiling our latest moonshot from Google[x]: balloon-powered Internet access.

We believe that it might actually be possible to build a ring of balloons, flying around the globe on the stratospheric winds, that provides Internet access to the earth below. It’s very early days, but we’ve built a system that uses balloons, carried by the wind at altitudes twice as high as commercial planes, to beam Internet access to the ground at speeds similar to today’s 3G networks or faster. As a result, we hope balloons could become an option for connecting rural, remote, and underserved areas, and for helping with communications after natural disasters. The idea may sound a bit crazy—and that’s part of the reason we’re calling it Project Loon—but there’s solid science behind it.

Balloons, with all their effortless elegance, present some challenges. Many projects have looked at high-altitude platforms to provide Internet access to fixed areas on the ground, but trying to stay in one place like this requires a system with major cost and complexity. So the idea we pursued was based on freeing the balloons and letting them sail freely on the winds. All we had to do was figure out how to control their path through the sky. We’ve now found a way to do that, using just wind and solar power: we can move the balloons up or down to catch the winds we want them to travel in. That solution then led us to a new problem: how to manage a fleet of balloons sailing around the world so that each balloon is in the area you want it right when you need it. We’re solving this with some complex algorithms and lots of computing power.

Now we need some help—this experiment is going to take way more than our team alone. This week we started a pilot program in the Canterbury area of New Zealand with 50 testers trying to connect to our balloons. This is the first time we’ve launched this many balloons (30 this week, in fact) and tried to connect to this many receivers on the ground, and we’re going to learn a lot that will help us improve our technology and balloon design.

Over time, we’d like to set up pilots in countries at the same latitude as New Zealand. We also want to find partners for the next phase of our project—we can’t wait to hear feedback and ideas from people who’ve been working for far longer than we have on this enormous problem of providing Internet access to rural and remote areas. We imagine someday you'll be able to use your cell phone with your existing service provider to connect to the balloons and get connectivity where there is none today.

This is still highly experimental technology and we have a long way to go—we’d love your support as we keep trying and keep flying! Follow our Google+ page to keep up with Project Loon’s progress.

Onward and upward.

Posted by Mike Cassidy, Project Lead

Excellent Papers for 2012

Posted by Corinna Cortes and Alfred Spector, Google Research

Googlers across the company actively engage with the scientific community by publishing technical papers, contributing open-source packages, working on standards, introducing new APIs and tools, giving talks and presentations, participating in ongoing technical debates, and much more. Our publications offer technical and algorithmic advances, feature aspects we learn as we develop novel products and services, and shed light on some of the technical challenges we face at Google.

In an effort to highlight some of our work, we periodically select a number of publications to be featured on this blog. We first posted a set of papers on this blog in mid-2010 and subsequently discussed them in more detail in the following blog postings. In a second round, we highlighted new noteworthy papers from the later half of 2010 and again in 2011. This time we honor the influential papers authored or co-authored by Googlers covering all of 2012 -- covering roughly 6% of our total publications.  It’s tough choosing, so we may have left out some important papers.  So, do see the publications list to review the complete group.

In the coming weeks we will be offering a more in-depth look at some of these publications, but here are the summaries:

Algorithms and Theory

Online Matching with Stochastic Rewards
Aranyak Mehta*, Debmalya Panigrahi [FOCS'12]
Online advertising is inherently stochastic: value is realized only if the user clicks on the ad, while the ad platform knows only the probability of the click. This paper is the first to introduce the stochastic nature of the rewards to the rich algorithmic field of online allocations. The core algorithmic problem it formulates is online bipartite matching with stochastic rewards, with known click probabilities. The main result is an online algorithm which obtains a large fraction of the optimal value. The paper also shows the difficulty introduced by the stochastic nature, by showing how it behaves very differently from the classic (non-stochastic) online matching problem.

Matching with our Eyes Closed
Gagan Goel*, Pushkar Tripathi* [FOCS'12]
In this paper we propose a simple randomized algorithm for finding a matching in a large graph. Unlike most solutions to this problem, our approach does not rely on building large combinatorial structures (like blossoms) but works completely locally. We analyze the performance of our algorithm and show that it does significantly better than the greedy algorithm. In doing so we improve a celebrated 18 year old result by Aronson et. al.

Simultaneous Approximations for Adversarial and Stochastic Online Budgeted Allocation
Vahab Mirrokni*, Shayan Oveis Gharan, Morteza Zadimoghaddam, [SODA'12]
In this paper, we study online algorithms that simultaneously perform well in worst-case and average-case instances, or equivalently algorithms that perform well in both stochastic and adversarial models at the same time. This is motivated by online allocation of queries to advertisers with budget constraints. Stochastic models are not robust enough to deal with traffic spikes and adversarial models are too pessimistic. While several algorithms have been proposed for these problems, each algorithm was known to perform well in one model and not both, and we present new results for a single algorithm that works well in both models.

Economics and EC

Polyhedral Clinching Auctions and the Adwords Polytope
Gagan Goel*, Vahab Mirrokni*, Renato Paes Leme [STOC'12]
Budgets play a major role in ad auctions where advertisers explicitly declare budget constraints. Very little is known in auctions about satisfying such budget constraints while keeping incentive compatibility and efficiency. The problem becomes even harder in the presence of complex combinatorial constraints over the set of feasible allocations. We present a class of ascending-price auctions addressing this problem for a very general class of (polymatroid) allocation constraints including the AdWords problem with multiple keywords and multiple slots.

HCI

Backtracking Events as Indicators of Usability Problems in Creation-Oriented Applications
David Akers*, Robin Jeffries*, Matthew Simpson*, Terry Winograd [TOCHI '12]
Backtracking events such as undo can be useful automatic indicators of usability problems for creation-oriented applications such as word processors and photo editors. Our paper presents a new cost-effective usability evaluation method based on this insight.

Talking in Circles: Selective Sharing in Google+
Sanjay Kairam, Michael J. Brzozowski*, David Huffaker*, Ed H. Chi*, [CHI'12]
This paper explores why so many people share selectively on Google+: to protect their privacy but also to focus and target their audience. People use Circles to support these goals, organizing contacts by life facet, tie strength, and interest.

Information Retrieval

Online selection of diverse results
Debmalya Panigrahi, Atish Das Sarma, Gagan Aggarwal*, and Andrew Tomkins*, [WSDM'12]
We consider the problem of selecting subsets of items that are simultaneously diverse in multiple dimensions, which arises in the context of recommending interesting content to users. We formally model this optimization problem, identify its key structural characteristics, and use these observations to design an extremely scalable and efficient algorithm. We prove that the algorithm always produces a nearly optimal solution and also perform experiments on real-world data that indicate that the algorithm performs even better in practice than the analytical guarantees.

Machine Learning

Large Scale Distributed Deep Networks
Jeffrey Dean, Greg S. Corrado*, Rajat Monga, Kai Chen, Matthieu Devin, Quoc V. Le, Mark Z. Mao, Marc’Aurelio Ranzato, Andrew Senior, Paul Tucker, Ke Yang, Andrew Y. Ng, NIPS 2012;
In this paper, we examine several techniques to improve the time to convergence for neural networks and other models trained by gradient-based methods. The paper describes a system we have built that exploits both model-level parallelism (by partitioning the nodes of a large model across multiple machines) and data-level parallelism (by having multiple replicas of a model process different training data and coordinating the application of updates to the model state through a centralized-but-partitioned parameter server system). Our results show that very large neural networks can be trained effectively and quickly on large clusters of machines.

Open Problem: Better Bounds for Online Logistic Regression
Brendan McMahan* and Matthew Streeter*, COLT/ICML'12 Joint Open Problem Session, JMLR: Workshop and Conference Proceedings.
One of the goals of research at Google is to help point out important open problems--precise questions that are interesting academically but also have important practical ramifications. This open problem is about logistic regression, a widely used algorithm for predicting probabilities (what is the probability an email message is spam, or that a search ad will be clicked). We show that in the simple one-dimensional case, much better results are possible than current theoretical analysis suggests, and we ask whether our results can be generalized to arbitrary logistic regression problems.

Spectral Learning of General Weighted Automata via Constrained Matrix Completion
Borja Balle and Mehryar Mohri*, NIPS 2012.
Learning weighted automata from finite samples drawn from an unknown distribution is a central problem in machine learning and computer science in general, with a variety of applications in text and speech processing, bioinformatics, and other areas. This paper presents a new family of algorithms for tackling this problem for which it proves learning guarantees. The algorithms introduced combine ideas from two different domains: matrix completion and spectral methods.

Machine Translation

Improved Domain Adaptation for Statistical Machine Translation
Wei Wang*, Klaus Macherey*, Wolfgang Macherey*, Franz Och* and Peng Xu*, [AMTA'12]
Research in domain adaptation for machine translation has been mostly focusing on one domain. We present a simple and effective domain adaptation infrastructure that makes an MT system with a single translation model capable of providing adapted, close-to-upper-bound domain-specific accuracy while preserving the generic translation accuracy. Large-scale experiments on 20 language pairs for patent and generic domains show the viability of our approach.

Multimedia and Computer Vision

Reconstructing the World's Museums
Jianxiong Xiao and Yasutaka Furukawa*, [ECCV '12]
Virtual navigation and exploration of large indoor environments (e.g., museums) have been so far limited to either blueprint-style 2D maps that lack photo-realistic views of scenes, or ground-level image-to-image transitions, which are immersive but ill-suited for navigation. This paper presents a novel vision-based 3D reconstruction and visualization system to automatically produce clean and well-regularized texture-mapped 3D models for large indoor scenes, from ground-level photographs and 3D laser points. For the first time, we enable users to easily browse a large scale indoor environment from a bird's-eye view, locate specific room interiors, fly into a place of interest, view immersive ground-level panoramas, and zoom out again, all with seamless 3D transitions.

The intervalgram: An audio feature for large-scale melody recognition
Thomas C. Walters*, David Ross*, Richard F. Lyon*, [CMMR'12]
Intervalgrams are small images that summarize the structure of short segments of music by looking at the musical intervals between the notes present in the music. We use them for finding cover songs - different pieces of music that share the same underlying composition. Wedo this by comparing 'heatmaps' which look at the similarity between intervalgrams from different pieces of music over time. If we see a strong diagonal line in the heatmap, it's good evidence that the songs are musically similar.

General and Nested Wiberg Minimization
Dennis Strelow*, [CVPR'12]
Eriksson and van den Hengel’s CVPR 2010 paper showed that Wiberg’s least squares matrix factorization, which effectively eliminates one matrix from the factorization problem, could be applied to the harder case of L1 factorization. Our paper generalizes their approach beyond factorization to general nonlinear problems in two sets of variables, like perspective structure-from-motion. We also show that with our generalized method, one Wiberg minimization can also be nested inside another, effectively eliminating two of three sets of unknowns, and we demonstrated this idea using projective struture-from-motion

Calibration-Free Rolling Shutter Removal
Matthias Grundmann*, Vivek Kwatra*, Daniel Castro, Irfan Essa*, International Conference on Computational Photography '12. Best paper.
Mobile phones and current generation DSLR’s, contain an electronic rolling shutter, capturing each frame one row of pixels at a time. Consequently, if the camera moves during capture, it will cause image distortions ranging from shear to wobbly distortions. We propose a calibration-free solution based on a novel parametric mixture model to correct these rolling shutter distortions in videos that enables real-time rolling shutter rectification as part of YouTube’s video stabilizer.

Natural Language Processing

Vine Pruning for Efficient Multi-Pass Dependency Parsing
Alexander Rush, Slav Petrov*, The 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL '12), Best Paper Award.
Being able to accurately analyze the grammatical structure of sentences is crucial for language understanding applications such as machine translation or question answering. In this paper we present a method that is up to 200 times faster than existing methods and enables the grammatical analysis of text in large-scale applications. The key idea is to perform the analysis in multiple coarse-to-fine passes, resolving easy ambiguities first and tackling the harder ones later on.

Cross-lingual Word Clusters for Direct Transfer of Linguistic Structure
Oscar Tackstrom, Ryan McDonald*, Jakob Uszkoreit*, North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL '12), Best Student Paper Award.
This paper studies how to build meaningful cross-lingual word clusters, i.e., clusters containing lexical items from two languages that are coherent along some abstract dimension. This is done by coupling distributional statistics learned from huge amounts of language specific data coupled with constraints generated from parallel corpora. The resulting clusters are used to improve the accuracy of multi-lingual syntactic parsing for languages without any training resources.

Networks

How to Split a Flow
Tzvika Hartman*, Avinatan Hassidim*, Haim Kaplan*, Danny Raz*, Michal Segalov*, [INFOCOM '12]
Decomposing a flow into a small number of paths is a very important task arises in various network optimization mechanisms. In this paper we develop an an approximation algorithm for this problem that has both provable worst case performance grantees as well as good practical behavior.

Deadline-Aware Datacenter TCP (D2TCP)
Balajee Vamanan, Jahangir Hasan*, T. N. Vijaykumar, [SIGCOMM '12]
Some of our most important products like search and ads operate under soft-real-time constraints. They are architected and fine-tuned to return results to users within a few hundred milliseconds. Deadline-Aware Datacenter TCP is a research effort into making the datacenter networks deadline aware, thus improving the performance of such key applications.

Trickle: Rate Limiting YouTube Video Streaming
Monia Ghobadi, Yuchung Cheng*, Ankur Jain*, Matt Mathis* [USENIX '12]
Trickle is a server-side mechanism to stream YouTube video smoothly to reduce burst and buffer-bloat. It paces the video stream by placing an upper bound on TCP’s congestion window based on the streaming rate and the round-trip time. In initial evaluation Trickle reduces the TCP loss rate by up to 43% and the RTT by up to 28%. Given the promising results we are deploying Trickle to all YouTube servers.

Social Systems

Look Who I Found: Understanding the Effects of Sharing Curated Friend Groups
Lujun Fang*, Alex Fabrikant*, Kristen LeFevre*, [Web Science '12]. Best Student Paper award.
In this paper, we studied the impact of the Google+ circle-sharing feature, which allows individual users to share (publicly and privately) pre-curated groups of friends and contacts. We specifically investigated the impact on the growth and structure of the Google+ social network. In the course of the analysis, we identified two natural categories of shared circles ("communities" and "celebrities"). We also observed that the circle-sharing feature is associated with the accelerated densification of community-type circles.

Software Engineering

AddressSanitizer: A Fast Address Sanity Checker
Konstantin Serebryany*, Derek Bruening*, Alexander Potapenko*, Dmitry Vyukov*, [USENIX ATC '12].
The paper “AddressSanitizer: A Fast Address Sanity Checker” describes a dynamic tool that finds memory corruption bugs in C or C++ programs with only a 2x slowdown. The major feature of AddressSanitizer is simplicity -- this is why the tool is very fast.

Speech

Japanese and Korean Voice Search
Mike Schuster*, Kaisuke Nakajima*, IEEE International Conference on Acoustics, Speech, and Signal Processing [ICASSP'12].
"Japanese and Korean voice search" explains in detail how the Android voice search systems for these difficult languages were developed. We describe how to segment statistically to be able to handle infinite vocabularies without out-of-vocabulary words, how to handle the lack of spaces between words for language modeling and dictionary generation, and how to deal best with multiple ambiguities during evaluation scoring of reference transcriptions against hypotheses. The combination of techniques presented led to high quality speech recognition systems--as of 6/2013 Japanese and Korean are #2 and #3 in terms of traffic after the US.

Google's Cross-Dialect Arabic Voice Search
Fadi Biadsy*, Pedro J. Moreno*, Martin Jansche*, IEEE International Conference on Acoustics, Speech, and Signal Processing [ICASSP 2012].
This paper describes Google’s automatic speech recognition systems for recognizing several Arabic dialects spoken in the Middle East, with the potential to reach more than 125 million users. We suggest solutions for challenges specific to Arabic, such as the diacritization problem, where short vowels are not written in Arabic text. We conduct experiments to identify the optimal manner in which acoustic data should be clustered among dialects.

Deep Neural Networks for Acoustic Modeling in Speech Recognition
Geoffrey Hinton*, Li Deng, Dong Yu, George Dahl, Abdel-rahman Mohamed, Navdeep Jaitly, Andrew W. Senior*, Vincent Vanhoucke*, Patrick Nguyen, Tara Sainath, Brian Kingsbury, Signal Processing Magazine (2012)"
Survey paper on the DNN breakthrough in automatic speech recognition accuracy.

Statistics

Empowering Online Advertisements by Empowering Viewers with the Right to Choose
Max Pashkevich*, Sundar Dorai-Raj*, Melanie Kellar*, Dan Zigmond*, Journal of Advertising Research, vol. 52 (2012).
YouTube’s TrueView in-stream video advertising format (a form of skippable in-stream ads) can improve the online video viewing experience for users without sacrificing advertising value for advertisers or content owners.

Structured Data

Efficient Spatial Sampling of Large Geographical Tables
Anish Das Sarma*, Hongrae Lee*, Hector Gonzalez*, Jayant Madhavan*, Alon Halevy*, [SIGMOD '12].
This paper presents fundamental results for the "thinning problem": determining appropriate samples of data to be shown on specific geographical regions and zoom levels. This problem is widely applicable for a number of cloud-based geographic visualization systems such as Google Maps, Fusion Tables, and the developed algorithms are part of the Fusion Tables backend. The SIGMOD 2012 paper was selected among the best papers of the conference, and invited to a special best-papers issue of TODS.

Systems

Spanner: Google's Globally-Distributed Database
James C. Corbett*, Jeffrey Dean*, Michael Epstein*, Andrew Fikes*, Christopher Frost*, JJ Furman*, Sanjay Ghemawat*, Andrey Gubarev*, Christopher Heiser*, Peter Hochschild*, Wilson Hsieh*, Sebastian Kanthak*, Eugene Kogan*, Hongyi Li*, Alexander Lloyd*, Sergey Melnik*, David Mwaura*, David Nagle*, Sean Quinlan*, Rajesh Rao*, Lindsay Rolig*, Dale Woodford*, Yasushi Saito*, Christopher Taylor*, Michal Szymaniak*, Ruth Wang*, [OSDI '12]
This paper shows how a new time API and its implementation can provide the abstraction of tightly synchronized clocks, even on a global scale. We describe how we used this technology to build a globally-distributed database that supports a variety of powerful features: non-blocking reads in the past, lock-free snapshot transactions, and atomic schema changes.

Getting healthy just got a little easier

We’re all looking for ways to get a little healthier and smarter about the choices we make. Having tools and information at your fingertips might help bring a bit of motivation to your routine, and of course good tunes and a strong community doesn’t hurt either.

What’s in that cupcake?
Want to know how many calories are in a cupcake, or how much potassium is in a banana? You can now find nutrition information for over 1,000 foods in search - helping you stay informed about what you eat more quickly and easily. While using voice search, on desktop, your iPhone, or Android device you can ask, “how many calories are in a cupcake?” and you can follow-up and ask, “how about a cookie?” without needing to repeat parts of your question. Fruits and vegetables don’t have labels, and it’s often hard to track down the nutritional info for wine or more complex dishes like a burrito, so type or tap the microphone and easily ask your question for these foods and more.

Explore what’s around you, on two wheels
If you want a change of scenery from the gym, use Google Maps on your Android device to find nearby biking routes. Mount your device on your handlebars to see the turn-by-turn directions and navigation, or use speaker-mode to hear voice-guided directions for more than 330,000 miles of trails and paths around the world. Dark green lines on the map show dedicated bike trails and paths without cars, light green lines show streets with dedicated bike lanes, and dashed green lines show other streets recommended for cycling.

Team up to get fit
Looking to get healthy with a friend? Join a Google+ Community and connect with others that share your diet and exercise goals. Check out Communities such as Eating Right and Fitness & Weight Loss for motivation, tips and inspiration to keep you on track. Use Hangouts On Air to learn what experts like The Biggest Loser are saying about nutrition or jump into a yoga class.

Don’t stop the music
A good beat will keep you moving and motivated. Sign up for All Access, our new music subscription service, and you can listen to millions of songs from Google Play Music. Build an awesome workout mix or start a radio station from your favorite pop song like “We Can’t Stop!” Miley Cyrus says it best.

Keep track—no matter which device you’re on
Counting calories? Apps such as Diet Diary can be easily accessed through Chrome or on your mobile device—that way it’s with you when it‘s on your mind. If spreadsheets are more your style, try one of several Google Docs templates, like this weekly meal planner.

Get inspired by the pros
Need a little more motivation? Why not watch fitness gurus do their thing on YouTube: you can watch Sadie Nardini and her amazing yoga classes, or Cassey Ho will get you in top shape for summer - all in the comfort of your own living room.

Posted by Roya Soleimani, Google Search team

Play Cube Slam face-to-face against your friends

My friends and I used to play videogames all the time, squashed together on the couch, engaged in structured intellectual discourse about exactly how badly we were going to destroy each other. Now that we live spread out around the world, it’s a bit harder to dance in each other’s faces and yell “booyah!” every time we win a game. Enter: Cube Slam.

Cube Slam is a video game that you can play face-to-face against your friends. It’s a Chrome Experiment built using WebRTC, an open web technology that lets you video chat right in the browser without installing any plug-ins. That means you can quickly and easily play Cube Slam with your friends, no matter where they are in the world, just by sharing a link.

To win Cube Slam, hit the cube against your friend’s screen three times until the screen explodes. Shields, obstacles, and gravity fields change with every new level, and you can unlock power-ups including fireballs, lasers, multi-balls, mirrored controls, bulletproof shields, fog, ghost balls, time bombs, resized paddles, extra lives and death balls––though you might want to avoid the death balls. If none of your friends are online, you can always play against Bob the Bear and see what level you can reach. If you install the Cube Slam app, you can even play Bob when you’re offline.

Cube Slam’s graphics are rendered in WebGL and CSS 3D, and its custom soundtrack is delivered dynamically through Web Audio. WebRTC, which enables the two-person game, is available on desktop Chrome and Chrome OS, and will be available on mobile later this year. In the meantime, you can play Cube Slam against Bob the Bear on your phone or tablet. To learn more about what’s going on under the hood, see our technology page and Chromium blog post.

Play a friend. Play a bear. Have fun!

Posted by Clem Wright, Google Creative Lab, Ursine Diversion Division

Improving Photo Search: A Step Across the Semantic Gap

Posted by Chuck Rosenberg, Image Search Team

Last month at Google I/O, we showed a major upgrade to the photos experience: you can now easily search your own photos without having to manually label each and every one of them. This is powered by computer vision and machine learning technology, which uses the visual content of an image to generate searchable tags for photos combined with other sources like text tags and EXIF metadata to enable search across thousands of concepts like a flower, food, car, jet ski, or turtle.

For many years Google has offered Image Search over web images; however, searching across photos represents a difficult new challenge. In Image Search there are many pieces of information which can be used for ranking images, for example text from the web or the image filename. However, in the case of photos, there is typically little or no information beyond the pixels in the images themselves. This makes it harder for a computer to identify and categorize what is in a photo. There are some things a computer can do well, like recognize rigid objects and handwritten digits. For other classes of objects, this is a daunting task, because the average toddler is better at understanding what is in a photo than the world’s most powerful computers running state of the art algorithms.

This past October the state of the art seemed to move things a bit closer to toddler performance. A system which used deep learning and convolutional neural networks easily beat out more traditional approaches in the ImageNet computer vision competition designed to test image understanding. The winning team was from Professor Geoffrey Hinton’s group at the University of Toronto.

We built and trained models similar to those from the winning team using software infrastructure for training large-scale neural networks developed at Google in a group started by Jeff Dean and Andrew Ng. When we evaluated these models, we were impressed; on our test set we saw double the average precision when compared to other approaches we had tried. We knew we had found what we needed to make photo searching easier for people using Google. We acquired the rights to the technology and went full speed ahead adapting it to run at large scale on Google’s computers. We took cutting edge research straight out of an academic research lab and launched it, in just a little over six months. You can try it out at photos.google.com.

Why the success now? What is new? Some things are unchanged: we still use convolutional neural networks -- originally developed in the late 1990s by Professor Yann LeCun in the context of software for reading handwritten letters and digits. What is different is that both computers and algorithms have improved significantly. First, bigger and faster computers have made it feasible to train larger neural networks with much larger data. Ten years ago, running neural networks of this complexity would have been a momentous task even on a single image -- now we are able to run them on billions of images. Second, new training techniques have made it possible to train the large deep neural networks necessary for successful image recognition.

We feel it would be interesting to the research community to discuss some of the unique aspects of the system we built and some qualitative observations we had while testing the system.

The first is our label and training set and how it compares to that used in the ImageNet Large Scale Visual Recognition competition. Since we were working on search across photos, we needed an appropriate label set. We came up with a set of about 2000 visual classes based on the most popular labels on Google+ Photos and which also seemed to have a visual component, that a human could recognize visually. In contrast, the ImageNet competition has 1000 classes. As in ImageNet, the classes were not text strings, but are entities, in our case we use Freebase entities which form the basis of the Knowledge Graph used in Google search. An entity is a way to uniquely identify something in a language-independent way. In English when we encounter the word “jaguar”, it is hard to determine if it represents the animal or the car manufacturer. Entities assign a unique ID to each, removing that ambiguity, in this case “/m/0449p” for the former and “/m/012x34” for the latter. In order to train better classifiers we used more training images per class than ImageNet, 5000 versus 1000. Since we wanted to provide only high precision labels, we also refined the classes from our initial set of 2000 to the most precise 1100 classes for our launch.

During our development process we had many more qualitative observations we felt are worth mentioning:

1) Generalization performance. Even though there was a significant difference in visual appearance between the training and test sets, the network appeared to generalize quite well. To train the system, we used images mined from the web which did not match the typical appearance of personal photos. Images on the web are often used to illustrate a single concept and are carefully composed, so an image of a flower might only be a close up of a single flower. But personal photos are unstaged and impromptu, a photo of a flower might contain many other things in it and may not be very carefully composed. So our training set image distribution was not necessarily a good match for the distribution of images we wanted to run the system on, as the examples below illustrate. However, we found that our system trained on web images was able to generalize and perform well on photos.

A typical photo of a flower found on the web.

A typical photo of a flower found in an impromptu photo.

2) Handling of classes with multi-modal appearance. The network seemed to be able to handle classes with multimodal appearance quite well, for example the “car” class contains both exterior and interior views of the car. This was surprising because the final layer is effectively a linear classifier which creates a single dividing plane in a high dimensional space. Since it is a single plane, this type of classifier is often not very good at representing multiple very different concepts.

3) Handling abstract and generic visual concepts. The system was able to do reasonably well on classes that one would think are somewhat abstract and generic. These include "dance", "kiss", and "meal", to name a few. This was interesting because for each of these classes it did not seem that there would be any simple visual clues in the image that would make it easy to recognize this class. It would be difficult to describe them in terms of simple basic visual features like color, texture, and shape.

Photos recognized as containing a meal.

4) Reasonable errors. Unlike other systems we experimented with, the errors which we observed often seemed quite reasonable to people. The mistakes were the type that a person might make - confusing things that look similar. Some people have already noticed this, for example, mistaking a goat for a dog or a millipede for a snake. This is in contrast to other systems which often make errors which seem nonsensical to people, like mistaking a tree for a dog.

Photo of a banana slug mistaken for a snake.

Photo of a donkey mistaken for a dog.

5) Handling very specific visual classes. Some of the classes we have are very specific, like specific types of flowers, for example “hibiscus” or “dhalia”. We were surprised that the system could do well on those. To recognize specific subclasses very fine detail is often needed to differentiate between the classes. So it was surprising that a system that could do well on a full image concept like “sunsets” could also do well on very specific classes.

Photo recognized as containing a hibiscus flower.

Photo recognized as containing a dahlia flower.

Photo recognized as containing a polar bear.

Photo recognized as containing a grizzly bear.

The resulting computer vision system worked well enough to launch to people as a useful tool to help improve personal photo search, which was a big step forward. So, is computer vision solved? Not by a long shot. Have we gotten computers to see the world as well as people do? The answer is not yet, there’s still a lot of work to do, but we’re closer.