Celebrating Twenty Years of Writing the Future

Recently, I was in Beijing to celebrate the 20th anniversary of Microsoft Research Asia (MSRA) along with Dr. Hsiao-Wuen Hon, who has been the lab’s director for the past decade. Spending time with this distinguished group of scientists, researchers and engineers made me think a lot about the beginnings of MSRA – and even MSR – and how far we have come since we first opened the doors in 1998.  

In 20 years, we have published over 5,000 research papers and created one of the most recognized industrial labs in the world. The quality of our research has always been at the heart of our success.

How did we do it? Some might say it was luck. Perhaps, but I would also say there are some very specific reasons. We started with an open research environment in Beijing, which at the time was groundbreaking for an industrial lab. We’ve had an unwavering belief in research from our leaders, starting with Bill Gates, and then Steve Ballmer, and today Satya Nadella. And we created a culture that was grounded in questioning the past, challenging the status quo, and constantly looking to innovate.

Globally speaking, our research was what put MSRA on the map. And that is still true 20 years later. So, since we are celebrating our 20th year, I thought it would be fun to go back and highlight a representative paper from each year, selected by the MSRA leadership team.

If you want to really get up to speed on some cutting-edge computer science research and impress your friends and family, there’s a list of links at the end of this post.

Each of these papers is the work of an amazing research team breaking new ground, creating new ideas, and shaping the future. They are the foundation for cutting-edge research projects, incredible breakthroughs and new technologies that have driven innovation. And many are now part of products that millions of people use every day.

Earlier this year, I wrote a post sharing my belief that the act of writing helps us think deeply, express who we are at our greatest, and expand the intellect of those around us.

This is particularly true of research writing. Each of these 20 papers is the expression of hundreds and perhaps thousands of hours of deep thought, creativity and incredibly hard work, intended to further the course of the field of computer science. That’s a pretty ambitious and important goal, and one I take very seriously as part of my job and career.

It’s amazing to think that each of these 20 papers all began as words on the page—expressions of ideas from individuals and groups who were passionate about imagining how the future could be different. Those words, translated through action, determination and ingenuity, profoundly changed the future in very real ways.

Whether you are a computer science researcher or just someone who has an opinion on what you would like to change or create in the world, that’s a powerful lesson!

Because of our research, MSRA’s reach has influenced the world in ways we probably can’t yet comprehend. Our global investments in Microsoft Research have really paid off—not just in our labs, but in shaping our industry and contributing to new products that have helped people around the world.

More specifically, MSRA has had significant impact in these five areas of computer science globally:

1.Web & Search—reinventing the search experience, moving from simple keywords to a more intelligent experience with advances in ranking, visual and video search.

2. Networking & Systems—building next-generation infrastructures and platforms for sensors, devices and the cloud. 

3. Machine Learning & Intelligence—developing and improving algorithms that help computers learn from data to create more advanced, intelligent computer systems.

4. Speech & Language—working to create computers that can understand human language naturally and communicate more like we do.

5. Vision & Graphics—advancing areas like rendering, modeling, face recognition and object detection; creating machines that can “see” like we can. This is my own area of study, so I know the incredible journey it has taken to get us from where I started to where we are today.

I’m so inspired by the beautiful story of what MSRA has been able to accomplish in 20 years. MSRA has educated thousands of young researchers, graduated thousands of alumni who have gone on to lead companies both large and small, and affected so many elements of everyday life with new technologies. 

MSRA has been an extraordinary result of the right place, at the right time, with the right people who created the right environment and built the right culture—a legacy which I hope will endure for the next generation of computer scientists and researchers.

There is a Chinese proverb that says, “If you plant for one year, plant rice. If you plant for ten years, plant trees. If you plant for 100 years, educate mankind.”  

I cannot wait to see what will come out of MSRA in the next 20 years!

Microsoft Research Asia - 20 Years of Research:

Web & Search

• Detecting Web Page Structure for Adaptive Viewing on Small Form Factor Devices (2003)
• Learning to Cluster Web Search Results (2004)
• Adapting Ranking SVM to Document Retrieval (2006)
• Learning to Rank: From Pairwise Approach to Listwise Approach (2007)

Networking & Systems

• R2: An Application-Level Kernel for Record and Replay (2008)
• Sora: High Performance Software Radio Using General Purpose Multi-Core Processors (2009)
• Performance Debugging in the Large via Mining Millions of Stack Traces (2012)

Machine Learning & Intelligence

• Support Vector Machines Classification with a Very Large-Scale Taxonomy (2005)
• Comet: Batched Stream Processing for Data Intensive Distributed Computing (2010)
• Global Illumination with Radiance Regression Functions (2013)
• Deep Residual Learning for Image Recognition (2016)

Speech & Language

• Conversational Speech Transcription Using Context-Dependent Deep Neural Networks (2011)
• Gated Self-Matching Networks for Reading Comprehension and Question Answering (2017)
• Joint Training for Neural Machine Translation Models with Monolingual Data (2018)

Vision & Graphics

• Rendering with Concentric Mosaics (1999)
• Plenoptic Sampling (2000)
• Real-Time Texture Synthesis by Patch-Based Sampling (2001)
• A User Attention Model for Video Summarization (2002)
• Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition (2014)
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (2015)