Editor's note: 30 years ago, Microsoft founder Bill Gates envisioned that computers would one day see, listen, read and learn, and founded Microsoft Research in 1991 with the goal of focusing on cutting-edge technology research and innovation. Today, 30 years later, we have entered the era of intelligence, and Bill Gates's original vision of computers has come true. How have Scientists at Microsoft Research collaborated to bring ideas to life step by step over the course of 30 years?
In a series of events celebrating the 30th anniversary of Microsoft Research, Microsoft Research organized a series of expert forums under the theme of "Generations of inspirational and impactful research." We have compiled and compiled the content of the forum, and will release it in the near future to share these wonderful feasts of ideas with you.
Co-hosted by Eric Horvitz, Chief Scientific Officer of Microsoft, and Jenna Wiens, Associate Professor of Computer Science and Engineering at the University of Michigan at Ann Arbor, this forum joined Scientists, Engineers, and partners from Microsoft Research from around the world to discuss the magic of scientific collaboration and share their views on the dynamism, creativity, and impact of scientific research.
Experts participating in the forum include: Rick Szeliski, associate professor of computer science and engineering at the University of Washington and honorary researcher at Microsoft, Jessica Wolk, senior research engineer at Microsoft Redmond Research, Zhang Dongmei, executive vice president of Microsoft Research Asia, Lidong Zhou, president of Microsoft Research Asia, and Sriram Rajamani, president of Microsoft Research India.
On the magic of collaboration in science & technology：「链接」
Here's a compilation of the forum's highlights:
Eric Horvitz: The development of science and technology in the world relies heavily on teamwork, and the best and most influential scientific research is often carried out in the form of teamwork. In teams, passionate researchers can draw on the talents of others to complement each other's strengths, and research innovation is based on the long-term contribution of this broad community of cooperation. Microsoft Research has also always regarded internal and external open cooperation as the lifeblood of sustainable development.
Today, it is a pleasure to hear about your experiences of cooperation. Before giving your time to the experts involved in the discussion, I would like to ask Jenna, you have interned at Microsoft Research, you have done excellent cooperation with colleagues in Microsoft Research and other institutions around the world, what do you think about scientific collaboration?
Jenna Wiens: I think collaboration is necessary to solve really difficult problems, especially those that are cross-cutting. Because research in cross-cutting areas is generally very difficult, you have to learn to use the language of another field, organize a lot of meetings, and communicate frequently. As someone who works in AI and medical research, I meet every day with one or more collaborators in clinical medicine through meetings that keep our work up and running and make sure we're solving really meaningful problems. I love this collaboration, and when we discuss new, exciting ideas or brainstorm with colleagues, the office is full of life and vitality.
Eric Horvitz: It's interesting to start, nurture and sustain a collaboration, and I'd love to hear how the experts started it.
Rick Szeliski: I joined Microsoft Research in 1995, and before that, I started doing industrial research. I'm happy that we have a lot of close collaboration with the University of Washington, where if Microsoft Research researchers are interested, the institute encourages them to connect with the University of Washington, mentoring graduate students, teaching courses, and so on.
Later, we had a postdoc teach at Carnegie Mellon University, and a few years later transferred to the University of Washington. After he joined, we talked about some research ideas, and he invited me to teach courses at the University of Washington. At the time, computer vision and computer graphics were two relatively separate fields, and we did cross-cutting research in these two areas, and we also discussed related topics at SIGGRAPH, so we opened a course related to this.
One day I was talking to a graduate student, Noah Snavely, over lunch about the topics we were working on, and found that our ideas were similar, and we all wanted to create a 3D model that could stereoscopically represent the contents of the photos by taking and analyzing a large number of photos. At lunchtime communication, we have already outlined the outline of this idea. We're not trying to solve a long-standing problem, but to experience photos in a whole new way. I would say that inspiration sometimes comes together by chance in a single communication, not planned in advance.
Eric Horvitz: I remember this project, it was very groundbreaking. Jessica, how did your most interesting and productive collaboration begin?
Jessica Wolk: I come from a different background, I'm a software development engineer on the microsoft research central engineering team at MSR, and our team is involved in different phases of the project, mainly solving engineering problems. In my 10 years at Microsoft Research, I've been involved in about 12 different projects and seen the various ways projects are launched.
Sometimes informally, as Rick just said, in a lunch conversation we found that everyone had a common interest in solving a problem; sometimes it was directly targeted at market needs, and researchers also had to focus on solving specific needs. Research projects can be combined in a variety of ways.
For example, we worked with Microsoft's Edge product team when a software engineering researcher in a chat with an engineer on the Edge product team found that Chromium-based Edge had a lot of problems in keeping pace with Chromium, such as merge conflicts. How does a problem in Chromium affect the downstream Edge? We experimented with different techniques, such as program integration, machine learning, and relatively traditional methods, which ultimately helped the Edge team improve performance. What made this project interesting and productive was that we kept Edge efficient by focusing on solving a problem using different technologies over and over again.
So once some of the approaches don't work or don't work, we'll find it fun to focus on and have the right researchers connect well with the product team and focus on one thing in multiple ways.
Jenna Wiens: What are the most important elements of a fruitful, successful collaboration?
Zhang Dongmei: My background is also a little different. I joined Microsoft in 2001 as a software engineer, and after 3 years in the product department, I joined Microsoft Research Asia in Beijing. I've worked in both the product department and the research institute, and I think there are three key points to the success of these two teams: a shared vision, trust, and common promotion.
The first is a shared vision. The goals that the two teams want to achieve together will not be achieved overnight, and there are many unknowns and uncertainties that need to be explored, and many trials and errors will be experienced. These make collaboration a long-term, step-by-step process, which is in line with the nature of research and meets the need for product teams to stay ahead of the competition.
The second is trust. In my experience, to build trust in cooperation, the most important thing is to keep promises, be responsible, be transparent and inclusive, which are reflected in daily work. For example, can we deliver on time? What are the quality standards we uphold for ourselves and our partners? Do we actively share information, decisions and experiences? Do we actively seek feedback from our partners and incorporate it into our planning and decision-making processes?
Finally, there is the common ascension. This requires us to grow with our partners and go further together. For me, successful collaboration between research teams and product teams should be win-win and mutually beneficial. As innovative technologies are transformed, product teams' products are enhanced and more competitive, and researchers gain experience in solving real-world problems, bringing a sense of accomplishment, while also gaining new inspiration from real-world challenges and discovering new research topics and directions.
Zhou Lidong: Cooperation is one of my favorite topics. In fact, cooperation is like marriage. Just as a happy marriage can have many different ways of getting along, successful cooperation can take many different forms. Also like a happy marriage, I think there are three key elements to a successful collaboration: the first is a chemistry of mutual respect, which is the basis of marriage and the basis of cooperation. The second is diversity, a couple needs to have different perspectives and skills to complement each other, which is also very important for collaboration, and teams need to learn from each other and solve problems that cannot be handled alone. But just like married life, our cooperation will also encounter bottlenecks from time to time, facing various difficulties, so how to get through the difficult times? Then a third element is needed: faith. We must have faith in our common goals, in cooperation itself, in the mission we undertake.
Jenna Wiens: Yes, faith and trust are closely related. Sriram what do you think?
Sriram Rajamani: I very much agree with what Dongmei and Li Dong said. I would like to add that it is also important to learn new knowledge in cooperation. I've worked with many people from different backgrounds. When I first joined Microsoft Research, my background was hardware verification. I have a lot of work with Thomas Ball, principal investigator at Microsoft's Redmond Research Institute, who has a lot of experience with compilers, and he tells me how compilers work. In addition, in my collaborations with researchers in different fields, I have learned how security systems work, learned about machine learning, and so on.
Learning something new is a very important part of the collaboration, and we also have to give back by sharing knowledge with each other. In addition, I would like to add that friendship and fun, I have become lifelong friends with many collaborators, which is also the benefit of being a researcher, you can build your own connections in your work. For example, Eric comes to my house a lot, and we know each other and each other's families, which are the most wonderful things that can be experienced as scientists.
Eric Horvitz: I think these are the key factors for a successful collaboration. But as you know, we do face challenges, especially in collaboration across research groups and teams. So, what are the challenges of cross-team collaboration?
Zhou Lidong: The most important thing for cross-team collaboration is to establish a common goal, this goal must be ambitious enough, it cannot be achieved independently, to achieve this goal, other parties need to participate, cooperation becomes necessary, which is an element of cross-team cooperation. In many cases, if we are not ambitious enough to be far-sighted, then it is possible to retreat into our own small areas and continue to do our own thing. When we share common ambitions and are ambitious enough, it's natural for us to work with teams with a shared vision. Just now Rick said that because of a common goal, computer vision and graphics researchers came together, and that's when the magic of collaboration came into being. People must recognize the importance of setting higher goals and going beyond the expertise we have developed in our respective fields.
Zhang Dongmei: I think consistency is the most challenging part of different team collaboration, especially when the research team collaborates with the product team, from senior management to the functional development team, in terms of goals, priorities, communication, working methods and so on.
For example, as researchers, we always strive to push the boundaries of technology and evaluate technology using metrics commonly used in academia. But when these technologies go into production, we may find that some assumptions don't hold true. In a production environment, product teams have different ways of evaluating metrics such as latency or reaction time requirements, memory usage limits, and so on. Sometimes product teams have strict release timelines and resource constraints, in which case they may prefer relatively short-term, quick, and simple solutions.
In order to achieve consistency, I would like to borrow what Li Dong just said, we need to build faith and trust between teams, be open-minded, understand each other's positions, ideas and problems, find common ground, be willing to negotiate, and compromise.
Jenna Wiens: What are some of the things you particularly like about the Microsoft Research collaboration environment?
Rick Szeliski: The institute's work is multifaceted, and one of the happy things is its multifaceted nature. Earlier I talked about collaborations with academia, joint research projects with graduate students, we also have great internship programs, and every summer the best and brightest people from all over the world work with us and make deep friendships. We also work with university teachers and scholars. Overall, Microsoft strongly encourages collaboration outside the company.
Internally, I also like the free atmosphere of working with different research teams. As mentioned earlier, people need complementary skills, and whether it's walking around the halls or having lunch with others, is the easiest, most direct way for people to stay in touch. Personally, I am pleased that Microsoft Research is an international research institution. I enjoy going to places like the UK, China and India, where I spend some time meeting new people and learning about new cultures.
Regarding the collaboration with the product team, I would like to say that fortunately all of us are in Redmond, and outside of work I make a lot of connections, understand what everyone is doing, and occasionally bump into some good ideas. These informal connections are key to building trust, and over the next 6 to 12 months, we focus on product delivery and get feedback from real-world users that wouldn't be available in a purely research-only environment. These are all delightful.
Sriram Rajamani: What I like most about the Institute is that the expertise we've come into at Microsoft Research is more breadth and depth, and whether you're doing some aspect of security here, or a branch of machine learning, or something else, the Institute always has top experts in that area. At the same time, everyone at Microsoft Research is curious, and the culture here promotes curiosity, and whoever you talk to will be curious about your area of interest and the problem you want to solve. What's more, people are willing to give time for you, no matter how busy they are, and when I go to them, they take the time to communicate with me.
Eric Horvitz: In any complex relationship, setbacks and resistance are common. I think all of you here have encountered and overcome difficulties. I'm curious, how did you cope with setbacks and work together smoothly?
Jessica Wolk: I agree with Sriram and Rick about Microsoft Research culture. The Institute is very diverse, and these are part of our "wealth", both in terms of the breadth and depth of our expertise, in terms of culture, and in terms of people's qualifications.
However, when it comes to putting different types of people together, it does take time to understand each other's pain points and their respective strengths and weaknesses. When things don't go well, some stereotypes can arise, usually out of ignorance or misunderstanding. I think the way to overcome similar difficulties is to give the project and team members more tolerance and space, because we are all different "fireworks" and need to take the time to get to know each other.
But because we're in a fast-paced, high-tech environment, tension and uneasiness are certainly there. We want to be the first to enter the market or to be the best in the market, so we have to be patient with ourselves and the project, and think differently to alleviate the uneasiness. It's a bit like what Satya Nadella, Chairman and CEO of Microsoft, calls empathy, to understand each other's pain points, purpose, what other teams are discussing, what the organization needs, what management needs, what individuals need. We can patiently listen to each other, build empathy, and add some sense of humor to our collaborations, creating a relaxed, expansive atmosphere. I want to emphasize again that diversity is definitely our asset, but we have to master some skills to learn to meet challenges in a very diverse environment.
Rick Szeliski: Jessica talks about people having different agendas and goals, which is easy to overlook because everyone's mindset is deeply ingrained. People tend to get excited about new ideas and overlook what's important and what the end goal is, they may be different, but they're all going to come across when working together.
Corporate organizations and rules, etc., will also become challenges. For example, working within Microsoft, intellectual property is not a problem, because we create the intellectual property, patents, and code ourselves, but once we expand outside the organization, that's a problem. So sometimes the company's lawyers don't like to cooperate openly, and some academic institution leaders want to monetize their academic achievements and so on. It is helpful to establish an established framework outside of a particular area of study. I've always been grateful to Microsoft and Microsoft Research for going to great lengths to build these frameworks.
The other point I would like to say has to do with relationships. Our academic backgrounds excite us for creative, novel and practical ideas, but forget about building relationships. Sometimes we need to make trade-offs and learn to take a step back when your ideas aren't being accepted. If doing so builds trust and makes the other person feel valued, then compromise is the right decision.
Zhang Dongmei: Let me briefly share my views on interpersonal relationship construction. I very much agree with what Jessica said that we have a lot to do in this area; I would like to add that because of the restructuring, there are a lot of things that we have to do many times. We know that restructuring happens a lot in the product team, sometimes the direction of the product team changes, sometimes our contacts change. These changes can be frustrating because we can't do anything about it, new directions may not align with our research, and we have to look for new opportunities; sometimes we can do things like actively connect with new people, and the documentation of our past work can help new teams get to know us and explore new opportunities for collaboration. As Rick said, we can also interact with them offline, although we can't officially cooperate with all of our contacts at the moment, but be brave, maybe in the near future they will become your new collaborators. We should keep trying to sow the seeds, trying to find opportunities for cooperation, and everything starts with trying.
Eric Horvitz: It's great to hear about successful collaborations and coping with setbacks. Now, please do some quick thoughts for each of you and share some of the most exciting moments of collaboration in years.
Jessica Wolk: My most exciting moment of collaboration is when a new product is delivered. It's a joy when you're researching novel ideas, working hard, and finally being implemented and delivered to Volkswagen or other Microsoft employees.
When I worked in the Language Understanding and Intelligence Services group, it was a standalone business, which later became an important part of Microsoft's Azure Cognitive Services. It's really cool to see an idea go from infancy to final public release and have an impact on both the company and the developers. I'm really happy to be part of the project.
Sriram Rajamani: We worked with Thomas Ball on a software model detection system, including model inspectors, extraction systems, path analysis systems, and so on, and we did that for several years, and then Thomas and I programmed them in pairs and wrote a Perl script to put them together.
When we first ran this multi-year development, we didn't know if it would work properly. The moment we were all staring at the screen and watching it run and produce useful results was so magical. I think that when every project first runs code, people don't know whether it succeeded or failed, and they remember the moment it worked.
Zhou Lidong: As a researcher in the field of systems, I also feel the same way, when I see our carefully designed and implemented system working normally, that moment is always so exciting. I also like what Sriram says about learning, I worked at three institutes at Microsoft Research for 19 years, during which time I learned a lot from everyone, such as distributed storage systems, mobile wireless systems, and other areas, and recently we also partnered with Cornell University to introduce the concept of social choice into distributed systems.
The collaboration that I'm most excited about is what I see my colleagues at the Institute doing right now — researchers in natural language processing, machine learning, computer vision, and systems and networks are working closely together to drive the convergence of AI and the breakthrough opportunities that this convergence brings. In the past, these disciplines have developed in their own independent ways, there has not been much communication between the fields, and they have their own academic conferences. Now, we see the benefits of this cross-border collaboration, bringing new insights and insights. Although there are many differences between the various fields, there are also many commonalities. These cross-cutting collaborations are not easy, but they have greatly expanded our horizons and perspectives, and we are very excited about the new opportunities we have discovered in the collaboration.
Zhang Dongmei: Simply put, the moment that excites me the most is the moment of solving difficult problems and celebrating success with our partners.
Rick Szeliski: Going back to the story I told at the beginning, it may not be a big moment, just some daily joy: when you come out of a conference room and you have a new idea, a new idea, or a new understanding, but without the group of people in the room, those ideas may never come up.
But the most exciting moment was when we had that idea at lunch, And Noah went back and wrote the program, and we could go through more and more 3D photos of our own creation on the screen every week. It only took us a week to put together a lot of photos to form this 3D model, and it felt like I'd gone through a magic spell because I'd never seen anything like it, and we thought the whole world would feel the same way. Eventually, it also became a very popular presentation at academic conferences. I think that before success in public, even just seeing something that exceeds our expectations is already a very happy moment, and that may be the most wonderful part of the collaboration.
Jenna Wiens: So Eric, what was the most exciting moment of collaboration for you?
Eric Horvitz: That was when I worked with you before, Jenna. Establishment of C. difficile in the Washington Hospital Center in Washington, D.C. Difficile) predicts the model, makes it work, docks with real-time electronic cases, and that's really interesting, and it's probably the first in the world.
Finally, thank you very much Jenna for co-chairing this forum with me. Thank you for sharing your views on cooperation and its importance to promoting the development of the frontier of science and technology! We look forward to seeing you forge ahead in ongoing and upcoming collaborations!