Wednesday, July 8
-
Chair: Margaret Mitchell, Microsoft Research | video
Speakers:
- Ray Mooney, University of Texas at Austin | slides
- Dan Roth, University of Illinois at Urbana-Champaign | slides
- Lucy Vanderwende, Microsoft Research | slides
- Benjamin Van Durme, Johns Hopkins University | slides
Everyday intelligence relies on a large store of background—or “common sense”—knowledge about the world. As humans, we acquire this knowledge in part through our experiences; artificially intelligent systems currently do not have access to this same kind of input. However, both explicit and implicit information about the world can be learned from data available in text and images. This session focuses on some key research in this area, with talks on extracting information from data in order to acquire commonsense knowledge about the human world.
-
Chair: Krysta Svore, Microsoft Research | video
Speakers:
- Edward Farhi, Massachusetts Institute of Technology | slides
- Matthias Troyer, Eidgenössische Technische Hochschule Zürich | slides
- Nathan Wiebe, Microsoft Research | slides
In 1982, Richard Feynman first proposed using a quantum computer founded on the laws of quantum physics to simulate physical systems and achieve exponential computational speed-ups over conventional computers. In the thirty years since, quantum algorithms have been invented to solve problems in fields like number theory, chemistry, and materials science that would otherwise take longer than the lifetime of the universe to solve on an exascale classical machine. Quantum algorithms promise ways to break RSA (a mainstay of e-commerce), combat global warming, and design room-temperature superconductors. In addition, recent advances show how quantum computers can learn better deep machine learning models for use in speech and vision tasks. This session highlights killer applications of quantum computers and the potential global impacts, both scientific and societal, as well as pose challenging open questions for the computer science community to tackle.
-
Chair: Carolyn Nguyen, Microsoft | video
Speakers:
- Susan Aaronson, George Washington University
- Laura DeNardis, American University
- Stephanie Forrest, University of New Mexico
- Brenden Kuerbis, Syracuse University
Internet governance (IG), at its core, is a global discussion between governments, businesses, civil societies, technical experts, academic researchers, and other interested parties on how to shape the evolution and use of the Internet. A central debate concerns whether a federated multi-stakeholder approach, in which all interested parties can participate, is a more appropriate model than a centralized intergovernmental model, where countries would enter into treaties that are negotiated at the government-to-government level. Another issue concerns the role of the US government in managing the operations of the Internet.
This session aims to explain the issues involved in Internet governance, the upcoming milestones that make 2015 a critical year for engaging in IG, and explore how academic researchers can enable more thoughtful, evidence-based dialogues with policy stakeholders in the ongoing debates on issues ranging from required technology innovations, to economic analyses, multi-stakeholder policy frameworks, and studies on socio-cultural impacts.
-
Chair: Judith Bishop, Microsoft Research | video
Speakers:
- Gul Agha, University of Illinois at Urbana-Champaign | slides
- Philip Bernstein, Microsoft Research | slides
- Sergey Bykov, Microsoft | slides
Building interactive services that are scalable and reliable is hard. Interactivity imposes strict constraints on availability and latency, as that directly impacts end-user experience. To support a large number of concurrent user sessions, high throughput is essential. Many mobile device applications are backed by cloud servers and storage, but the current technology for programming cloud applications is tedious and potentially error prone. Developing individual application components is not difficult, but developing an entire system that is scalable and fault tolerant and makes efficient use of resources is far more challenging, especially for mainstream developers who are not distributed system experts. In this session we look at modern approaches to solving these problems. A variety of applications require interactive services and we’ll discuss a few, such as game platforms and Internet of Things.
-
Chair: Larry Zitnick, Microsoft Research | video
Speakers:
- Julia Hockenmaier, University of Illinois-Urbana Champaign | slides
- Margaret Mitchell, Microsoft Research | slides
- Richard Zemel, University of Toronto | slides
The recent advances in computer vision, natural language processing and other related areas has led to a renewed interest in artificial intelligence applications spanning multiple domains. Specifically, the generation of natural human-like captions for images has seen an extraordinary increase in interest. In this session, the speakers provide insight into this area. They describe several techniques that combine state-of-the-art computer vision techniques and language models to produce descriptions of visual content with surprisingly high quality. The limitations of current approaches and the challenges that lie ahead are both emphasized.
-
Chair: Ed Cutrell, Microsoft Research | video | slides
Speakers:
- Elizabeth Belding, University of California-Santa Barbara
- Andrew Cross, Microsoft Research | slides
- Aaditeshwar Seth, Indian Institute of Technology, Delhi | slides
As the cost of information technology falls and the reach of communication systems grows ever more pervasive, hundreds of millions of people are being exposed to computing technologies for the very first time. This gives rise to the question: can all this technology be used to help people in the world’s poorest regions improve their lives? What can computing technology do for the unique context of a wage laborer in India, a farmer in Tanzania, or a child in a São Paulo favela? This session focuses on some of the ways that research in IT and computing can contribute to the well-being of people in communities under severe resource constraint in terms of finances, language, education, infrastructure (power, Internet access) and availability of devices. Our speakers describe innovations in networking, health-care administration, and media sharing for communities in challenging contexts in the global south. We hope to engage the audience in a dialog on the challenges and unique rewards of computing research in these areas.
-
Chair: Kristin Tolle, Microsoft Research | video
Speakers:
- Lili Cheng, Microsoft Research | slides
- Ann Quiroz Gates, University of Texas at El Paso | slides
- Charles Isbell, Georgia Institute of Technology | slides
- Ed Lazowska, University of Washington | slides
The world is moving to create buildings, products, devices, services, and real and virtual environments that are more universally designed. It is possible to create environments that are usable and enable the success of all people regardless of age, culture, gender, or disability. Given the importance of computing as an underlying infrastructure technology, it is critical to the success of this movement to graduate more computer scientists from diverse groups, in particular PhDs, who can contribute to solutions with perspectives that may challenge conventional thought. With only 3.3% of the 2012–2013 doctoral graduates in computing coming from underrepresented ethnic groups, new and different recruitment and retention approaches at our academic institutions are required. This session introduces the new field of universal design and then panelists discuss the challenge of keeping up with the increased demand in computer science programs and hiring of underrepresented minorities. They suggest some controversial solutions to the industry issue. Join us for a lively discussion.
-
Chairs: Nikolaj Bjorner and Ratul Mahajan, Microsoft Research | video | slides
Speakers:
- Nate Foster, Cornell University | slides
- Sharad Malik, Princeton University | slides
- George Varghese, Microsoft Research | slides
- Geoffrey Xie, Naval Postgraduate School | slides
Surveys reveal that network outages are prevalent, and that many outages take hours to resolve, resulting in significant lost revenue. Many bugs are caused by errors in configuration files, which are programmed using arcane, low-level languages, akin to machine code. Further, mistakes are often hunted down using rudimentary tools such as Ping and Traceroute. Taking our cue from other fields such as hardware design, we would like to explore fresh approaches and contrast them to standard approaches to verification using model checking, SAT solvers, and so forth. While network verification is similar to finite state machine verification, there is domain-specific structure that can be exploited and a different set of properties to verify. Early results suggest that concepts from EDA and program verification can be leveraged to create what might be termed Network Design Automation. What might the equivalent of Layout Versus Schematic tools or Specification Mining be? Could there be a theory of types for networks? Our panelists explore this vision, touching upon modular network semantics, language design, performance invariants, and interactive network debuggers. We also explore, with the help of SDN experts, the connections between this vision and the vision of Software Designed Networks.
-
Chair: Matthew Richardson, Microsoft Research | video
Speakers:
- Yan Ke, Microsoft | slides
- Benjamin Van Durme, Johns Hopkins University | slides
- Scott Wen-tau Yih, Microsoft Research | slides
- Luke Zettlemoyer, University of Washington | slides
Natural-language question answering (QA) has clear practical and scientific values, such as evaluating a machine’s understanding of a domain, or providing succinct and precise answers to search engine queries. While both Bing and Google have incorporated more “semantics” to return direct answers to queries, QA is far from solved, and is becoming more important as natural language interaction becomes popular (e.g., Siri and Cortana). In this session, we invite experts from both academia and Microsoft to present recent technologies for improving QA. The topics include traditional IR approaches for QA, machine reading for knowledge acquisition and representation, and semantic parsing for answering questions using structured databases like Freebase or Satori. In addition, we invite product groups (Bing QnA team) to discuss the technical challenges faced in the real-world scenarios and highlight the research need.
-
Chairs: Bill Buxton, Microsoft Research; Jeff Han, Microsoft
Speakers:
- Niklas Elmqvist, University of Maryland | slides
- Steven Feiner, Columbia University | slides
- Tracy Hammond, Texas A&M University
- Andries van Dam, Brown University | slides
Microsoft recently unveiled plans to go to market with a brand new large touch device, the Surface Hub. The opportunity for this device to enhance and even transform how some research areas visualize, capture and display findings is brand new territory. This session shows some of the earliest efforts to understand how these new, powerful devices can provide exciting new platforms for research.
-
Chair: Baining Guo, Microsoft Research | video
Speakers:
- Martial Hebert, Carnegie Mellon University | slides
- Tao Mei, Microsoft Research | slides
- Zhengyou Zhang, Microsoft Research | slides
Recent advances in image understanding have opened many avenues for computer vision-based intelligent services. These technological advances—which span a broad range of visual understanding tasks including object recognition, video event detection, image search, and scene reconstruction—can facilitate a variety of image analysis applications from visual identification of users to 3D city mapping and detecting child exploitation images. Moreover, by harnessing the power of the cloud, computationally-intensive vision tasks have become increasingly accessible to ordinary users. This session focuses on the latest developments in image understanding as well as a discussion on consumer services that they may enable.
-
Chair: Aakanksha Chowdhery, Microsoft Research | video
Speakers:
- Sharon Gillett, Microsoft Research | slides
- Sundeep Rangan, New York University | slides
- Xinyu Zhang, University of Wisconsin Madison | slides
- Heather Zheng, University of California Santa Barbara | slides
The demand for Internet speeds continues to explode with the new applications such as augmented reality, artificial intelligence on mobile devices, and Internet of Things. As more and more wireless devices connect to the Internet and more and more data flows to and from the cloud, the wireless spectrum once deemed sufficient to handle such traffic quickly is getting stretched. The challenges include enabling 1–100 Gigabits/sec of speeds per device to 1,000x more devices within 10-millisecond latency at lower costs and longer battery lives. Come and learn about how 5G promises to overcome these challenges from both technical and policy perspectives of a panel of prominent academic and Microsoft researchers.
Thursday, July 9
-
Speaker: Monica Lam, Professor, Stanford University | video
With the widespread adoption of proprietary social networks like Facebook and mobile chat platforms like Wechat, we may be heading to a future where all our communication are monetized and our online transactions are mediated by closed monopolistic big-data companies.
This talk describes an open social movement led by Omlet, an open messaging service and distributed computing platform that spun out of four years of research at Stanford University. To the user, Omlet appears as a super chat app with many plug-ins and extensions; deep down, Omlet is actually a distributed social OS and network. At the heart of Omlet is a universal messaging system where devices can communicate with each other via existing identities such as phone numbers or email addresses, without giving up ownership of the communication data. Built upon this messaging layer is a distributed semantic file system that enables collaborative apps be written easily while allowing data be distributed in the cloud service of the users’ choice.
Launched in March 2014, Omlet has already been distributed to millions of phones. Because of its data protection policy, Omlet is a suitable foundation for social applications in education, commerce, finance, and Internet-of-things.
-
Chair: Eric Horvitz, Microsoft Research | video
Speakers:
- Dan Bohus, Microsoft Research | slides
- Manuela Veloso, Carnegie Mellon University
- Larry Zitnick, Microsoft Research | slides
Over the last 15 years, significant advances have been made in inference and decision making under uncertainty, machine learning, vision, speech recognition, natural language processing, and dialog. Supported by these advances, efforts to develop end-to-end intelligent systems that integrate multiple competencies and act in the open world have brought into focus new research challenges. Interesting opportunities arise at the confluence of traditional AI fields, for instance between computer vision and natural language processing. The science of engineering larger, integrated systems that are efficient, robust, transparent, and maintainable is still very much in its infancy. The speakers in this session provide a sampler of research challenges, from work at the intersection of vision and language, to situated dialogue systems and interactive robots, and highlight opportunities in integrative AI. The session was followed by a panel discussion.
-
Chair: Ben Zorn, Microsoft Research | video
Speakers:
- Jaeyeon Jung, Microsoft Research | slides
- Ben Livshits, Microsoft Research
- Arvind Narayanan, Princeton University
Big data equals Big Brother? The big data explosion forces us to confront the privacy implications of this wide-ranging societal change. In this session, we talk about privacy-related efforts: both in Microsoft Research and in academia. What does privacy means in the context of always-on sensors such as Kinect? What happens to parents and teenagers living in Internet-connected homes? How do we sanitize documents on a large scale? How can we quantify privacy and privacy loss? These topics and others are discussed with the goal of creating shared understanding as well as collaborative research opportunities.
-
Chair: Kathryn S. McKinley, Microsoft Research | video
Speakers:
- Michael Carbin, Microsoft Research | slides
- Noah Goodman, Stanford University | slides
- Dan Grossman, University of Washington | slides
- Todd Mytkowicz, Microsoft Research | slides
Many emerging applications produce and use estimates in domains as diverse as probabilistic reasoning, machine learning, data analytics, approximate computing, and sensor programming. This session focuses on programming language models for expert developers and the masses (application programmers without statistics background). In particular, we explore how well current languages support developers in producing estimation models, computing with estimates, and reasoning about what the resulting programs mean.
-
Chair: Dan Bohus, Microsoft Research | video
Panelists:
- Eric Horvitz, Microsoft Research
- Charles Rich, Worcester Polytechnic Institute
- Manuela Veloso, Carnegie Mellon University
- Lidong Zhou, Microsoft Research
- Larry Zitnick, Microsoft Research
The panel discusses challenges and opportunities in integrative AI, an area concerned with the development of systems that efficiently and robustly integrate multiple AI technologies (e.g., vision, speech, learning, real-time control) and can act and interact in the open world. While over the past years we have seen significant progress on core AI technologies, the science of integrating these competencies into larger systems that are robust, transparent, and maintainable remains very much in its infancy. Numerous research and engineering questions come to the fore: from algorithmic challenges in multimodal representations and modeling, to the science of systems that self-monitor, diagnose their limitations and improve through interaction, all the way to the necessity of research platforms, data, and broader community efforts in this space.
-
Chair: Jaron Lanier, Microsoft Research
Speakers:
- Mark Bolas, University of Southern California | slides
- Steven Feiner, Columbia University | slides
- Henry Fuchs, University of North Carolina | slides
- Shahram Izadi, Microsoft Research
- Ken Perlin, New York University
This is a special time for research in VR and mixed reality. Consumer products such as Valve Vive, Oculus Rift, and HoloLens will soon be available. While long anticipated, current developments bring up questions that are not yet answered: Should researchers still seek to create VR hardware that remains ahead of the commercial curve? Can that even be done? Or, are there cloud architectural challenges that the research community is better suited to attack than industry? Broadly, how can researchers best interact with an explosive, competitive industry? Are there important insights that commercial developers haven’t yet gleaned from the VR research community? How can and should the wave of “consumerization” influence basic VR research? What are the best experimental methods for engaging large numbers of users in VR-based human factors/cog sci/social research? Might commercial interests steer research away from important methodologies or topics? Are there new ethical questions?
-
Chairs: Alex Wade and Yang Song, Microsoft Research | video
Speakers:
- C. Lee Giles, Pennsylvania State University | slides
- Oren Etzioni, Allen Institute for Artificial Intelligence | slides
- Jie Tang, Tsinghua University, China | slides
- Kuansan Wang, Microsoft Research | slides
As web search is becoming more and more structured thanks to the availability of knowledge bases, scholar engines are also evolving towards that direction while facing the similar challenges and issues. This session focuses on bridging the gap between unstructured text and structured relationships to create new sematic scholar engines. We hope this session can serve as a discussion forum to bring together researchers from both academia and industrial labs from different disciplines to work together and exchange ideas on how to jointly optimize scholar engines and knowledge base systems while sharing their respective experiences in dealing with common issues in both areas.
-
Chair: Judith Bishop, Microsoft Research | video
Speakers:
- Ross Gardler, Microsoft
- Phil Haack, GitHub | slides
- Carlos Jensen, Oregon State University | slides
- Rustan Leino, Microsoft Research | slides
Open source is a powerful way of advancing software development and sharing data for experimentation. Microsoft has started putting research projects as well as key software such as .NET in open source repositories. Developers are jumping on to fix bugs and extend the systems. Academia has long wanted access to industrial projects, but now that it is a reality, what exactly can be done and by whom? There are many options, for example: instructors can use the software as illustration in their classes; software can be extended for research purposes; capstone projects can be based on OSS; and open data can be shared for analysis. The panel of experts bring their experiences with using open source, from repositories, creation, and education.