Microsoft Research hosted events at our labs on Sept. 27, 2011, to celebrate the past, present, and future of computing research on our 20th anniversary. This was a globally integrated live celebration and webcast which featured both our own researchers and invited special guests presenting across six locations, as we explored the future of computing and technology, and reaffirmed our commitment to fundamental research in the 21st century.
At the New England lab, we held a symposium for more than two hundred attendees, mostly academics but also for media, and technology and policy makers. Our symposium featured a dozen short non-technical talks and a panel highlighting research areas in which the Microsoft Research New England lab is actively engaged, including Computer Science theory, machine learning, cryptography, security, cloud computing, computational biology, healthcare, empirical economics, social media and privacy. What follows are the recordings from this symposium, which capture the essence of breadth of interdisciplinary research being conducted at Microsoft New England.
|Welcome to the Microsoft Research New England 20th Anniversary Symposium
Jennifer Chayes, Microsoft distinguished scientist and managing director of Microsoft Research New England, welcomes visiting researchers and staff to Microsoft Research’s 20th-anniversary event in Cambridge, Mass. The event also marked the third anniversary for the New England lab. Vision research, computational biology, and additional theoretical areas have been expanded at that lab, in addition to its general research and collaborative efforts with other institutions.
|Communication, Computing, and Technology
Madhu Sudan, principal researcher at Microsoft Research New England, discusses the differences between theoretical computation versus communication. “Communication,” in this context, refers to mobile phones, email, and other messages, while “computation” refers to things one does with a computer. These technologies are interdependent, and currently, one can no longer exist without the other. The contradiction lies in the practice where these technologies are being developed separately, creating missed opportunities and potentially slowing development in both fields.
|Cryptography Resilient to Physical Attacks
Yael Kalai, researcher at Microsoft Research New England discusses how, with all the progress made in the field of cryptography, security continues to be breached. Complicating matters are hackers, who are becoming increasingly sophisticated, almost making security worse than it was more than 20 years ago. Communications remains the largest field of encryption leaks. Physical attacks can consist of timing attacks, power attacks, and acoustic attacks, while tampering attacks include fault and radiation attacks.
Classic security uses access control, and, essentially, that does not work. Current systems are designed to say “no” and otherwise limit access, while another reason is code bugs. Real-world security currently is retroactive, an example being the financial system, in which mistakes or breaks can be undone after they occur. Typical access controls consist of both physical and principal boundaries, isolation of an item or information, control of access into this isolation, and policies stating who can have access.
|Algorithmic Pricing of Online Services
Cloud computing is a quickly growing business. Questions exist as to the proper business model, and this translates into how to model this, considering auction and game theory. Models exist to purchase on-demand and per-use, as well as to buying in bulk and making longer-term commitments. Regarding supply and demand, marginal costs in cloud computing are among the challenges. External forces, such as electrical costs, are variables out of the control of the hosting company. Another factor is simply time, where optimal prices cannot be exactly determined because of time shifting by consumers. Finding the optimal portfolio of computing products requires both strategic and tactical aspects of pricing, found through combining optimization, algorithms, and game theory.
|Systems Biology: Where Computer Science, Engineering and Biology Meet
During the last decade an entirely new approach to studying biology has emerged from the collaboration of traditional biologists with those trained in other fields. New measurement techniques are used to observe thousands of properties of cells, and algorithmic approaches are applied to assemble these data into a coherent picture of how cells function in particular conditions. These methods are beginning to reveal previously unrecognized cellular pathways that could be targeted to treat diseases.
|Machine Learning and Crowdsourcing
Adam Kalai of Microsoft Research New England—in association with Serge Belongie of the University of California, San Diego; Ce Liu of Microsoft Research New England; George Pierrakos of University of California, Berkeley; Ohad Shamir of Microsoft Research New England; and Omer Tamuz of the Weizman Institute of Science—explains the research focus and issues with machine learning and crowdsourcing. Much of machine learning works with similarity: how similar two separate items are to one another. Having computers create associations or similarities requires context at times, as in the use of numbers. People might look for shape similarities, while the values of numbers are vastly different to the computing system.
|Data-Driven Decision Making in Healthcare Systems
Mohsen Bayati of Stanford University explains how machine learning could and is being examined and used to determine ways to make health care more cost-effective. One example is patient readmission rates, with the most common occurrences in from past studies being from elderly and Medicare patients. A variety of reasons persist, but the surprising fact is many of these readmissions could have been avoided with a small amount of preventive care in the first place. Medication mismanagement is among the top reasons, and heart failure also is listed. A patient’s lack of access to care outside of the hospital is also a major factor for readmissions.
|Health Care Decisions in the Information Age
Benjamin Handel of UC Berkley focuses on how individuals interact with the medical care sector to make important decisions. This involves using the information available and the incentives they face. The incentives can be financial and non-financial for both the consumer and the physician. For the consumer there can be the incentive of deductibles and high costs associated with switching plans. For physicians, fees are assigned to procedures. IT support in the health field is far behind IT support in other areas of the economy, and there is no model for physicians to compare to one another. This, and the patient or consumer and physician usually have no previous knowledge of one another's history.
|Designing Auction-Based Marketplaces: Research and Practice at MSR
Platform markets have two groups of users, and these have both direct and indirect network effects throughout the market as a whole. Platform markets include media markets, credit cards, dating, and video games. There are also auction-based platforms, dealing more with advertising, display, and search. What matters in market design is the focus on efficiency and long-term participation, extracting rents and determining participation and first-order issues between competitions.
|Economics of Privacy
Markus Mobius and Susan Athey of Harvard University discuss changes over the last five years in the way people consume online news. In the past, individuals would consume online news similarly to how they would read a traditional newspaper. Currently, the trend has become using intermediaries, such as content aggregators and social media. While some major organizations disagree with the aggregation model, other organizations defend it, citing new reach to readers for their sites.
|Six Provocations for Big Data
With big data come big responsibilities, specifically in the realm of social media. There is a hint an emergence of a philosophy that big data is all one needs and that other scientific methods and disciplines are becoming outdated and unimportant. But big data can be incomplete because of how fragile data sources such as the Internet can be. Different social behaviors can be mapped but not necessarily compared because of the social channels being used. For example, a relative to whom one is close might not be online or involved with social media. Therefore, an incorrect assumption could be made that an individual is actually not close to that relative because of the visible data collected.
|Privacy in the Age of Augmented Reality
Alessandro Acquisti, an economist by training and a self-described behavioral economist, explains that across history, time, and culture, people have felt the need for publicity, as well as the need for privacy. These are not contradictions, but these concepts are at times in conflict. Current technology has opened the way to potential issues and problems, such as where an individual can share non-sensitive information that could eventually lead to the sharing of sensitive data.
|Privacy: Issues and Perspectives
Moderator Ethan Zuckerman hosts a panel in Cambridge, Mass., discussing issues of human rights, as well as individual and public privacy. Difficult questions on privacy include personal privacy and marketing data, and technology and tagging that could erode privacy. Privacy is also culturally sensitive, not only in relation to the law and respect for laws, but also with regard to who is and is not entitled to privacy and which technologies are self- or socially monitored.