SIGMOD 2022: Keynote Talks
Keynote Speaker 1: Barbara Liskov (MIT)
Reflections on a Career in Computer Science
Abstract
Computer Science is a wonderful field with many interesting and important problems to work on. This is true today and was also true in the past; of course what is considered interesting changes as a field matures. In this talk I will discuss some of the problems that I found intriguing. In each case I will discuss the state of the field at the time I did this work to provide an historical context for my work and that of others doing related work. I will also discuss how I selected problems, what I brought to the table, and where the ideas for my solutions came from.
Bio
Barbara Liskov is an Institute Professor at MIT. Her research
interests include distributed and parallel systems, programming
methodology, and programming languages. Liskov is a member of the National Academy of Engineering, the National Academy of Sciences, the National Inventors Hall of Fame, and the
Massachusetts Academy of Sciences. She is a fellow of the
American Academy of Arts and Sciences and the Association for
Computing Machinery, and a charter fellow of the National
Academy of Inventors.
She received the ACM Turing Award in 2009, the IEEE Von
Neumann medal in 2004, the IEEE Pioneer Award in 2018, a lifetime achievement award from the Society of Women Engineers
in 1996, the ACM SIGPLAN Programming Language Achievement
Award in 2008, the ACM Sigops Hall of fame award in 2012, and
the Stanford Hero of Engineer ing award in 2019.
Keynote Speaker 2: Laks V.S. Lakshmanan (University of British Columbia)
On A Quest for Combating Filter Bubbles and Misinformation
Abstract
The advent of social networks and media has made it easier than ever for users to access up-to-date information as well as share news and views on matters of the world with many of their peers. Unfortunately, it has also led to increased societal polarization as well as deteriorating trust in institutions. Two of the problems that are blamed for this are filter bubbles and misinformation1. Filter bubbles are the result of excessive personalization which has the benefit of enhancing relevance but comes at the price of limiting the exposure of users to a specific viewpoint. They are amplified by the so-called echo chambers that exist in social media, whereby members of a community mutually reinforce a fixed opinion or viewpoint on an issue. Misinformation, on the other hand, tends to propagate through the network, and studies show it does so faster and more virally than truth.
Both problems manifest themselves in the form of groups of actors working in concert and providing mutual reinforcement. How can we recognize these groups? Having detected them, how can we counteract these problems? The first question can benefit from an examination of techniques developed to search for communities or more generally dense subgraphs from an underlying network. As for the second question, a natural approach for countering filter bubbles is to launch some kind of counter-campaign to try and enhance the balance in users’ exposure to viewpoints. Countermeasures for misinformation propagating through a network, on the other hand, are manifold and can depend on who is planning the countermeasure. For example, the network host can intervene and take steps to limit the propagation of misinformation, but these actions come with a cost. Besides the political sensitivity and cost of limiting freedom of expression, what if the intervention was by mistake done on genuine information? As another example, a third party interested in countering the propagation of misinformation may launch a counter-campaign. Interestingly, some of the ideas behind designing such campaigns have strong connections to a classic problem called Influence Maximization, studied in a very different context, driven by different applications like viral marketing, infection containment, and revenue or welfare maximization. In this talk, we will examine research on detecting dense subgraphs as well as competitive influence maximization and discuss how that can inspire techniques for addressing the two problems above.
Bio
Laks V.S. Lakshmanan is a Professor in the
Computer Science Department at the University
of British Columbia, Vancouver, Canada.
His research interests span a wide spectrum of
topics within Database Management and Mining
as well as applications of database technology
for recommender systems, social media,
machine learning, and natural language processing. He was named
ACM Distinguished Scientist in 2016. He received ACM SIGMOD
2020 Research Highlights Award and a Best Paper Award in IEEE
Conference on Data Science and Advanced Analytics (DSAA) 2018.
Keynote Speaker 3: Chris Ré (Stanford)
Is Data Management the Beating Heart of AI Systems?
Abstract
The era of artificially intelligent systems is here, and the breadth and speed of the change has been breathtaking. Less than a decade ago, Go ogle did not use AI or machine learning in its products. Today, we interact with a wide range of AI -powered systems when we search the web, write an email, watch a video, or make a purchase. AI has spread beyond the web as well, we use it every time we toutouch phones, open our laptops, turn on our television, enter (or request) a car, and many more interactions. What’s perhaps more remarkable than the ubiquity of these systems is how they operate: they optimize a small number of machine learning architectures on increasingly vast and diverse quantities of data. As a result, data management is at the heart of this new breed of systems. This talk explores the idea that the next phase of AI systems might have historic parallels with the growth of relational data management. In this talk, I survey some of the exciting recent developments that have affected my own work in industry and research building systems to support AI applications. My main message is that it is an incredibly exciting time for researchers and industrial practitioners from the data management community to bring their expertise to the next generation of AI systems. To that end, I plan to focus on three larger trends through vignettes based on my own work:
- Data-Centric AI is a broad and exciting research agenda with a burgeoning community. Via our work on Snorkel, the first data -centric AI platform, we learned that treating data management as a firstfirst-class machine learning problem can change how AI applications are built and maintained.
- Foundation models are large pretrained models that have obtained statestate-of -thethe-art qu ality in natural language processing, vision, speech, and other areas. The underlying techniques allow us to integrate multiple modalities ( images, structured data) in ways that would have been difficult to imagine even a few years ago. Their natural language interfaces have altered who and how easily developers can use them. I plan to my own journey using these models at Apple with structured data the data management challenges.
- Robustness is a hot topic in the AI community that is increasingly impor tant, e.g., benchmarks like WILDS. I plan to describe our own work on a specific form of robustness called hidden stratificationstratification—a kind of schema error. These results arose in AI for medical imaging, in which systems unintentionally overstated their accuraaccuracy due to nonobvious subpopulations in the dataset. I will also report on the growing body of techniques in AI to help automatically improve the robustness in the face of hidden stratification and distribution shift.
Bio
Christopher Ré is an associate
professor at Stanford in the
AI lab associated with the
Machine Learning Grouproup— but no sensible community
wou ld claim him as their
own . His recent work is to unde rstand how computing
systems will cha nge due to machine learning along with
a continuing petulant drive to work on math problems. Due
to work with amazing
collab orators and brilliant students, ideas and software from our work have been used in a variety of widely us ed productproducts in Google Ads, YouTu be, App le productproducts, dozedozens of la rge and small
enterpriseenterprises, and human itarian efforts.