Welcome Keynote - The Practice of Machine Learning
Peter Norvig, Google Research Director
Peter Norvig is a Director of Research at Google; previously he directed Google's core search algorithms group. He is co-author of Artificial Intelligence: A Modern Approach, the leading textbook in the field, and co-teacher of an Artificial Intelligence class that signed up 160,000 students, helping to kick off the current round of massive open online classes. He is a Fellow of AAAI, ACM, the California Academy of Science and the American Academy of Arts & Sciences.
Crowdsourcing Images for Global Diversity
Peggy Chi, Google Research Scientist
Abstract: Crowdsourcing enables human workers to perform designated tasks unbounded by time and location. As mobile devices and embedded cameras have become widely available, we deployed an image capture task globally for more geographically diverse images. Via our micro-crowdsourcing mobile application, users capture images of surrounding subjects, tag with keywords, and can choose to open source their work. We open-sourced 478,000 images collected from worldwide users as a dataset "Open Images Extended'' that aims to add global diversity to imagery training data. We describe our approach and workers' feedback through survey responses from 171 global contributors to this task.
Navigating Academic and Industry Careers
Lauren Wilcox, Google UX Researcher
Nicolas Papernot, Google Research Scientist
Abstract: You're at an exciting time of your life. This panel will go over some of the pathways to an academic or industry career, including some tips on how to navigate the job market. We'll tease out some of the differences between academia and industry as an environment for contributing to the research community. The talk will be followed by a panel discussion.
Deep Learning for Tackling Real-World Problems
Jeff Dean, Google Senior Fellow & SVP Google Research
Abstract: In this talk, I'll look at how recent advances based on deep learning have made significant strides in fundamental areas such as computer vision, speech recognition, language understanding and translation. Given these advances, I'll then look at how deep learning can now help us tackle some of the major challenges in the world, such as improving access to healthcare and creating new tools for scientific discovery. Finally, I'll touch on how deep learning is changing the way in which we think about building computational hardware, creating a resurgence in new and interesting computer architecture work to design systems that target and dramatically accelerate deep learning workloads.
Machine Learning Fairness
Vinodkumar Prabhakaran, Google Research Scientist
Abstract: As machine learning techniques are increasingly being used in various day-to-day applications, there is growing awareness that the decisions we as researchers and developers make about our data, methods, and algorithms have immense impact in shaping our social lives. In this talk, I will outline the growing body of research on ethical implications of machine learning technology, especially around questions about fairness and accountability of the models we build and deploy into the world. I will focus specifically on natural language processing (NLP) techniques, and discuss ways in which machine learned NLP models may reflect, propagate, and sometimes amplify social stereotypes about people, potentially harming already marginalized groups. I will also briefly discuss various ways to address these issues, both through mitigation strategies and through increased transparency.
The Role of Generative Models in Music, Art and other Media
Douglas Eck, Google Principal Scientist
Abstract: I'll talk about recent progress on a project from the Google Brain team called Magenta (g.co/magenta). Magenta is an open source research project exploring the role of machine learning as a tool in the creative process. I'll relate this to work done on the Brain team in the area of generative models for text, images and video. My talk will consist of a high-level overview of work we've published in domains like music composition, audio generation, drawing and text generation. I'll also address some related non-technical questions such as: What is the relationship between AI and artistic creation? and What are the societal implications of using generative models for communication?
Self-Driving Networks: Is AI the Answer to Automating Network Management?
Jeffrey Mogul, Google Principal Software Engineer
Abstract: It's increasingly hard for humans to manage computer networks against requirements for high reliability and rapid evolution. Automation seems to be the answer, but is AI the right way to automate network management? And do self-driving cars offer a useful or misleading analogy?
Studying High Risk Users
Sunny Consolvo, User Experience Researcher
Abstract: This talk will present the results of two exploratory studies of people’s privacy- and security-related practices. In the first study, we report on the experiences of financially vulnerable users and discuss challenges participants faced that tended to impact their online safety. In the second study, we explore the digital privacy and security motivations, practices, and challenges of survivors of intimate partner abuse.
Internships Q&A (Optional)
Allison Kemmerling, Google Intern Recruiter + Staffing Partner
Abstract: Learn about summer internships for PhD Fellows. Allison Kemmerling from Intern recruiting will provide an overview of the intern process specifically tailored for Google PhD Fellows, and will answer your questions. Note: This session is optional, but encouraged if you are interested in a Google internship.
Erik Vee, Google Software Engineer
Abstract: The area of online algorithms has a long and successful history. But its set-up -- where we assume we know nothing about the future and compare ourselves to the best algorithm that knows everything about what's to come -- is usually too pessimistic. In practice, we often have predictions that limit inputs to a smaller set of possibilities.
To better understand this gap, we propose "semi-online" algorithms, which is a simple and robust extension to the classic online setting. By assuming we know something about the future, we get results that naturally interpolate between fully online algorithms and fully offline algorithms. We study several classic problems, including matching, ski-rental, and caching.
Joint work with Ravi Kumar, Manish Purohit, Aaron Schild, and Zoya Svitkina.
Accelerating Your Research with TensorFlow Research Cloud
Jonathan Caton, Google Product Support Manager
Abstract: The TensorFlow Research Cloud (TFRC) program provides free access to Cloud TPUs to researchers worldwide. In this talk we will briefly describe the hardware used to make this possible, give a few examples of the work being supported by our program, and inform attendees how they can take advantage of Cloud TPUs to further their research.