Service & Outreach

Systems that can learn interactively from their end-users are quickly becoming widespread in real-world applications. Humans offer a wealth of implicit information (such as multimodal cues in the form of natural language, speech, eye movements, facial expressions, gestures etc.) which interactive learning algorithms can leverage during the process of human-machine interaction to create a grounding for human intent, and thereby better assist end-users. A closed-loop sequential decision-making domain offers unique challenges when learning from humans -– (1) the data distribution may be influenced by the choices of the algorithm itself, and thus interactive ML algorithms need to adaptively learn from human feedback, (2) the nature of the environment itself changes rapidly, (3) humans may express their intent in various forms of feedback amenable to naturalistic real-world settings, going beyond tagged rewards or demonstrations. By organizing this workshop, we attempt to bring together interdisciplinary experts in interactive machine learning, reinforcement learning, human-computer interaction, cognitive science, and robotics to explore and foster discussions on such challenges.

I am lead organizer and HRI representative for the DREAM/CPAR Seminar at UC Berkeley, a weekly seminar hosting professors/professionals in robotics, control, and human-centered autonomy.

Robots deployed in the real world will interact with many different humans to perform many different tasks in their lifetime, which makes it difficult (perhaps even impossible) for designers to specify all the aspects that might matter ahead of time. Instead, robots can extract these aspects implicitly when they learn to perform new tasks from their users' input. The challenge is that this often results in representations which pick up on spurious correlations in the data and fail to capture the human’s representation of what matters for the task, resulting in behaviors that do not generalize to new scenarios. In this workshop, we are interested in exploring ways in which robots can align their representations with those of the humans they interact with so that they can more effectively learn from their input. By bringing together experts from representation learning, human-robot interaction, and cognitive science, we believe we can foster an environment where we can exchange ideas for how the robot learning community can best benefit from learning representations from human input and vice-versa, and how the HRI community can best direct their efforts towards discovering more effective human-robot teaching strategies.

Social intelligence is at the core of both human and artificial intelligence. From a young age, humans can understand, interact, collaborate, and communicate with each other. Most of what we learn is taught by others, or learned in a social context. Thus, a truly intelligent AI agent should be able to understand and work with humans as well as other AI agents. This workshop focuses on the challenges and developments in building AI systems equipped with social intelligence, and leverages theories and insights from studies of human social intelligence for achieving such goals. In particular, the workshop will explore what it would take for machines to: a) understand the behaviors and mental states of humans, and b) engage in rich and complex interactions with humans. We brought together experts from cognitive science and developmental psychology to better understand the principles and origins of human social intelligence, and experts from AI and Robotics, to discuss how to engineer socially intelligent artificial agents, and how these paradigms can be deployed in both virtual and real scenarios.

As robots and other intelligent agents increasingly address complex problems in unstructured settings, programming their behavior is becoming more laborious and expensive, even for domain experts. Frequently, it is easier to demonstrate a desired behavior rather than to manually engineer it. Imitation learning seeks to enable the learning of behaviors from fast and natural inputs such as task demonstrations and interactive corrections. However, human generated time-series data is often difficult to interpret, requiring the ability to segment activities and behaviors, understand context, and generalize from a small number of examples. Recent advances in imitation learning algorithms for both behavior cloning and inverse reinforcement learning—especially methods based on training deep neural networks—have enabled robots to learn a wide range of tasks from humans with relaxed assumptions. However, real-world robotics tasks still pose challenges for many of these algorithms, and in this workshop we bring together AI and robotics experts to discuss the greatest challenges facing imitation learning for robotics.

I co-organized the SemiAutonomous seminar at UC Berkeley, a weekly seminar series well established over the past few years that brings together students and professors (internal and external to Berkeley) from machine learning, control theory, and robotics.

GiE is a week-long non-residential summer camp where female middle school students can learn engineering in an engaging, hands-on environment. I co-organized one of the Self-Driving Cars workshops, where I got to teach the girls about sensing, planning, and control in autonomous driving, and work together on experimenting with an Evo robot.

AI4ALL is a 5-day non-residential summer program, that exposes current Bay Area 9th to 11th grade students from underrepresented groups to the exciting field of AI. I mentored a team of students as they learned to train a deep reinforcement learning agent in MuJoCo.

As a BAIR mentor the past few years, I have been meeting up regularly with underrepresented undergraduate students and mentoring them in research and career planning. I helped one student find a robotics summer internship at Dishcraft, a research position in Ruzena Bajcsy's lab, and a Master's position in Shankar Sastry's group.

As part of WiCSE, I was paired with and mentored early-stage female PhD students in career planning and navigating life at UC Berkeley.

As part of WiSE, I mentored high school girls from the Greater Boston area during monthly sessions designed to introduce them to engineering at MIT.

As part of ESP, I taught courses on "Water Security in Asia", "Introduction to Probability", and "Group Theory" to middle school students in the New England region.