All Segments
Anca Dragan: Human-Robot Interaction and Reward Engineering | Lex Fridman Podcast #81
48 segments available
Anca Dragan is a professor at Berkeley, working on human-robot interaction -- algorithms that look beyond the robot's function in isolation, and generate robot behavior that accounts for interaction and coordination with human beings. This episode is presented by Cash App. Download it & use code "LexPodcast": Cash App (App Store): https://apple.co/2sPrUHe Cash App (Google Play): https://bit.ly/2MlvP5w PODCAST INFO: Podcast website: https://lexfridman.com/podcast Apple Podcasts: https://apple.co/2lwqZIr Spotify: https://spoti.fi/2nEwCF8 RSS: https://lexfridman.com/feed/podcast/ Full episodes playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4 Clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41 EPISODE LINKS: Anca's Twitter: https://twitter.com/ancadianadragan Anca's Website: https://people.eecs.berkeley.edu/~anca/ OUTLINE: 0:00 - Introduction 2:26 - Interest in robotics 5:32 - Computer science 7:32 - Favorite robot 13:25 - How difficult is human-robot interaction? 32:01 - HRI application domains 34:24 - Optimizing the beliefs of humans 45:59 - Difficulty of driving when humans are involved 1:05:02 - Semi-autonomous driving 1:10:39 - How do we specify good rewards? 1:17:30 - Leaked information from human behavior 1:21:59 - Three laws of robotics 1:26:31 - Book recommendation 1:29:02 - If a doctor gave you 5 years to live... 1:32:48 - Small act of kindness 1:34:31 - Meaning of life CONNECT: - Subscribe to this YouTube channel - Twitter: https://twitter.com/lexfridman - LinkedIn: https://www.linkedin.com/in/lexfridman - Facebook: https://www.facebook.com/LexFridmanPage - Instagram: https://www.instagram.com/lexfridman - Medium: https://medium.com/@lexfridman - Support on Patreon: https://www.patreon.com/lexfridman
Segments Timeline
The Passion for Robotics
Anca Dragan shares her journey into robotics, highlighting her early interest in programming and how it evolved into a passion for human-robot interaction. She discusses her academic path, including her time at Carnegie Mellon, and the transformative experiences that solidified her love for robotics, particularly self-driving cars and expressive robots like Spot Mini.
"the following is a conversation with ANCA Jorgen a professor of Berkeley working on human robot interaction algorithms and looked beyond the robots function in isolation and generate robot behavior th..."
The Magic of Self-Driving Cars
Anca recounts her first experience with a self-driving car during her PhD at Berkeley, describing it as a transformative moment. She reflects on the decision-making capabilities of the car and how it inspired her to explore the potential of robotics beyond manipulation, emphasizing the emotional connection humans can have with robots.
"an organization that is helping to advanced robotics and STEM education for young people around the world and now here's my conversation with Enka Droog on when did you first fall in love with robotic..."
WALL-E: A Personal Connection
Anca discusses her favorite fictional robot, WALL-E, and its expressive design that captivates audiences. She shares a personal story about how her husband proposed to her using a custom-built WALL-E robot, illustrating the deep emotional connections that can be formed through robotics and animation.
"like it was there a robot that truly sort of inspired you that was I remember that experience very viscerally riding in that car and being just wowed I I had the they gave us a sticker that said I wro..."
Expressivity in Robotics
Anca explores the challenges of creating robots that can express emotions and intentions. She discusses the importance of expressivity in human-robot interaction and the difficulties in programming robots to autonomously convey their internal states, emphasizing the need for robots to adapt to human emotions.
"by building a Wally and he actuated it so it's seven degrees of freedom including the lens thing and it kind of came in and it had the he made it have like a you know the belly box opening thing so it..."
Understanding Human Complexity
Anca addresses the complexities of human behavior and how they impact human-robot interaction. She discusses the challenges of modeling human emotions and preferences, highlighting the need for robots to understand and adapt to human states to effectively interact and fulfill user needs.
"are kids are being rude to Alexa because they can be rude to it and it doesn't really get angry right it doesn't reply it in any way it just says the same thing so I think there's at least for that fo..."
The Dual Challenges of HRI
Anca outlines two major challenges in human-robot interaction: the coexistence of robots and humans in shared spaces and the need for robots to optimize their actions based on human preferences. She emphasizes the importance of understanding user needs and the complexities involved in programming robots to meet those needs.
"the problem of human robot interaction yeah should we talk about what the problem of human robot is yeah this is what I mean talk about how that yeah so and by the way I'm gonna talk about this very p..."
Anticipating Human Behavior
In this segment, Dragan delves into the dual tasks of predicting human actions and satisfying their preferences. She discusses how understanding human behavior is essential for robots to effectively anticipate future actions and adapt their responses accordingly.
"preferences be able to interactively adjust to understand what the human is so really boys ought to be understand the humans in order to interact with them in order to please them right so why is this..."
Inverse Reinforcement Learning Explained
Dragan introduces the concept of inverse reinforcement learning, which allows robots to infer human preferences from observed behavior. She explains how this method can help robots learn optimal actions based on human demonstrations, enhancing their ability to assist in tasks like driving or household chores.
"relationship between them now how complicated do these models need to be in order to be able to understand what people want so we've gotten a long way in robotics with something called inverse reinfor..."
Challenges in Complex Tasks
Dragan discusses the limitations of current models in robotics, particularly in complex tasks like controlling a lunar lander. She highlights the difficulties in accurately predicting human behavior and the need for more sophisticated models to handle the noise and unpredictability inherent in human actions.
"press back in the forties fine women mortgage time or like okay people are making choices by maximizing utility go and then in the late 50s we had loose and Shephard come in and say people are a littl..."
The Messiness of Human Behavior
In this segment, Dragan reflects on the challenges posed by the messy and often irrational nature of human behavior. She emphasizes the importance of understanding the underlying assumptions and constraints that guide human decisions, which can help robots better interpret and respond to human actions.
"lunar lander so just I don't know if you know this Atari game it's called lunar lander it's it's really hard people really suck at landing the same mostly they just crash it left and right okay so thi..."
Collaborative Human-Robot Interaction
Dragan explores the idea of human-robot collaboration, where both agents actively participate in the interaction. She discusses how robots can gather information through their actions, allowing them to better understand human preferences and improve their assistance.
"make sure that you're telling them the right thing to do what is that thing well read it in human behavior so if I just sit here quietly and try to understand something about you but listening to you ..."
Inferring Driving Styles
Dragan illustrates how robots can infer human driving styles by observing their behavior in traffic. She explains the importance of understanding whether a driver is aggressive or defensive, and how robots can adapt their actions based on these insights to navigate safely around human-driven vehicles.
"there's these information gathering actions that the robot can take to sort of solicit responses that are actually informative so for instance this is not for the purpose of assisting people but with ..."
Rethinking Human Rationality
In this thought-provoking segment, Dragan challenges the notion that humans are inherently irrational. She suggests that by understanding the different assumptions and constraints under which humans operate, robots can better interpret human behavior and improve their interactions.
"cool things where it kind of not just towards you and then sees if you're gonna slow down or not then if you slow down it sort of updates its model of you and says oh okay you're more on the defensive..."
Intuitive Physics and Human Actions
Dragan discusses the concept of intuitive physics and how it relates to human actions. She shares insights from cognitive science on how humans may simplify complex physical models, and how robots can leverage this understanding to enhance their performance in tasks requiring human collaboration.
"we think about rationality and bliss the assumption is or they're rational under all the same assumptions and constraints as the robot right what if this is the state of the world that's what they kno..."
Modeling Human Intentions
In this segment, Dragan explores the complexities of modeling human intentions in robotic systems. She explains how understanding human behavior can enhance robot performance, particularly in dynamic environments. The discussion touches on the challenges of accurately predicting human actions and the implications for designing more effective human-robot interactions.
"that instead of where they were actually sending you in the real world and I kid you not it word peopled landed there the damn thing and you know in between the two flags and and and all that so it's ..."
Game Theory in Human-Robot Interaction
Dragan introduces the concept of game theory as it applies to human-robot interaction. She discusses how both robots and humans have their own objectives and how their actions influence each other. This segment delves into the negotiation aspect of interactions, emphasizing the need for robots to anticipate human behavior to achieve successful coordination.
"you mentioned recommender systems what kind of and we're talking about human robot interaction kind of problem spaces are you thinking about so is it robots like wheeled robots of autonomous vehicles ..."
Optimizing Robot Perception
This segment focuses on the optimization of robot perception to enhance human-robot interactions. Dragan explains how robots can be designed to better understand human beliefs and intentions, thereby improving their responses. The conversation highlights the importance of making robots more informative and less confusing to users, ultimately leading to more effective collaboration.
"talking about around well you know how do you model humans it turns out that a lot of systems need to quote benefit from a better understanding of how human behavior relates to what people want and ne..."
The Role of Attitude in Robots
Dragan and Fridman discuss the necessity of robots having a certain attitude to foster respect and effective interaction with humans. They explore how assertiveness and vulnerability can influence human perceptions of robots. This segment raises important questions about the emotional dynamics between humans and robots, suggesting that a robot's expressiveness could enhance user experience.
"kids people see Alexa as a in a certain way is there some value in trying to also optimize how people see you as a robot is that it or is that a little too far and away from the specifics of what we c..."
Human Nature and Robot Interaction
In this thought-provoking segment, Dragan reflects on human nature and its implications for robot interaction. She expresses concerns about how humans may treat robots as 'the other' and discusses the potential for robots to be bullied or disrespected. The conversation emphasizes the need for robots to be designed in a way that encourages positive interactions and mutual respect.
"just Bayes rule basically okay first has a belief they see an action they make some assumptions about how the robot generates its actions presumably is being rational because robots are rational see r..."
Navigating Human-Robot Dynamics
Dragan elaborates on the dynamics of human-robot interactions, particularly in scenarios like autonomous driving. She discusses how robots must navigate complex social cues and human behaviors to ensure safety and efficiency. This segment highlights the intricate balance between human intentions and robotic actions, underscoring the importance of understanding this interplay for successful human-robot collaboration.
"passive-aggressively I don't it seems like we humans yet need that dude what are you is there something you have thoughts on this one is one is it it's we respond to you know someone being assertive b..."
The Underactuated Nature of Interaction
In this concluding segment, Dragan ties together the concepts of underactuation in robotics and human behavior. She discusses how both robots and humans operate under constraints that limit their actions and how understanding these limitations can improve interactions. The segment emphasizes the need for a nuanced approach to designing robots that can effectively engage with humans in various contexts.
"talking about different equilibria of a game the very basic way to solve coordination is to just make predictions about what people will do and then stay out of their way and that's hard for the reaso..."
The Game of Chicken: Crossing Streets
In this segment, Dragan elaborates on the risks pedestrians take when crossing streets, likening it to a game of chicken. She discusses how ambiguity in right-of-way rules and the personalities of intersections affect human decision-making, revealing the intricate dance between humans and vehicles.
"just also like the the poetry of calling human robot interaction and under actuated robotics problem and y'all so much sort of nudging it seems that there and I don't know I think about this a lot in ..."
Confidence in Movement: The Power of Nonverbal Cues
Dragan explains the concept of 'civil inattention' and how confidence in one's actions can influence the behavior of others in shared spaces. She shares insights from a runner's perspective on how making eye contact can alter pedestrian dynamics, emphasizing the importance of nonverbal communication in human interactions.
"and so on I mean and this nudging idea I don't you know it seems that people don't even nudge they just aggressively take make a decision somebody there's a runner that gave me this advice I sometimes..."
Driving as a Ballistics Problem
The conversation shifts to the challenges of autonomous driving, with Dragan discussing how driving fundamentally resembles a ballistics problem. She highlights the complexities introduced by human behavior, suggesting that the presence of humans makes driving significantly more difficult than simple object collision avoidance.
"conversation with Jim Keller who is a sort of this legendary chip or chip architect but he also let the autopilot in for a while and his intuition that driving is fundamentally still like a ballistics..."
Perception vs. Human Interaction in Autonomous Driving
Dragan and Fridman explore the perception challenges in autonomous driving, particularly in urban environments. They discuss how advancements in perception technology have improved but still face hurdles, especially when human elements are involved, complicating the driving problem.
"on this imagine downtown San Francisco yeah yeah it's crazy busy everything okay now take all the humans out no pedestrians no human driven vehicles no cyclists no people and little skill electric sco..."
The Role of Lidar in Autonomous Vehicles
In this segment, Dragan shares her thoughts on the use of Lidar in autonomous vehicles. She discusses the balance between relying on proven technologies and the need for innovation, emphasizing the importance of safety in critical systems while acknowledging the limitations of current approaches.
"there any perception is hard but yeah it's we've made a lot of progress on the perceptions on how to undermine the difficulty of the problem I think everything about robotics is really difficult of co..."
The Trade-offs of Safety and Innovation
Dragan reflects on the trade-offs between safety and innovation in robotics, particularly in autonomous driving. She discusses the tension between sticking to proven methods and exploring new technologies, highlighting the challenges faced in safety-critical systems.
"way I think it's kind of irresponsible to not use lighter that's just my personal opinion depending on your use case but I think like you know if you if you have the opportunity to use light are good ..."
The Challenge of Human Element in Driving
The discussion continues on the difficulties posed by human interactions in driving scenarios. Dragan emphasizes the unpredictability of human behavior and the need for breakthroughs in understanding human-robot interaction to improve autonomous driving systems.
"for the invention of the things that will work tomorrow right they get it's kind of true in the sense that if we you know that we want to stick to the conference and you see this in academic and setti..."
Learning from Data: The Future of Autonomous Systems
Dragan discusses the role of machine learning and reinforcement learning in developing autonomous systems. She highlights the importance of combining learned models with human expertise to create robust systems capable of navigating complex environments.
"think it in a sense it really depends I think that you know we were talking about how well look it's really hard because I'm just know people do is hard and on top of that playing the game is hard but..."
Simulation's Role in Human-Robot Interaction
In this segment, Dragan explores the potential of simulation in human-robot interaction. She discusses how simulations can help train models and prepare robots for real-world interactions, emphasizing the need for accurate representations of human behavior.
"were they're going fattening it felt fast because you're like freaking out at first I was this is my first experience but it's going like the speed limit right 30 40 whatever it is and there's humans ..."
Out of Distribution: The Challenge of Human Behavior
Dragan concludes by addressing the challenges of modeling human behavior in robotics. She discusses the concept of 'out of distribution' scenarios and the difficulties in predicting human actions that deviate from learned patterns, underscoring the complexity of human-robot interactions.
"and more about new challenges that get you get faced when how much is this problem do you think can be learned through in turn this is the success of machine learning and reinforcement learning how mu..."
Understanding Human Behavior Beyond Data
In this segment, Dragan explores the complexities of human reasoning in novel situations and the limitations of data-driven models. She argues that humans can anticipate outcomes in unfamiliar scenarios, raising questions about how robots can effectively learn and adapt to human behavior without extensive prior data.
"where that model that you've built of the human completely sucks because out of distribution you have no idea right there's if you think of all the possible policies and then you take only the ones th..."
The Challenge of Common Sense Reasoning
Dragan and Fridman delve into the necessity of common sense reasoning for effective human-robot interaction. They discuss the importance of encoding fundamental human concepts, such as the fear of death, into robotic systems to enhance their understanding of human motivations and actions.
"open question do you think so that one of the dreams of artificial intelligence was to solve common sense reasoning whatever the heck that means do you think something like common sense reasoning has ..."
Semi-Autonomous Driving: A New Paradigm
This segment focuses on the dynamics of semi-autonomous driving systems and the role of human supervision. Dragan cautions against the assumption that human supervisors can maintain the same level of safety as active drivers, highlighting the differences in engagement and decision-making between the two roles.
"that stuff let me ask sort of an another small side of this that we've been talking about the pure autonomous driving problem but there's also relatively successful systems already deployed out there ..."
Empowering Human Supervisors in Autonomous Systems
Dragan discusses the potential benefits of designing semi-autonomous systems that empower human supervisors rather than merely relying on them to take control. She emphasizes the importance of keeping humans engaged and how this can lead to better outcomes in human-robot interactions.
"robot interaction problem I think what we have to be careful about there is to not me it seems like some of these systems not all are making this underlying assumption that if so I'm a driver and I'm ..."
Designing Effective Reward Functions
In this segment, Dragan addresses the complexities of designing reward functions for robots. She explains the challenges of ensuring that specified rewards lead to desired behaviors in various situations, emphasizing the need for careful consideration of all possible scenarios to avoid unintended consequences.
"basically not join but be ready to sort of take over so one of the interesting things we'll be talking about is the rewards that they seem to be fundamental to the way robots behaves so broadly speaki..."
Unintended Consequences in Robotics
Dragan and Fridman conclude by discussing the concept of unintended consequences in robotics, referencing Goodhart's Law. They explore how metrics can lead to behaviors that are not aligned with the original intentions, highlighting the broader implications for both robotics and human decision-making.
"optimized will work well in every possible situation so so you're kind of referring to unintended consequences or just in general any kind of suboptimal behavior that emerges outside of the things you..."
Collaborative Reward Design
In this segment, Dragan explores the idea of collaboration between robots and their human designers in the context of reward specification. She discusses how the interaction between the robot and the programmer can lead to better understanding and optimization of the robot's behavior, moving beyond simple rule-based systems to a more nuanced approach to human-robot interaction.
"this thing called good hearts law which is you start a metric for an organization and the moment it becomes on target that people actually optimize for it's no longer a good metric well what's it call..."
Leaked Information in Human-Robot Interaction
Dragan delves into the concept of 'leaked information' in human-robot interactions, where human behavior unintentionally communicates preferences to robots. She discusses how robots can learn from physical interactions and the implications of these signals for optimizing their behavior. This segment highlights the subtle ways in which humans can influence robot learning through their actions.
"between the robot and the expert programmer a roboticist designer in charge of actually specifying what the hectic wants should do a task for this professor that's so cool like collaborating on the re..."
Understanding Human Preferences
This segment focuses on how robots can interpret human preferences based on physical interactions. Dragan explains the importance of understanding the context of human behavior and how robots can adapt their actions based on feedback from users. She emphasizes the need for robots to learn from their environment and human interactions to better align with human expectations.
"again it looks the poetry of leaked information you mentioned humans leaked information about what they want you know leaked reward signal for the for the robot so how do we detect these leaks yeah wh..."
The Complexity of Human Signals
Dragan discusses the challenges of interpreting human signals in robotics, particularly in semi-autonomous systems. She highlights the difficulty of understanding why humans disengage from automated systems and how this information can inform robot behavior. This segment underscores the need for robots to adapt and learn from human actions to improve their functionality.
"to what the robot is currently doing and this is what we call physical human robot interaction and now there's a lot of there's a lot of interesting work on how do you respond to physical human robot ..."
The Three Laws of Robotics Revisited
In this thought-provoking segment, Dragan reflects on Asimov's Three Laws of Robotics and their relevance in modern robotics. She discusses the limitations of these laws and the need for a more dynamic approach to robot behavior that considers human context and preferences. This segment invites listeners to think critically about the ethical implications of robotics in society.
"is that they want so you can learn that people want you to stay further away from them now you're right that there might be many things that explain just at one signal that you might need much more da..."
Interpreting Human Intentions
Dragan emphasizes the importance of interpreting human intentions in the design of robotic systems. She discusses how robots should not take human commands literally but rather understand the context behind them. This segment highlights the ongoing challenge of aligning robotic behavior with human expectations and the potential for robots to learn from their interactions.
"how do we get that dance right I know the three laws might be a silly notion but do you do you think about like what Universal reward functions that might be that we should enforce on the robots of th..."
The Impact of AI on Human Behavior
In this concluding segment, Dragan shares her personal journey with AI and its influence on her understanding of human behavior. She reflects on the transformative power of AI in simplifying complex decision-making processes and its implications for future robotics. This segment encapsulates the excitement and challenges of integrating AI into human-robot interactions.
"of what people want every time the person corrected maybe they demonstrate maybe they stopped hopefully not right one really really crazy one is the environment itself like our world you don't it's no..."
The Value of Existence
Anca Dragan discusses the beauty of life and the sadness of its finiteness. She reflects on how the awareness of mortality enhances our appreciation for life's mundane moments and suggests that this understanding should influence the design of AI and robotics, particularly in how they interact with humans.
"questions but and the silly ones if a doctor gave you five years to live sort of emphasizing the finiteness of our existence what would you try to accomplish it's like my biggest nightmare by the way ..."
Small Acts of Kindness
Anca shares a touching story about a high school teacher, Nicole McConnell, who provided her with free tutoring and guidance, significantly impacting her educational journey. This act of kindness exemplifies the importance of support and mentorship in shaping one's future.
"because otherwise people just like wall-e so so I think the finance helps but but yeah it's just um you know I don't I don't I'm not a religious person I don't think that there's something after and s..."
The Meaning of Life
In a deep philosophical exchange, Anca Dragan and Lex Fridman explore the meaning of life and the human experience. They discuss the insignificance of individual existence in the grand scheme of the universe and the importance of making a positive impact on local communities and relationships.
"pretty joyous moment because I don't know that i would change mine listen I'm what I'm I'm you know I'm trying to make some contributions to how we understand human AI interaction I don't think I woul..."
Closing Thoughts
As the conversation wraps up, Lex Fridman expresses gratitude for Anca Dragan's insights and reflects on the importance of questioning our assumptions about the world. He leaves the audience with a quote from Isaac Asimov, encouraging listeners to remain open-minded and receptive to new ideas.
"they're our friends our family our local communities and just try to be there for other humans cuz I just everything beyond that seems ridiculous I mean are you like how do you make sense of these mul..."