Socially Assistive Robotics: A Possible Future of AI for Care | Maja Matarić | TEDxPaloAlto

[Music] AI it's everywhere right it's on every billboard that I Past coming here um it's in every ad and it's in every vision of the future um you're either in it you're trying to get into it or you're worried about it or maybe all three well I got into AI a long time ago when I was a college student that was a long time ago um because I'm really interested in the way the brain creates behavior in the world around us and by that I don't mean just words but actual Behavior how do we move about how do we relate to one another and to study that I got into robotics because robotics is AI in the real world and so I went on to grad school to MIT to the AI lab there and I could spend quite a few years playing in the space of ideas about Ai and I was particularly interested in biological systems because biological systems are the beginning of intelligence so how for example does a rat navigate so well and can we learn something from that and then make better robots so I made my first robot Toto um that navigated just like a rat but didn't exactly look like one but that's okay then I was also fascinated by things like flocks of birds that create these amazing 3D structures in the sky and no one's in charge how does that happen and could we use that some how to coordinate teams of robots and I had a team of robots I was very lucky or not depending on how you look at it I had my nerd herd it was like a bunch of little toaster likee robots 20 of them and uh it was basically an entire PhD dissertation to get these 20 little toasters to navigate and her and even learn from each other they learned from one another I learned a lot eventually I finished my dissertation I went off to be a professor so that I could play in the space of AI ideas with amazing graduate students and under undergraduates and I did that for a while and it was wonderful and really satisfying and then boom something completely amazing happened we would call it a phase transition um like an epic moment in my life and I think most of you can relate I became a mom actually three times definitely my best work by far um a lot of learning happening there and it was an amazing thing because suddenly everything that I had been doing in my work no longer meant as much um I was no longer interested enough to just play with ideas because I had someone much cuter and much more interesting to play with um but now you know what was I going to do and this new interesting person pretty soon asked me mama why do you make robots H why you know I hadn't really thought about that for a long time right I got on a kind of a an a rail I you know went to school went to grad school became a professor why why am I doing this um and I I could say to my sweetie well you know Ai and Robotics in particular is intelligence in the real world nerdy or I could say you know our papers are really well published and well cited and we get a lot of grant money boring kids don't care about stuff like that and you know what kids are really smart kids have good actually great values because kids are always asking why and we maybe forget to ask that and so I thought about it how can I impress my kid how can I impress all the kids um and I knew there was one really good answer and that answer is Mama makes robots that help people so suddenly I knew what I needed to do this was it this was my purpose I had to somehow bring Ai and Robotics together in a way that it could help people pretty soon because you know kids are impatient and they grow up and you know I'm going to grow up and so you know it had to happen quickly so suddenly I had this Mission this purpose um and it seemed like it would be really hard like what could I do turns out it wasn't hard at all to figure out what to do and the thing that I realized we should work on is Bridging the care Gap so what is this care Gap that I like to talk about so the care Gap is the gap between the needs of literally millions of people in the kinds of care that they need to receive and the care that is actually available to them and examples are everywhere everywhere you look so this was the easy part for example um one in six children in the US has some kind of a developmental disorder if you can diagnose that early and intervene even with the slightest of interventions you can change the course of their entire life for the better autism one in 45 children today is diagnosed with autism the number is has quadrupled over the last 20 years and it continues to grow children with autism don't even have the luxury of playing with other children and learning social skills because they often don't have peers who will play with them and now even though it's shortly after the covid pandemic the actual big pandemic that we have is anxiety and depression and especially you see it in adolescence because it affects the Adolescent brain in particular in strong ways but he goes across the ages span adults and elderly so we have this epic ch challenge also as we get older I'm sorry to say stroke is on the rise and if you've had a stroke and survived um you basically have to retrain your brain to relearn to do things that you learned when you were a little kid like you know walking and reaching for something and to do that it's it's a slog you have to basically do these boring exercises for hours and hours per day and you have to keep failing that's really hard what's going to keep you motivated to do that and then finally as we really get older um unfortunately people get more and more isolated lonely and depressed um and they lose the the will and the desire to keep going so that is the care Gap so how can we help millions of people while keeping in mind that while we can use AI for this in the end we're profoundly human and for us as humans human care is the best care so we have this in inherently human need to be with others like ourselves so how do we create technology that can supplement human care and serve that very profound human need so answering that question brought about this new field of socially assisted robotics it's not so new now it's about 21 years old so it's a young adult you know getting out there um and this is really exciting so this this field is based on the premise that these robots are physical robots they're in the real world you'll see they're Out Among Us but at the same time they're not doing any physical work in fact they're helping each of us do our own work through social support so that's the idea of socially assistive robotics so let me give you some examples so in my lab we started out well in a lot of different things but we started out in particular in stroke rehab from the very beginning because remember it's a slog to recover even the basic of functions so we created these robots that would come up to a stroke patient and kind of encourag them and well let's see let's see what they did how about we do a fun activity where I'm the robot and you're the librarian and you put the books away oh so she likes it you sure that's great immediately though she cheats look at that she immediately tries to District the robot that's humans for you good job no how does he know she keeps cheating again maybe it'll work this time how does he know how does he know and also why is it a he so we learned that people are really interesting and you can't just Creative transactional Technologies and hope that people will like them in fact you have to make really engaging interesting personalized robots and so one of the things that we did early was to create robots that actually had personalities so we know that people get along better with people with similar personalities so we thought well you know if if we created robots that are maybe really kind of assertive and more like coaches that would really work with users that are more in the extra version side of the spectrum of extroversion introversion and if we created kind of more supportive you're doing great it'll be fine robots for those who are maybe more in the intr version side of the spectrum and we did and we tried it and we found that actually users who had a robot that matched their personality did these boring exercises much longer and they reported liking them more so we took a lot of those lessons and then we also worked with um altimer patients who are not expected to really recover unfortunately or improve on task and yet we were able to create these robots that would sing songs very badly I will not play that video for you you'll have to look it up on the web um they would sing songs and encourage users to kind of guess the game and push the right buttons and we were able to get Alzheimer's patients to improve um this was a long six-month study and they were actually able to improve and most importantly they really enjoyed interacting with the robot they enjoyed interacting with the robot and that's important because they have less and less going on in their lives um we also took robots and we put them in the homes of elderly people who were living alone and were feeling pretty lonely and tended to be quite sedentary now that's not one of the elderly participants in our study that's the PHD student who built the robot um but as you can see the robot is small and cute and he would sit on its little pedestal he would tell Dad jokes and he would also do a little dance and um people love that they would get up they would dance with the robot they would you know want to hear more jokes it was extremely effective we found that basically the people who had this robot in their home for 2 weeks would sit a lot less but unfortunately at the end we had to take the robots back home you know to the lab to their home um and our users unfortunately went back to their sedentary ways because they missed their robot buddy I mean who wouldn't and while we were working with users on the sort of other side of the age Spectrum we were also always very interested in infants and babies um and young children because if you can help them early it really changes the course of their lives and so it's never too early to start working with a robot to help you remember I told you about rates of Developmental disorders in uh very young babies so we were interested in getting babies who are not moving properly to exercise their body properly but how do you get a baby to move in a particular way spoiler alert you can't you also can't get anyone to do anything you want but that aside how do you coach babies really hard unless you can give them a robot that's their size and they look at it and they may be motivated to imitate so so let's watch that so first the baby's watching and when the baby kicks the robot kicks to reinforce the baby sometimes the robot kicks and the baby imitates and after only about a couple of minutes the baby learns and they even like it and what's important is they actually practice more which helps them recover and it helps them learn movement so it's kind of like a really cute coach for an even cuter baby and then we were inspired by that so we developed robots for children with cerebral paly who need to do exercises that are really really boring and we said well if you have to keep doing this maybe the funnest way to do that is to do a number guessing game with a robot and the robot guesses a number and you say no higher lower higher lower so that was very effective um we developed a robot that would help children at Children's Hospital Los Angeles who were about to receive IV injection treatments which are quite painful to cope with the upcoming pain and anxiety that they were feeling so it's really heartwarming doing this and then as I mentioned before autism which is so prevalent now um has been a real challenge for us for the Healthcare Community for the educational Community but interestingly autism is a really great challenge for AI because the way AI works now is that it takes big data to train big models in order to exhibit you know big general intelligence well in autism is the other way around you have tiny amounts of data and the data are really messy and really inconsistent because each individ ual is different and each individual changes over time often in unexpected ways and in fact I think that's true for all people all the time so autism is a really great challenge case for AI so how do you develop systems that can adapt to each user each child and so we started out by developing robots that would just stimulate and engage children in doing Play Because often they don't know how to start playing and then when you play you have to do things like take turns and make eye contact and that that's really a lot of social skill learning and that's an old robot as you can tell they get better looking and then we were also interested in the fact that imitation is a really important way to learn social skills but often um children on the Spectrum have issues with learning imitation and so and then learning by imitation and so we created a robot that would help children be inspired and motivated to imitate and would teach them to imitate so let's look at what that looks like yes yes now move like this move like this so do you hear an echo basically the robot tells the child what to do but the child is imitating the movement and the voice because even knowing what to imitate what not to imitate is not obvious so we learned a great deal from these studies over the years and in the end we we got to a point where we could actually create a complete robot that we could leave in the homes of families with autism for a month to 6 weeks and so we did this um and the robot was supposed to help children learn math skills but when you're learning you're not just learning any one thing you're not just learning math you're also learning social skills that's a really important thing in school kids are learning social and cognitive and emotional regulation skills and that's what this robot was trying to do meet a Adrian age six this has done great work and his robot friend kiwi you are doing an amazing job on this weekend morning they've settled in to play some games along with big brother Darren Adrien is on the autism spectrum and kiwi is no toy it's a socially assisted robot you doing really great keep up the good work so the thing that's interesting here is that Adrien has a screen he could just be working from the screen but then he won't learn any of the social skills and he won't have the motivation and the engagement that's what the robot is for and the children really love the robots that would you know wrap them in towels that was a little scary for us from the engineering side never know what's going to happen um but you know they they named them and they didn't want to let them go at the end of the study and we keep having this experience and this was especially sad because this was two years before the covid pandemic happened and before quarantine and I just I so wish that millions of children and parents could have had robots like to help them through what we all went through so now as I mentioned anxiety and depression are at you know epidemic levels and so what we have developed lately is actually much smaller cheaper lighter 3D printed robots as the one that you see it has a crochet skin you can actually just make it out of cloth as well the skin you can decorate it you can knit it you can do whatever you'd like and we have actually deployed these robots in the dorms at USC to help students deal with stress to focus um to help them study and even to help them do some breathing exercises and to do behavioral um and to do um cognitive behavioral therapy and really encouragingly we found that after two weeks of working with our robots we found that the students had lower rates of psychiatric distress and interestingly we did a control experiment where we compared it to chatbots and interacting with chatbots did not lower their psychiatric distress so it's really important to have that embodied robot that is your buddy like you in this physical world so now you've met some of our robots and you can see that our goal has been always to create these machines that are for each person and we're aiming to make them ever lower cost Ever more accessible hopefully ever cuter ever more engaging these are really low cost they're on the web you can download the design you can build your own you can have your own coach buddy whatever you'd like you can run research studies that's the idea to really make this possible for as many people as possible and to that to me overall is the vision of human centered AI so AI can be many things and it will be many things but one thing that I think it needs to be more of is human centered so the idea of these socially assisted robots these robots that care is that they're there to help each person pursue their purpose and this is really different for most of AI because if you think about what is most of AI about most of AI is about doing work so that people don't do that work but if we don't have any work to do then what is our purpose purpose is really important to what makes us human so the idea with social assistive robotics is actually to help you maintain that purpose and to me that's a great reason to be in AI so that's why Mama makes robots um that and the amazing people especially students that I've gotten to work with and that join the privilege of seeing do amazing things so that's my story that's why I'm in AI That's why I'm in AI um what about you why are you interested in [Applause] AI

Channel: TEDx Talks

Views: 2,015

Keywords: AI, Engineering, English, Mental health, Robots, TEDxTalks, Technology, [TEDxEID:55171]

Id: ybm9TleOQss

Length: 18min 20sec (1100 seconds)

Published: Fri Mar 15 2024