Latest / Trojan Talks / Ep. 10: President Kim with Maja Mataric
Transcript
- 0:04Welcome to Trojan Talks.
- 0:05My name is
- 0:06Beong-Soo Kim, president of the University of Southern California,
- 0:08and this podcast is a space for open conversation
- 0:12with thought leaders inside and beyond our university
- 0:15about the ideas shaping USC’s future.
- 0:18Today, I am delighted to be joined by Distinguished
- 0:21Professor Maja Mataric,
- 0:23one of the world's leading researchers
- 0:25in robotics, artificial intelligence
- 0:27and socially assistive robotics,
- 0:30technology designed to coach, motivate and support
- 0:33people in areas like rehabilitation,
- 0:35autism therapy,
- 0:37eldercare and mental health.
- 0:39What I find so compelling about her work is that it puts people
- 0:41at the center of technology and innovation.
- 0:45Maja, so delighted to be with you.
- 0:47Thank you for having me.
- 0:48So there are so many questions I have,
- 0:51but why don't I just dive right in
- 0:54and ask you about this
- 0:57very powerful narrative
- 1:00and set of concerns that people have
- 1:03at this moment more than ever, about robots
- 1:06and AI replacing humans.
- 1:09Your research is so beautiful
- 1:11because I feel like it flips the question
- 1:14a little and asks, how can robots
- 1:17and AI actually help humans reach their full potential?
- 1:22Thank you
- 1:22for the opportunity to talk about my favorite topic.
- 1:25So people often like to think about AI for good.
- 1:28And I've actually thought about what does it mean
- 1:30to do good, at least for humanity.
- 1:33And if we look back at what we know
- 1:34about what makes people happy, what makes people thrive,
- 1:37it tends to have to do with
- 1:39the being connected with other people.
- 1:41And so I've taken that as a cue
- 1:44in work to think about rather than replacing
- 1:47purpose in human lives, purpose
- 1:49that comes from being with others and from doing something in life,
- 1:53like giving us some sense of accomplishment.
- 1:55So instead of removing that, which is the notion of automation,
- 1:58so automation is about replacement,
- 2:00and instead augmenting it,
- 2:02so augmentation is enhancement.
- 2:05So I always say, instead of too much automation,
- 2:08we should have more augmentation.
- 2:11If we look at technology today, because of, of course
- 2:14the opportunities for monetization,
- 2:17it's generally cheaper if we replace human labor.
- 2:19And so that is the threat in many people's eyes of
- 2:22AI is the replacement.
- 2:24And what is it balanced with?
- 2:26But I see this amazing
- 2:27set of opportunities for augmentation,
- 2:29for enhancement of human ability and human purpose.
- 2:33And I would love to see more research
- 2:35going into that direction.
- 2:37So there have obviously been incredible
- 2:40leaps forward over the last several years
- 2:42with respect to large language models.
- 2:44Your focus has largely been
- 2:47on physical intelligence
- 2:51as a way
- 2:52to form those connections with people.
- 2:55How have these developments
- 2:57with large language models
- 2:59enhanced your research and gotten
- 3:02you excited or even worried about what's to come?
- 3:05Now, with the
- 3:08tremendous explosion of AI, this idea
- 3:10that if we really have both enough data
- 3:12and enough computing and this representation of transformers,
- 3:16wow, these amazing, unprecedented
- 3:18reasoning abilities happen,
- 3:20or at least the behavior that appears intelligent.
- 3:23Because let's be clear, we don't really understand
- 3:24exactly how it works, so we have to be honest about that.
- 3:27Even all of us who are full-on technologists, we've created something
- 3:30that is so complex that we don't fully understand it,
- 3:33and it is not fully predictable.
- 3:35So first, we have to accept that.
- 3:36We've created something.
- 3:37We brought it into the world.
- 3:38It's like a child, and now you have to bring it up.
- 3:41And so it's tremendously powerful.
- 3:43It has advanced, just ...
- 3:45not just cognitive AI, right?
- 3:47You can talk to a computer.
- 3:48It talks back.
- 3:48It's like talking to a human, a
- 3:50very smart human, potentially.
- 3:53But for robots, for behavior in the physical world ...
- 3:56So what have these models done
- 3:58for actual robots?
- 4:00Not quite as much, because it turns out,
- 4:02while language is very special,
- 4:04language is how we relate to one another,
- 4:07behavior in the physical world,
- 4:08like understanding the scene around
- 4:10me, understanding that you're
- 4:12here, that it’s during the day, that the lights are maybe a little hot
- 4:15and the camera is wherever, and so on ...
- 4:17Understanding all of that takes a tremendous amount
- 4:20of learning early in life.
- 4:22And so embodied systems,
- 4:24including physical AI, need a lot more training.
- 4:27Not just data, but training, trial and error.
- 4:30There is a reason why all of us humans during development,
- 4:33you know, we take months to figure out how to grasp something,
- 4:37and there's really no shortcut for that.
- 4:39People would like to think
- 4:40that there's a shortcut,
- 4:40that you're just going to, well, it's like showing a baby.
- 4:43Oh, look, you know, here's how you grab a coffee cup.
- 4:45Go. Doesn't work.
- 4:47So the trial and error, the tremendous amount of experience,
- 4:51that is something that we don't know how to shortcut.
- 4:54I know that some of your work in robotics
- 4:56and in socially assistive
- 5:00interaction
- 5:01hasn't really focused as much on language.
- 5:03I'm, I'm thinking of one of your experiments
- 5:07where you have a robot
- 5:10cohabiting with an elderly person,
- 5:12and the robot will simply get up
- 5:14at a particular point in time
- 5:16as a reminder to the person
- 5:18to get up and, and be physically active.
- 5:22That didn't necessarily involve
- 5:24any kind of language, right?
- 5:26Right.
- 5:26So before we even had language,
- 5:28we were still humans and we related to one another.
- 5:31And, you know, ultimately we're animals, and language is not necessary.
- 5:34It is just tremendously powerful.
- 5:36And it's created all the other technologies that we have.
- 5:39So we don't want to set it aside.
- 5:40But we did studies, goodness, 20 years ago where we showed that
- 5:44if you just put a physical robot
- 5:46next to a person that is doing
- 5:47maybe stroke rehabilitation exercises,
- 5:49something really boring and demoralizing,
- 5:51just having a robot there that doesn't say anything, doesn't even move,
- 5:55makes the person feel more alert, and they exercise harder.
- 5:58Now, of course, we know now that there's this idea of body
- 6:01doubling that, for example, for individuals with ADHD,
- 6:04just having another person present, not even paying
- 6:06attention to them, helps them focus.
- 6:08So this notion of presence and embodiment
- 6:11is something that we are deeply rooted in.
- 6:14And then you build from there, the physical embodiment, the way that,
- 6:17you know,
- 6:17when I'm doing all of this with my hands,
- 6:18your brain is ready to imitate, but it's not.
- 6:21And that's good.
- 6:22But you really, the way you understand
- 6:24my behavior is by mapping it to your own body.
- 6:27So our physical embodiment is really important.
- 6:29And that's why that part is still missing in AI.
- 6:32And so getting back to the question
- 6:34about how these developments in
- 6:37large language models in AI has impacted your work.
- 6:41Sure, the, the robots that you're working with
- 6:43now have this ability to speak
- 6:45conversationally in a way that passes the Turing test,
- 6:48but it sounds to me that it hasn't had
- 6:53as much impact on the
- 6:55issues that you're interested in because you're focused
- 6:58not just on the language part of intelligence, but also on
- 7:01the physical intelligence and the social interaction.
- 7:05Absolutely.
- 7:05And in fact, I would be happy if language were enough.
- 7:09But we find in really profoundly
- 7:12critical contexts that it's not enough.
- 7:14So, for example,
- 7:16billions and billions of people
- 7:17are being, wanting to believe ...
- 7:19Sorry.
- 7:20So for example, there are millions
- 7:23and more being invested in, for example, apps for mental health.
- 7:26So, you know, it's an app, you talk to it
- 7:28and it's trying to help you.
- 7:30It turns out that you can get the absolute
- 7:33best mental health advice, support, any,
- 7:36any basically therapeutic method that humans can deliver.
- 7:40You can get it on an app, but it doesn't work the same way.
- 7:43And, in fact, it doesn't seem to work in the long run.
- 7:46So there's something about that human connection that's critical.
- 7:49In fact,
- 7:49in mental health, we do know that the best predictor of recovery
- 7:52is actually how well the therapist and the client get along.
- 7:56That is the most important thing.
- 7:57It's not what you do.
- 7:59It's not how often you do it.
- 8:00It's how you get along with this other person,
- 8:02and this idea of getting along, which is a social construct,
- 8:06an emotional construct, an embodied construct,
- 8:09these things really important.
- 8:11And I actually really worry that we're missing
- 8:14some of that as we're developing AI that as yet
- 8:17doesn't have it, but will have to have it.
- 8:20So we've been talking
- 8:22about the importance of getting along
- 8:24and even having a, a
- 8:26body double there to
- 8:29really help with
- 8:30learning and with motivation.
- 8:32But let me just back up a little bit.
- 8:35The robots that you work with, Maja,
- 8:38they don't look like humans or people.
- 8:40So how, how
- 8:43similar looking to a human
- 8:45do they need to be according to the research that you've done?
- 8:49This is such a good question because maybe 10 years ago,
- 8:52it was pretty much impossible to have a humanoid robot
- 8:56that could do anything at all.
- 8:58So it was, you know, a really hard problem.
- 8:59People could build them, but they couldn't really move them.
- 9:01Now we're slightly better, although it's still going to take a while.
- 9:05So we were, for many reasons in my lab,
- 9:07not focusing on humanoids, partly because of cost.
- 9:10I would like to believe
- 9:11that we can create technologies that will be accessible and affordable
- 9:14for people very soon and not, you know, decades from now.
- 9:17And humanoids are not cheap.
- 9:19They're also not inherently safe.
- 9:20Anything that you build that has enough strength to do, for example,
- 9:24pick up objects in the environment also,
- 9:26unfortunately, has enough strength to hurt people, and safety
- 9:28is still very hard to guarantee.
- 9:30And so we focused away from humanoids for those practical reasons.
- 9:34And then we discovered these tremendous effects.
- 9:36And when I say “we,” I mean the field, you know, it’s not just my lab,
- 9:40that it turns out that what's important
- 9:42is the physical embodiment
- 9:44and the presence in the same space, the,
- 9:46the body,
- 9:48the specific body that you have
- 9:50for the robot, is not as important.
- 9:52For example, it's important that it be expressive, right?
- 9:54So if I'm trying to do something, like I'm trying to show you
- 9:56how to do stroke exercises, well then I probably need arms.
- 10:00But if I'm trying to help you with mental health
- 10:02exercises, I don't.
- 10:04And so we have thought about robot bodies
- 10:07as serving the purpose of the interaction.
- 10:09For our robots that are being used in USC dorms
- 10:12for cognitive behavioral therapy and other kinds of support,
- 10:16these are just cute little squash and stretch
- 10:18tabletop robots, and they're fun to have in your dorm.
- 10:20And they are just enough
- 10:23for what they have to be, and no more.
- 10:26Your answer makes me think of pets.
- 10:28So pets don't look like humans,
- 10:30but obviously
- 10:32people are able to form very, very strong connections.
- 10:35And I would imagine
- 10:37that they also actually facilitate
- 10:40some of the positive benefits that we've been discussing.
- 10:44Oh, absolutely.
- 10:45People love their pets.
- 10:46And it is known that, for example,
- 10:48having a dog makes people generally live longer.
- 10:51Now, if we dig into why it does that,
- 10:52it does it for all the reasons that you might expect.
- 10:54When you have a dog, you have something that gets you out to do things.
- 10:58You go for a walk.
- 10:59You talk to other people.
- 11:00You're being physically active.
- 11:01You're outside.
- 11:02So you're doing all the things that are good for you, but
- 11:04it gives you a sense of reason or purpose for it.
- 11:07And a creature
- 11:09is much better at doing that,
- 11:11including a creature being a peer, eventually a human.
- 11:14But it's important to have that.
- 11:16Without that, we lack the motivation
- 11:19to do the things that we need to do.
- 11:20So when you're designing robots
- 11:22or working with robots or choosing a robot,
- 11:24Maja, you're not necessarily looking for a humanoid.
- 11:28In fact, you could even imagine
- 11:31replicating a pet or something
- 11:33cute that's not humanoid,
- 11:35could be as effective or even more effective?
- 11:37Exactly.
- 11:38In fact, we don't want humanoids by default.
- 11:41Because here's the problem with humanoids:
- 11:42When you have a human form,
- 11:45the user expects a lot, right?
- 11:46Oh, it looks like a human.
- 11:47So it should be ...
- 11:48It should move like a human, and it should be as smart as a human.
- 11:51Now we're getting with AI to a point where, you know, we can look like
- 11:54the robot is as smart as a human, but it can barely move.
- 11:58It cannot do the very basic
- 12:00things very well.
- 12:02That will change.
- 12:03It's just going to take much longer than people would like.
- 12:05Recently, you began a study
- 12:07looking at the uses
- 12:11of socially assistive robotics for mental health.
- 12:14And I understand that you've chosen college students
- 12:17as the cohort you're going to be focusing on.
- 12:19Can you share a little bit more about that?
- 12:20I would love to talk about this.
- 12:22So I'm really, first of all, honored
- 12:24that we're doing an NIH clinical trial.
- 12:26So that's really an acknowledgment that the technology's mature enough
- 12:30to where we can actually deploy it in people's lives, in this case
- 12:33university students.
- 12:35And, you know, it's safe and it's important to test the outcome.
- 12:38So that's, you know, a wonderful stage that we're in.
- 12:41Now, in most of my lab
- 12:43studies up until now, we would do some kinds of evaluations
- 12:46with USC students
- 12:47because they're called, so-called convenience population.
- 12:49So you see that in studies all the time.
- 12:51You have the convenience population, and then eventually
- 12:53you go to your real population that you care about.
- 12:56But with mental health support, in particular anxiety management,
- 12:59because anxiety rates are very hard in university students,
- 13:02so anxiety’s high across the board in the population,
- 13:05it is very high during the educational settings,
- 13:07high school and college.
- 13:09And so now we're working with university students
- 13:12as our actual population, not just convenience population.
- 13:16But to me, it's much more than that.
- 13:18You know, these are students who are in my classes.
- 13:21These are students who are in my lab working as undergrads.
- 13:24And the fact that they've chosen to work on
- 13:25this is because it's very personal for them.
- 13:28And so it's very special
- 13:30to be in that space, to be able to work on something
- 13:32that is so presently
- 13:34acutely real in their lives.
- 13:36And how long is this study going to go for?
- 13:38And tell me, tell us a little bit
- 13:40more about what you're looking for.
- 13:43So this is a ...
- 13:45the clinical trial, it's a four-year grant, of which the last
- 13:49three years of the clinical trial, we are in it right now.
- 13:52So we're collecting data right now.
- 13:53There are robots, like at this moment, there aren't because they're
- 13:56starting ...
- 13:56we're starting a six-week trial soon.
- 13:59But basically the, the, at the heart of the study
- 14:02is getting at this issue of the robots’ embodiment
- 14:06making a difference in the outcomes and the engagement
- 14:09that the user feels, the alignment with the robot,
- 14:12just like the alignment with the actual therapist
- 14:15that motivates them to practice.
- 14:17And then,
- 14:19they've ...
- 14:19over the practice,
- 14:20the ability to continue to do it so that over a long period of time,
- 14:24six weeks at least, you can see outcomes.
- 14:26The reason that this is important is because
- 14:28in mental health,
- 14:30there is a menu of skills
- 14:32one can develop for coping,
- 14:33but they only work if you really are well practiced at them.
- 14:37And of course practice is hard.
- 14:39People don't make time.
- 14:40It's, it’s emotionally difficult.
- 14:41It's cognitively difficult.
- 14:43So how do we motivate people to practice?
- 14:45And in many ways, it's the same thing
- 14:47that we've done with socially assistive robotics for,
- 14:50gosh, 20-plus years.
- 14:51Because we started with stroke rehabilitation.
- 14:54We worked with children on the autism spectrum learning
- 14:57to make eye contact again, behaviors that are not natural,
- 15:00that are hard for a variety of reasons,
- 15:03but you still need to do them.
- 15:04So in many ways, I think about what
- 15:07we do as helping people help themselves.
- 15:09So how do we motivate you to do this thing
- 15:11that you should be doing but you find hard?
- 15:14And what we're finding is that because it's a robot,
- 15:17it has a different effect
- 15:20on the interaction than if it's just a chatbot.
- 15:23Ironically, students
- 15:25prefer to interact with the chatbot.
- 15:27They much prefer to have a nice screen or their phone,
- 15:30but unfortunately those things are not working for them.
- 15:34We have even had students who have used AI,
- 15:37rather than answer the questions
- 15:39themselves about how they feel.
- 15:41So I say that with a very big heart.
- 15:43I'm not saying, “Oh, they're cheating!” I mean, they are cheating,
- 15:45but why are they cheating?
- 15:48The important question is
- 15:49what is motivating them, right?
- 15:51They are, they're part of this study.
- 15:53They have elevated anxiety.
- 15:54They have something that they know they need to work on.
- 15:57But it's so hard.
- 15:59And I think it's all about
- 16:00understanding human motivation.
- 16:02And something about these physically
- 16:04embodied robots is allowing us
- 16:06to probe at that question and to motivate people more effectively.
- 16:10So that's why we do it.
- 16:11And going back to the question earlier
- 16:12about how developments in AI
- 16:15and LLMs have impacted your research,
- 16:18are the robots in the study
- 16:20that you're conducting, do they speak?
- 16:23How do they engage in terms
- 16:25of language with the participants?
- 16:27Right.
- 16:28So, you know, if you're thinking about cognitive
- 16:30behavioral therapy, the, the robot needs to talk, right?
- 16:32So the robots do talk.
- 16:34And in fact, we find that some study participants
- 16:37prefer to type in their answers, maybe because of privacy.
- 16:40Maybe that's just, they're so used to typing.
- 16:43Others just want to speak their answers.
- 16:44So we want to give them the opportunity to do whatever.
- 16:47But the robot is talking to them.
- 16:49And then we also have other kinds of exercises,
- 16:52like we have a breathing exercise
- 16:53where they actually put their hands on the robot
- 16:55and the robot breathes with them.
- 16:57So it kind of expands its little belly.
- 16:59There's no chest, really.
- 17:00It's like a head and, you know, so it expands.
- 17:02And that's very important because when they're
- 17:04putting their hands on it, they can feel it.
- 17:06And that embodied experience actually motivates the breathing.
- 17:09And I'm really excited to say that we see that their,
- 17:13their experience of anxiety, their arousal, declines.
- 17:17They calm down after
- 17:18doing these exercises with the robot.
- 17:20So that's a really good sign.
- 17:22It must be really gratifying to be able
- 17:24to do this work with students who are in your lab
- 17:28and whom you're also mentoring, right?
- 17:30I love mentoring.
- 17:31I’m, I'm always the mom, might be a little too much, but ...
- 17:34so it is hugely satisfying.
- 17:36I'm, I really have, have been, throughout my career,
- 17:39just honored to have these amazing students come in, do
- 17:42really, really hard things, and do them well.
- 17:46And then I look good.
- 17:47So, so this is ...
- 17:48what a great career to have.
- 17:50But truly, I think all times are hard
- 17:53and students are going through hard times and going through
- 17:56graduate school, through undergraduate,
- 17:57it's hard at any time
- 17:59and possibly especially hard now.
- 18:01And so it really is both an honor to kind of
- 18:04be there along for the ride and to support them,
- 18:06but also to see this,
- 18:09you know, to witness this amazing
- 18:10sense of wanting to do good.
- 18:12These students come in, they want to do something amazing.
- 18:14They want to help other people.
- 18:16They learn how hard it is to do that, but they stick with it.
- 18:20And, you know, I'm just I'm honored by that.
- 18:22And these students
- 18:24that you get, they’re from all sorts of different fields,
- 18:26just as you would expect at USC, right?
- 18:28Absolutely.
- 18:28We get plenty of engineering students, as you'd expect.
- 18:31But the part that I find really heartwarming
- 18:34is that we will get students from across USC,
- 18:37and what I love is they'll come in and they'll say, “Well,
- 18:39I don't know what I can contribute.” But in fact,
- 18:41they contribute perspective.
- 18:43They contribute their time.
- 18:45You know, we've had students
- 18:46take our group to see the Special Olympics
- 18:48when they were on campus to broaden our minds.
- 18:51And so it really sounds like a cliche, but it’s, it's really true.
- 18:54And also, if a student wants to volunteer in my lab,
- 18:57of course I will let them.
- 18:58They want to do something meaningful, come on in.
- 19:01tell me a little bit more
- 19:03about how our environment, in particular
- 19:06the ability to work with researchers
- 19:09at CHLA and work with the neuroscientists,
- 19:12how has that been helpful in terms of advancing your research?
- 19:16So the thing that I always used
- 19:17to say about USC, and it's still true,
- 19:19but even for like decades ago,
- 19:21I used to say this.
- 19:23One, it's an incredibly interdisciplinary place.
- 19:25But every place says that.
- 19:26They always say, “Oh, you know, we're interdisciplinary.”
- 19:28Everyone says that.
- 19:30The way I experienced that personally,
- 19:32all the time, is that I can reach out to anybody at USC.
- 19:35They can be at the Keck School of Medicine, which is ultimately
- 19:38how we ended up starting this entire field
- 19:40of socially assisted robotics, is that it was folks from Keck
- 19:43who reached out to me and said, “Hey, you know,
- 19:45can you make our surgical robots better?” And I said, “I cannot.
- 19:48That's a whole other field.” But how about post-op?
- 19:51And then we started this whole new direction.
- 19:54Annenberg, Gerontology, Dornsife, anywhere.
- 19:58I've never had anybody at USC,
- 20:00and this is true over many decades, I've never had any colleague
- 20:03look at me and say, “OK, crazy robot lady.
- 20:06Like, what, what do we have to talk about?”
- 20:08People would always have this attitude of like,
- 20:10“Hmm, that's interesting.
- 20:11Well, that's, tell me more.” And then we would team up
- 20:15and completely new things would emerge.
- 20:17So I love that because it's not just talk.
- 20:20It's not even, you can say, “Oh,
- 20:21I like interdisciplinarity.” Interdisciplinarity is hard.
- 20:24The language is different.
- 20:26The goals might be different.
- 20:27But that's never stopped my colleagues.
- 20:29And so I love that.
- 20:30That is, to me that's been
- 20:32one of the, the major just
- 20:36cultural aspects of USC.
- 20:37It's not, no one tells you to be like this.
- 20:39Maybe we're just drawn to it and that's why we're here.
- 20:42Or maybe I just have an extremely biased experience,
- 20:45but I think after this many decades ...
- 20:47Yeah, I think we're just like this.
- 20:49There are two things I always say about USC,
- 20:50and I didn't say the second thing,
- 20:52which is that one is interdisciplinarity.
- 20:54The other one is actually focus on societal relevance.
- 20:57So this is a part of a sort of an implicit mission for us
- 21:00is that we feel like we should be doing research
- 21:04that is somehow going to help the world, help society.
- 21:06I see a lot of my colleagues feeling that almost as a default.
- 21:09And that also brings us together.
- 21:11That's why
- 21:11we like to team up and look across, because these are hard problems.
- 21:14I feel in your work that you're really trying
- 21:17to have as much impact that you can have in
- 21:20as short a period of time as possible.
- 21:21It definitely helps me sleep better at night.
- 21:24Yeah, Maja, so we've talked about
- 21:28the different forms of intelligence that you appreciate.
- 21:30I think you've described this
- 21:33as a symphony of physical
- 21:36and social, emotional and cognitive intelligence,
- 21:39and it seems like
- 21:41we've made so much progress
- 21:43recently in terms of cognitive intelligence,
- 21:46and that the physical
- 21:47and the social, emotional intelligence
- 21:50has not catapulted as quickly forward.
- 21:53I was wondering
- 21:55what your thoughts are on
- 21:58how to accelerate
- 22:00our understanding of these other forms of intelligence
- 22:03and where this is all taking us in five or 10 years?
- 22:06This is the perfect segue,
- 22:10because I think what is missing in the development of AI right now
- 22:13is actually the interdisciplinary perspective.
- 22:15So the reason that everyone is focused
- 22:18on cognitive AI,
- 22:20which is language-based AI,
- 22:22is that people are not taking human psychology into account.
- 22:25And so people who develop AI are
- 22:27generally engineers, computational people, you know, so,
- 22:30you know, they just don't have the background.
- 22:32And so they don't understand that
- 22:34social intelligence is really important.
- 22:36And it is not the same as cognitive stuff.
- 22:39And then emotional intelligence ...
- 22:40Oh, well, that's just completely alien to people.
- 22:42They haven't ...
- 22:42It's not even on their horizon.
- 22:44And yet emotions, affect, is what drives everything we do.
- 22:48It informs everything we do.
- 22:49It's absolutely necessary for reasoning.
- 22:51It's not secondary. It's right there.
- 22:54So I think the lack of interdisciplinary perspectives
- 22:57is what's keeping AI from moving forward.
- 23:00Because, you know, you have an engineering race
- 23:02and people want short answers and something that's quick and easy.
- 23:05And, you know, interdisciplinarity is not quick and easy,
- 23:08but it is fundamentally important.
- 23:10Hopefully, thanks to your research, we’ll
- 23:12be able to make more advances in both physical intelligence
- 23:15but also social and emotional intelligence
- 23:18and bringing it all together
- 23:20in the robots that you work with.
- 23:23Where do you see this heading,
- 23:25and what are you excited about?
- 23:27And also, what are you concerned about?
- 23:30Oh, goodness.
- 23:32I really want to be optimistic
- 23:34because I think the optimism would have to come
- 23:37from what I think about human nature.
- 23:39So we have created AI,
- 23:42you know, we're a very smart species.
- 23:44We've created this thing,
- 23:45but it is very complex.
- 23:47And this is why I compare it to having a child,
- 23:49because you've created something
- 23:51and you don't really understand it, and you can really mess it up.
- 23:55And then you have to let it into the world.
- 23:57But before you let it into the world,
- 23:58there's a whole lot of nurturing that needs to happen.
- 24:00And I think the concern with AI is that we don't understand it.
- 24:03We're not doing enough nurturing and trying to figure it out
- 24:06before we let it out into the world.
- 24:09So, if we become good parents,
- 24:12then I think we can direct
- 24:13AI towards the massive unsolved world problems, right?
- 24:17I mean, there's so much for us to worry about.
- 24:19You know, there's climate, there's poverty, there's just all ...
- 24:22disease, everything.
- 24:24But if we're not focusing on
- 24:25that and if we're racing to just get the next thing out
- 24:28so that we can claim AGI,
- 24:31that is concerning.
- 24:32And so
- 24:34I don't want to choose.
- 24:36There are many possible futures.
- 24:38Some futures are more probable than others based on human history.
- 24:41But I want to believe that we will rise to the challenge.
- 24:44And the place where I get my optimism is the students.
- 24:48That's why I love the work, because I see these students
- 24:50and they're really hoping and trying
- 24:53to do something that is good.
- 24:56Can you imagine a future focusing
- 24:58just on socially assistive robotics,
- 25:00where everyone has their own socially assistive
- 25:03robot?
- 25:05It is so funny because I always say, well, yes,
- 25:07let's everyone needs a robot like this to be their fundamental support,
- 25:10but I worry that that argument
- 25:12can be taken to a, one possible conclusion, which is,
- 25:16well, let's just have that and then we don't need humans.
- 25:18And what's really concerning about that is that we know
- 25:22that we already have increased rates of isolation and anxiety
- 25:25and depression, because people aren't
- 25:26seeing each other in person and talking and forming friendships
- 25:29and dealing with all the messiness of human relationships.
- 25:33But we need that.
- 25:34We need that to be well and to thrive.
- 25:36And if you provide people
- 25:38even with robots that are, you know,
- 25:41wonderful and supportive and more effective
- 25:43than just apps,
- 25:45it's still not nearly enough.
- 25:47People need people.
- 25:48And all of neuroscience and psychology tells us that.
- 25:52But we're right now not paying attention,
- 25:54and that's concerning.
- 25:56Maja Mataric, thank you so much for your time.
- 25:59And I really appreciated the conversation
- 26:02and everything that you're doing in your research.
- 26:05So good luck and hopefully we can check in soon.
- 26:07Thank you.