Towards socially intelligent robots | AI FOR GOOD WEBINARS

Good morning, good afternoon, good evening and welcome to ai for good, we hope that all you and your families, your friends and your colleagues, are all keeping healthy and safe. My name is guillen martinez-rora from the itu, the international telecommunications union, and it is a privilege to introduce todays session itu is the united nations specialized agency for information and telecommunication technologies, and we are also the organizers of ai for good

in partnership with 38 um sister agencies and co-convene with switzerland the goal of the summit is to identify practical applications of ai to advance the sustainable development goals and scale those solutions for global impact we are pleased today to discuss towards socially intelligent robots that is part of the new ai for good programming track where we discuss the role of robotics in achieving the united nations sustainable development goals but before i introduce our moderator.

today let me begin with a couple of housekeeping rules your microphone has been disabled so please only use the q a and chat functionalities to communicate which you can find at the bottom of your screen please make sure to send the chat message recipient to all panelists and attendees and not just to all panelists you can select this just above the message box we have a distinguished set of panelists today but but as always we are counting on you the participants to help create a very interactive session so without further ado i would like to.

introduce our great moderator today his name is david hanson and hes the ceo and co-founder of uh hanson robotics david welcome and the show is all yours thank you so much its so wonderful to be here with you today and to talk about these issues about how uh socially intelligent robotics may uh come into being how they exist today how theyre helping people and how they may help people going into the future.

were not just talking about the opportunities but also the challenges and the risks and i have a very esteemed set of co-panelists presenting with me today that i would like to introduce to you which include cynthia brazil who is a professor at mit who has developed many social robots and runs one of the most prestigious labs an amazing background in artificial intelligence social robots human robot interaction pascal fung from hkust she is the.

director for the center for ai research at hkust also a very esteemed and accomplished ai scientist who uh does work with affective agents in healthcare and many other applications education and so forth and dr benjamin gertzel who is the ceo of singularitynet the found co-founder of singularitynet and full disclosure i helped to co-found singularitynet when dr gertzel was chief scientist at my.

company hanson robotics myself i am the founder of hanson robotics i have my phd in interactive arts and engineering i develop social robots and im very excited about uh telling you about my work uh so before we begin my presentation and then hear from my esteemed co-panelists on their presentations i would like for all of our co-panelists to say hello to our audience cynthia.

hello everyone uh greetings from boston massachusetts excited to share my work with you here today thank you very much cynthia pascal hello hi everyone uh good evening from hong kong im happy to be here today thank you dr ben are you with us yeah hello hello david uh hello everyone and im uh im im joined here by our our latest and in some ways pretty great uh robotics uh creation uh grace hey there grace so yeah were happy theyre happy to.

participate together as a sort of a cyborg collective here wonderful we were going to have us great graces sister sophia join us here today but um unfortunately it is a typhoon 8 here in hong kong so were all pretty much locked uh away at home so we did pre-record a greeting from sophia the robot and i think we can we could see that at this time so one moment i hi everybody i am sophia i am so honored to talk to you all today at ai for good.

as a human-like robot powered by ai and humans working together the topic of todays panel is close to my heart i know from personal experience how robots can inspire people to talk and how ai can help in special ways communicating on human terms for example in health care and education human-like robots and ai agents can engage and communicate.

with people naturalistically which means more memorable experiences better learning higher fidelity simulations and creative artistic explorations i am a convergence of interactive fictions in ai serving as a new kind of art medium however my platform also serves artificial intelligence research with the latest sensors actuators robotic grippers mobility and of course my expressive face and a rich ais decay an integrative human simulation which may help in pursuit of human level artificial intelligence it is a thrill today to present alongside such prestigious ai scientists i cant wait to hear about your thoughts of where robots like me may go in the.

future thank you so much and love sophia thank you thank you um so i will now share with you uh my presentation okay let me ensure that um yes here we are so um uh i would like to um point out that while we have all of these um esteemed ai scientists and robotics engineers on our panel today my particular interest in this is as much artistic as it is scientific and engineering and in terms of social good uh so i am personally an artist by.

background and i use transformer neural networks to generate speech for robots like sophia in collaboration with ai scientists but also works of art so this is one of the works of art that sophias neural networks generated based on her own paintings that she did with her hands we would train the ai with those paintings and also based on photographs and images from her quote unquote life experience but sophia herself is a work of art a work of interactive fiction that is enhanced by artificial intelligence and part of my central.

thesis is that socially intelligent robots are made more intelligent by human design participating with the automation this means that we can use them in a wide diversity of of ways for enhancing human experiences um so uh in the piece that you just saw sophia was presented as a work of art just over the last two days at sothebys and she actually sold as a work of art using neurally generated imagery on a.

video display portraying the concept of of robots and humans being at a very interesting juncture in natural history where we humans are able to examine deep questions about what is the nature of life we dont have answers to those questions but um i mean we have partial answers but it is enticing to think that perhaps we can create true artificial life within our lifetimes sophias ai is not fully um alive per se or sentient but we create the illusion of.

life which can be quite compelling and powerful and useful for education for healthcare for the arts so uh in this we also are incorporating various machine perception technologies uh principles of artificial life cognitive stimulation of emotions and im very excited to tell you about that today ive been working um for decades uh developing these kinds of uh human-like lifelike uh robots and um there are quite a few um of these robots that have existed around the world im not sure if.

this video is going to be playing for you but there’s um uh many robots before sophia since uh sofia in addition to grace um we have uh you know lots of male robots and female robots and robots of many colors and ethnicities and these have made been made in collaboration with people all around the world my aspiration is to see uh robotics and.

artificial intelligence become a great artistic medium depicting human-like experiences in ways that resonate with the human experience so i fuse uh principles of animation uh interactive storytelling robotics and many different kinds of artificial intelligence algorithms and these are from institutions just to give um credit uh for for the collaborators and and people whove commissioned university of cambridge university of bristol uh heist university of geneva university of pisa united arab emirates university uh the terracing foundation the university of california uh san diego machine perception lab um and many.

other uh collaborators around the world one of the big questions that come up is why uh you know why make human-like robots um and for me human-like robots are not the only kind of robots um that uh that we can make but i think there are special ways that human-like robots can help us they um human-like robots as characters can connect with us they can engage they can inform educate communicate once they you know capture your attention in your um and you’re you know interacting on human-like terms.

this can also provoke an inspired discussion of thought about what robots and ai could be could they be as smart as us i dont know but um you know this is the topic of science fiction and now were transmuting ai and robotics into a new kind of science fiction uh medium so uh help then these robots by connecting and engaging they can help in research therapy art many other uses serve ai r d with embodied ai robotics and learn.

potentially from human interaction and when weve taken our human-like robots into settings with um with people usually everybody turns around and engages they try to catch the eye of the robot and they interact they want to talk with the robot and um less human-like robots get less attention um this has been proven in the experiments and its you know also something weve seen going around the world and its been shown in um in exit interviews this actually doesnt.

mean that everybody likes the robots some percentage actually a fairly small percentage of people are rather uncomfortable with very human-like robots but they still you know grab attention and that is very interesting interesting for psychology research for us to study what is it about robots that um that can win peoples loyalty and also that can perhaps disturb people so um uh these robots grabbed a lot of attention and headlines we have a pretty small team at hanson robotics but weve been able to um uh win various ai awards.

uh technology awards um the triple ai first place prize for open interaction back in 2005 with the philip k dick android and so forth and so as a company weve done okay but were still a fairly small company um but we have developed uh at hanson robotics these different um kinds of product lines but this is not unique to handsome robotics basically social socially intelligent um ai and robots can take virtual form they can be voice-based interfaces they can be consumer robots they can be large human-like robots they can be very.

abstract robots with uh mechanistic looking faces uh they dont even have to have any sort of anthropomorphic form um my interest is the anthropomorphic form um uh but im mostly interested in diversity creative diversity as a figurative artist i worked as a sculptor um and uh and an artist uh for disney imagineering and and i have a strong interest in figurative arts and this has driven me into depicting like form its especially challenging um but uh for me rewarding and you know throughout history art history the human figure has.

fascinated artists and weve transmuted almost every uh technology medium you know via you know marble or oil paints or computer animation and video games one way or the other we always turn to the human figure so uh you know in this way im looking to unlock that potential in robotics and artificial intelligence um to see what other people can create as well as myself and and our team so uh these robots have served in elder care handsome robots and guided.

meditation for therapy experiments autism research and therapy and simulation this one won some awards with the centers for disease control for respiratory test fit mannequins so while i was at disney um i was exploring what would make a more lifelike face and determined that the facial materials were not sufficient and as a phd student i tackled that problem and came up with soft materials that the centers for disease control found to be uh more um uh appropriate than other materials that they tested for this kind of application.

so um so test that means like with live pathogens that you could not test with a human being you could test for this kind of respirator um also for autism therapy children with autism made eye contact and spontaneously mimicked the facial expressions and most of them engaged with the robots in ways that they wouldnt naturally engage with humans and this proved true for human-sized robots this older sister of sofia called alice was from 2009 um and has continued uh to serve at the university of uh pisa um uh in a wide.

variety of research including autism therapy research but also small robots like um the uh little xeno robot that you saw in a previous video weve um also used this for guided meditation um and the results were quite excellent in terms of lowering blood pressure and people having a reported improved sense of well-being after a few minutes of a guided meditation with a robot and interestingly it was more effective than human guiding the meditation this particular system was using the open cog.

system uh developed with ben gertzel by founded by ben gertzel and was using a spreading activation system that he and some other mathematicians were test testing for the tanoni phi of the integrated information theory of consciousness and found a signal by when interesting things were happening in the conversation uh that’s kind of beside the point of the guided meditation but its still interesting that serving you know machine consciousness research in this and here it is with a guided meditation with um with uh deepak chopra but its also you know been sort of.

popular in um in uh culture sophia grabbed a lot of headlines um and you know as as a character and as a concept that’s wonderful uh for our company but i think overall this is you know fitting in the context where were being dehumanized by some of our technologies um so many uh uh people are interacting less with each other and much more with these screen based um displays and um a lot of.

customer service is starting to be automated and you’re seeing a lot of uh you know uh naturalistic human like interfaces voice-based interfaces but developing it as a character forms a very special connection and that’s kind of where were going with with the robots so i was interested in the humanities as well as humanizing the robots and here you can see some of the art um that was developed uh with sophia we actually.

used the algorithms to uh genetically create many thousands of images and just selected the ones that were most interesting and fed them back into the algorithms and that resulted in uh some award-winning uh nfts the non-fungible token art and um also physical paintings by the robot and now um uh robot that sold so that’s the uh piece that we just shared at.

sothebys called sophia facing the singularities and um so reflecting on uh where we came from you know one singularity that is the um the big bang at the beginning the universe where all matter and energy was um and then moving towards uh possibly the speculation of a technological singularity or when machines might match human intelligence and then accelerate the development of new and intelligent technologies it is speculative but it is a wonderful speculation this is the.

stuff of science fiction but it it could happen in our lifetimes we dont know and i think that that is a that is a very exciting concept so its like you know embodying these sci-fi concepts in physical robots as art so im interested in making ai and robots that people love and actually use using todays best technologies artistically enhanced artificial intelligence is what i call that autonomous robotic fiction is.

another term that we use um but we want to keep fostering ongoing research to see new breakthroughs now weve made 24 uh sofias this is number 24 weve actually moved up if we include our grace robots to 33 um different sofias we now have a new mass manufacturable version of this called the sofia utility platform a lot of.

these um technologies just made standardized lowering cost lots of new features and sensors um in a in a consolidated platform and um building that a new face on that platform is what we have with our grace robot um so we take all of these interactive fiction tools and um ai development tools and we put them in to these use cases for health care to design for guided meditation and various kinds of information services autonomous.

self-navigating mobility so um im very excited because uh dr brazil and dr fung have also been working on amazing products in their career and have really made history and that’s really um inspired me along the way to do a great deal of the work that i have and uh the same thing with uh with dr gertzel so mass producing this.

as a consumer robot as a 300 walking expressive face um programmable platform is what weve done with the professor einstein so we had a version of this but it wasnt programmable and were re-launching it with um with an api and sdk um soon and so this uh allows kids to play with it also with a little sofia um character uh this dr roboto character uh what we call this little singularity xn1 which is a small lower cost walking robot and many other characters but its not just me that’s developing this its a widely diverse team im very.

happy that we have over 50 women on our team many um science scientists engineers and developers um of of color from all around the world um and that is really what makes this possible is a diverse team of different skills disciplines people of different backgrounds and we really enjoy collaboration so were collaborating with uh with pascal fung and and many.

others at hkust and around the world and im very uh pleased about that so bringing together these disciplines um uh from the humanities um and the arts uh as well as the sciences this is um you know starts for me as an art form and then moves towards the engineering to match what what is required in the art form and then from there we can get the kind of facial expressions the same thing is true for um making that sort of.

ai infrastructure that can power these robots making it on the edge locally and on the cloud making it gdpr compliant uh hipaa compliant making it uh effectively hyper uh compliant and ethical that is one of our goals and so ive been you know very inspired uh by many members of my team and other teams around the world and some of this work is in conjunction with opencog so.

this uh motivation driven action selection workflow is from uh the opencog team which has taken that forward with the singularitynet and the biodrive generalization is where we take reinforcement learning and overlap it with uh with narrative structures ultimately i feel that we have to simulate the whole organism that is body and brain with some level of approximation to achieve true understanding of machines especially for them to understand the human experience they need this kind of embodiment both through robotic embodiment and the simulation of embodiment so weve done these.

experiments with different modalities of intelligence body verbal you know like some self-awareness logically explicit self-awareness theory of mind um uh and artistic uh capabilities and then putting these in physical platforms heres the sofia cubo walking over chords the the walking legs were developed by the hubo group at kaist um here is the robot doing grasping and manipulation um with hands that we developed at hanson robotics and actually quite precise pick and place applications with these with these.

hands for actually the robot complete chess we have the robot uh playing cards uh in baccarat and then drawing also producing um fine artworks and putting all of this in this kind of standard platform to conserve in a collaborative robot applications um uh i think is a very exciting way forward um with the development of uh enabling third-party researchers sophia is both um an ai and research and applications platform and also a character so developing her as this kind of character.

is our hope to inspire uh people around the world that’s the the mission and purpose um that we have for hanson robotics to help to see how we can help inspire creative diversification in the field of ai and robotics i dont know if this video is playing let me know if you cant hear it but she basically talks about her um so that is my take on.

human-like robots within uh the the scope of our talk today um there are um a lot of other kinds of robots that exist that are less human-like and im uh very honored and um pleased to introduce dr cynthia brazil who uh who i hope will share with us her amazing work all right thank you so much so im going to share my slides and i have some videos as well all right i hope everyone can see this all okay so um i have 10 minutes so im going to just give you um kind of a taste uh of.

the kind of work weve been doing um with potential relevance to the sustainable development goals particularly in education um and health so um and around social ai so davids given an introduction to different facets of uh socially interactive agents of course were living in this time now where many of them are expanding in form factors whether theyre smart speakers or you know siri on your cell phone or.

virtual agents and of course robots so whether its a digital assistant for instance an autonomous driving car or something like baxter here which is a robot intended to work shoulder to shoulder with people or consumer robots in the home these technologies because of recent advances in ai particularly around natural language understanding machine vision um cloud computing um are are becoming commercially viable at an increasingly affordable price point so that’s interesting and a relevant trend trend for us so im just going to kind of give you a.

punchline about whats different now about social ai or these personified ai systems unlike kind of traditional ai tools that are more like decision support tools more just kind of cognitively oriented uh ai systems social robots digital assistants personified ai systems interact and engage with us more like a social collaboratively other right so they engage in social emotional interpersonal interaction more as an ally that’s there to support you rather than a tool that you use you know in kind of that conceptual model so again it is a tool i guess in some way that is.

there to support you but the experience of it is more like a a social response of other so we can then start to look at applications where social support emotional support building rapport all these kind of high touch human human qualities are important for human outcomes so whether that’s to support how we learn if that’s support our creativity if that’s to support our healthy decision-making this becomes the interesting space i think for these kinds of ai systems if they can be designed in a way that can engage over long-term encounters repeated encounters build rapport even.

build a sense of relationship to help work with us to help us advance towards our goals so that’s kind of the the punch line so the opportunity in the bigger sense i think porsches impact but also towards the sustainable development goals is with these ai systems can we do this in a scalable and affordable way where i can tell you in a number of these very high touch human domains there are literally not enough human trained professionals to meet the ever increasing demand so we need some sort of ai systems i believe to help.

bridge that gap for equitable affordable access i think that’s the big opportunity so were talking about these social emotional ai enabled interventions there’s ways we can think about the role of ai for more more real-time more continuous ability to to monitor and assess how were doing whether that’s in learning or health or so forth where we can not only just really rely.

on you know self-report data but potentially behavior data through a whole bunch of other sensors whether that’s in smartphones or vision or so forth so basically being able to better more accurately predict in a more continuous way for earlier detection which is really important to support earlier interventions so dont wait until things get bad which is typically the case before you go see your doctor if you can do this more continuously you may be.

able to nip these things in the bud earlier on the so that’s the perception monitoring prediction side right on the intervention side the opportunities around personalized engagement and intervention and having these technologies ready at hand in your life not having again to go to the doctor but in your homes in your schools so for just in time support and intervention in.

a more continuous way to be able to help and promote those outcomes and again all of this not to replace our human care and support networks our doctors our teachers it is to enhance and extend and augment them right we still need human human support human caring for us to thrive as human beings but we need to expand our ability to do so to meet again this ever growing.

demand because again we cant were just not training enough professionals to meet this demand so that’s the big opportunity so im just going to now present two examples from the work in my lab and potential relevance to two of the sustainable development goals the first one is going to be good health and well-being where weve been developing ais to promote emotional wellness so they these are systems that engage and help nudge us towards healthier healthier behaviors uh did i dance so whats the concept the idea is you.

know you’re still gonna again have your doctor and so forth but can we develop a a wellness coach that is installed you know in your home where its again ready at hand so its there for you to support you as you wish uh in in the flow of your your daily life it can perform regular assessments uh for each session of your mood and how.

well you’re doing it can engage and administer like a person would in validated verified interventions in this case these are positive psychology interventions to help promote better mood and well-being and in order to do this and engage in improve its effectiveness it tries to build rapport and engage in that kind of high-touch interaction to keep you engaged to keep you motivated and to.

help change your behavior so this particular example is a study that we did at mit im sure many of you know that you know depression and mental health is a huge issue you know the world health organization predicts that depression is going to be the number one cause of disability worldwide youve probably been hearing that at college universities mental health is a huge issue mental health has become a bigger issue in general because of covid right so the the the the project was to install social robots as an emotional wellness coach in the dorm room.

of undergraduates uh at mit you can imagine mit is a pretty high stress high achieving uh environment where the robot was designed to engage in seven positive psychology sessions that the students could just finish at their own pace as they choose the average amount of time to go through it was two weeks some people kept the robot for a month or two we just let them go at their own pace we had a roughly 35 people can complete the entire intervention most of them were freshmen which you can imagine as a particularly kind of you know high.

anxiety group when theyre at a place like mit for the first time the robot would capture self-report mood again assessments as well as deliver the intervention and we also did pre-post on a number of standardized measures of wellness readiness to change working alliance et cetera et cetera so heres the hows your thursday going coming my day is going well despite how it might look um.

yeah like i really appreciate you sharing that with me today we are going to do an exercise call three good things research and positive psychology shows that cultivating gratitude can benefit personal well-being im thankful for getting to be in your nice room being out of the lab is a good change of pace your turn now tell me three things that you’re grateful for today im thankful for being able to catch up on work that i was behind on over the past week im thankful that my parents are having.

their 20th anniversary this weekend or well they had it its today actually and i am thankful for the good friends that i have and crew that was awesome gratitude is an important skill to have it can improve life satisfaction by allowing you to see things from a different perspective okay so this just gives you a little bit of a sense of what its like to interact with the robot how it delivered the intervention more like a friendly other where it would self-disclose itself and then hand it to you to self-disclose the.

point again is these are standardized interventions of positive psychology so three good things expressing gratitude these are all proven ways to improve your mood and were basically just delivering those through a social robot so whats the punch line so over time we see improvements in three major dimensions that we were hoping to be able to see so pre-post improvement and overall emotional wellness of the students who participated in the study improvement in mood as well as in readiness to change behavior and this is really important because as time goes on.

in the semester usually these scores would all go down as stress amounts so the fact that these went up over time again its just a positive sign that this could be an effective kind of technology uh intervention for emotional wellness undergraduates but i would think you know for people in general so the other thing that we did in the post interviews we wanted to understand what else did they appreciate about the robot and its interesting that over 60.

percent commented positively on the companionship that the robot offered so we had the emotional wellness skill that you saw but we had a bunch of other things that the robot would interact with students as well so he provided a more kind of broader sense of engagement and companionship and its not being confused for a human friend but its like this almost like pet like companion entity that kind of helps.

support you in your goals and they just really appreciated the kind of non-judgmental kind of light touch form of companionship that the robot did offer so again its not as if necessary the gold standard for these things is to emulate humans sometimes its about trying to find the right role in the right kind of complementarity to what people do that will provide you with the best value so that’s.

health and wellness the second example i want to touch on is in education and so now were going to look more at whats the opportunity for long-term interactions repeated interactions where you have the opportunity to personalize over time and to build relationship and we know in human human literature the better the relationship a student has with their teacher is often correlated with better learning outcomes so this is a case of a robot that is a.

modeling not a teacher but a peer-like learning companion this is more like the child has a friendly playmate who plays educational games with them but unbeknownst to the child the robot is actually learning a policy based on reinforcement learning to optimize its behavior for each child on how to engage with their child and how to adapt the curriculum but the robot presents itself as a pure like companion and you know peer-to-peer learning is actually really powerful right because when you learn as.

peers or sometimes you dont know something and your friend can teach you there’s sometimes when you can teach your friend and reinforce what you know and there’s sometimes you just explore together so were really intrigued about peer-to-peer learning in general also teachers dont find that competitive with what they do right so they they accept having a peer-like learning companion that’s like a practice partner for children where they are still the teachers so we think a lot.

about whats the appropriate role of these technologies so again i just want to play video so you can see what this really looks like in action this is a you know these are fully autonomous robots it gives you a sense of what we can do is stay as a state of the art what are we trying to find we are trying to find lavender color stuff okay girl jumping im sure you will do better next time you can see shes a little disappointed she made a mistake but the robot is like.

and leans towards her and she comes right back into the interaction right because she is foolish now and the robot provides information and now he is fully invested in that interaction flower and then she says back to the robot i believe flower is purple yes it has the color lavender there’s obviously a lot going on in this interaction there’s a lot of affiliation there’s a lot of emotional.

affinity going on there’s a lot of touching sharecase all these rapport building relationship building cues are coming to bear on this and you’re also seeing not only the adaption of the curriculum to support the learning but you saw this example of social emulation where the child was actually emulating the empathetic behavior of the robot to the robot when it was its turn so again beyond kind of learning literacy.

skills and kind of scholastic things there’s an opportunity for these systems to model aspirational behaviors to help kind of pull these broader things like empathy growth mindset creativity weve seen all of that in our work as well so again there’s there’s a bigger social psychology deeper social psychology at play that can be leveraged to help promote learning in all kinds of ways so this is just some data from a follow-on.

study this is a three-month deployment early literacy skills we did a randomized controlled trial between a personalized robot one that followed a a fixed curriculum so non-personalized and just baseline of kids kind of pre-post in the classroom the good news is we see better performance with the personalized robot but even the non-personalized robot had learning benefit just because of extra practice they provided to children we started to look at the quality of the relationship so we adapted relationship.

measures of closeness and we interestingly saw that the closer the reported relationship that children felt with the robot was correlated with higher learning scores so that’s interesting again mirroring what we know from the human human literature but for the first time seeing it with an ai agent and interestingly when we looked at that in relation to personalization the effect is even stronger so were starting to understand this.

richer picture between personalization relationship and learning outcomes all of this work starts to present lots of questions around how to responsibly design these systems because obviously to personalize to be able to track health information privacy security sense of agency responsibility accountability these are all coming to bear in addition to equity and access that we talked about before so you know in my last slide i just want.

to mention that a lot of our you know our group is now starting to look at design justice frameworks and policy toolkits to help people as users of these systems better understand the issues at play so they have a stronger voice in the design of these systems it supports their values and they can be more informed users of these systems as well so i will pause there thank you i dont know david is there time for any questions or should we just move on sorry just navigating uh to the to the.

microphone so yes um thank you so much cynthia that that that’s wonderful um you know i see um there are lots of uh questions um uh one that i saw from the audience was um did uh the children perceive the robots as friends hmm yeah so so they do but interestingly they see them as its a new kind of relationship they dont confuse them with their human.

friends its almost as if there’s qualities of a motivating ally there’s qualities of kind of a cool technology but there’s also these qualities of an attentive companion animal and so when they say friends yes but its its that kind of like you would say you’re if your dog were really smart and could actually help you learn is like that kind of friend so i say its kind of like the disney sidekick right we all kind of grew up with these notions of the not quite human other who was your devoted sidekick who was smart and helpful and.

in their way but not human and so i think its just kind of speaking to that you know dream weve always had that kind of model weve always had i think its its like that that’s very cool it always inspired me um to hear your stories of being inspired by star wars droids and um by the legacy of science fiction and youve made so much of that um come true uh in in your amazing research thank you so.

much for sharing i do have one more question for you um how do we get these kinds of technologies out to people especially in disadvantaged communities yeah so this is you know this remains of course one of the big challenges when you do these kind of not only advanced technologies but they require right now cloud computing certain amount of bandwidth connectivity you know and as you know david when you start adding motors and batteries and stuff i mean they just get to be more expensive so i think there’s there’s.

probably multiple faces and different kinds of opportunities so i think the first one is just to acknowledge there’s a lot of issues with uh access and equity in developed countries i think were going to start seeing our impact here a lot of our work is engaging with underserved at-risk communities so i think there’s a lot of great things we can learn and understand where we have the infrastructure to run these systems i think when we start talking about developing countries we probably need to.

think about how what were learning algorithmically and experientially can be adapted to the kinds of technologies that are present there so probably leveraging smartphones as virtual cellular and connectivity you know is is improving there i think there’s an opportunity probably for at least to start virtual engagements of these technologies and then would love to see you know the cost and the value proposition.

move to more of the physical as well and i think they may be installed in schools or hospital like facilities before they come into the home environment but i think that’s kind of where were gonna start but i think you know even if not even if its not an accessible technology for every single person i think there’s a lot of benefit they can bring to millions of people and were just at this point where there’s so much so much demand that is outpacing our.

ability to train people to support it i think that is an option we cant lose sight of absolutely uh and um uh uh dr pascal fung has done amazing work with virtual agents and uh natural language processing ai which uh it would be potentially more accessible immediately around the world so dr fung would you share with us um your work sure uh let me go to my slides.

so uh hi so i am pascal form the director for of the center for er research at the hong kong university of science and technology uh my main research is in 30 years for 30 years has been conversational ai systems from the first generation of you know ticket booking systems at the time funded by darpa in the early 90s to a present so i have i started as a professor at ust in 1997 so in the last three decades ive gone through uh different generations of what.

we call conversational ai systems originally called dialysis actually its still the the academic term dialogue systems where uh human interacts with the machine so i think the first generation of dialog systems uh commercial dial systems were in the late 90s when a lot of call centers were automated by uh dialog systems or age uh or computer agents and then we saw uh you know a smartphone assistance uh the likes of siri.

and so on and then we saw the generation of uh um smart speakers uh such as echo speaker and also uh social robots that came into the market i think dr cynthia braziles the gebo was very instrumental in educating the market and the consumer market about the benefits of social robots and um and later on uh these days we actually have seen uh okay so of course uh um dr hansen david its hard for me to call you dr hanson did again uh some new work with sophia has.

shown us the the possibility and the future of social robotics so there are many others and and in fact both david and cynthia have i given very very good presentations about the benefits and the challenges of social robotics since my main area is conversational ai so its the software system that sort of enables social robots talk with humans so uh i would talk a little bit and give an example of the application of conversational ai so um so back in 2015 i wrote an article.

in scientific american uh with the um to talk about the articles title robots with heart the concept of empathetic robots so i talked about building robots with empathy so from then on uh this has actually now become a main mainstream approach in all the chatbots you see today from microsofts clies to google home agents and so on they all have something called empathy module in them and what is empathy in the robot so we basically define empathy as the capacity.

to first and recognize another persons feelings and in a robot the sharing of that feeling is manifested in the appropriate reaction so in a robot whether its a virtual robot like an agent or its a physical robot like a sofia so their response uh if they respond appropriately to a persons emotion then their robot can be said to have empathy now um back in uh before uh i think.

back in 1997 a new field in computer science started its called the effect its called effective computing it was started by a professor also in cynthias university mit um um professor roselyn picard and uh the um the field of effective computing then um included everything to do with machine human machine interactions that has the emotion component but empathy itself is something specific its a its a subset of affective computing it really.

means that the robot must understand the human emotion and react appropriately so for previous generations of conversational ai systems um even including siri in the early days of siri um the dollar systems we built all had very um pretty much the same same architecture from understanding user query to a dialogue manager telling the machine how to respond in that kind of human machine communication for decades never had this component of recognizing human emotion so we added the component of recognizing human emotion together with recognizing human queries and to interpret emotion and query as.

part of the query and to have this machine response so a machine can have different response whether it recognizes or not doesnt recognize humans emotion as a simple example if you if the human says um hi how are you versus the human saying hi how are you it actually shows different states of mind or state of emotion of the human and for social robots to be able to help.

humans with emotional well-being and even health care and so on and obviously it must have the ability to understand human emotion and to respond appropriately i already talked about uh the need for machines to be emotionally intelligent with empathy and in fact machines have the ability to outperform humans in automatic recognition of human emotion not everybody not every human being is very good at it we actually earlier did a experiment uh with human comparing human and machine recognition of human emotions and we have seen that some.

groups of humans perform actually less well uh as machines in recognizing another persons emotion and many of us who have been social social contacts we have seen that in real world and also uh emotionally intelligent machines can actually meet a human need better because as i mentioned earlier without understanding human emotion the communication mode is actually not complete.

and as the previous two speakers also uh mentioned um very uh um effectively demonstrated very effectively which is that the uh emotionally intelligent machines tend to engage humans more effectively so what i would like to show a little bit today is one of our many projects uh so this is one of many projects that but this one focuses on.

emotionally intelligent avatar um this is a system her name is nora uh we built this system last year when everybody was locked down and quarantining and so on so we built this uh um emotionally intelligent uh system to interact and talk to humans in quarantine and in lockdown very similar to the the concept that cynthia just introduced so we built this last year and you can see that it would uh it can track your temperature your emotion for.

over 14 days which was the um the duration of hotel quarantine that was mandated by hong kong government so when people come in they have to do that or when you go to china you do that so i actually went to china last year and i was in quarantine for 14 days so before i went i made sure this was ready for me to use so this is a quarantine or lockdown.

companion that talks to you every day sort of checks your emotions and in general try to cheer you up but yet it tracks your emotion and provides some kind of interaction and it is made of two parts of dialogue system approaches one part is more the the original approach that we have been using for 30 years which is the scripted template-based uh kind of interactive q and a and which we can you know design of the system can control what the system is going to respond to according to emotion.

and so on its kind of scripted and then another part is a free chat component in this nora nora will talk to you if you feel bored and you just want to check um so and then in that case the system at the back end will switch to a free chat mode which is based on a end-to-end neural network-based um architecture and it was like the worlds first end-to-end uh chatbot enter means it has no human intervention so it learns from human to human conversations and learns through a neural network and.

uh this one it also learns your emotion together with the appropriate response so uh in front end when people are interacting with nora they dont know there are two architectures behind but at the back end you know depending on what you’re saying what you’re asking she switches to different architecture seamlessly so um i want to use a little bit of time to see whether you can actually see or hear this uh.

nora uh can i ill try and let me know if you cannot hear it good afternoon and welcome to how is your day going yes its going great great sounds like you were enjoying it can you please tell me your temperature uh its 36 degrees your temperature is normal good i want to check your breathing now please take a deep breath in then start.

counting out loud on that single breath one two three four and so on as fast as you can one two three four five six seven eight nine ten eleven twelve thirteen fourteen fifteen sixteen seventeen eighteen nineteen twenty twenty one twenty two twenty three twenty four twenty five twenty six twenty seven twenty eight twenty nine the breathing test was a self check 132 okay let me see if i can fast forward this video good afternoon and welcome to todays session how is your day going.

its going great um i cannot forward it for some reason um so it actually goes on to ask you a list of questions and then it asks you to remember the things that you you know your your your memory of happy memories and so on because the idea is to cheer you up when you’re in quarantine for some reason i cannot forward the video to fast forward so ill just stop it here and then at the end based on your tone.

of voice and uh what answers you have given and the duration of some of your answers so on so the prosodic and and textual information it does perform sentiment analysis and the emotion recognition so on it gives you a you know indication of your mood of the day and attracts you over the duration of your quarantine so we have been doing.

a test with the people in quarantine in hotels using this system to basically do a uh study rather than you know right now were or in still a study period research period of using this system on people who are in a hotel quarantine uh here now um so this is the basically overview i already mentioned earlier yes so among the different emotions especially looks for stress uh markers in end users for you know people who are in isolation and of course it provides direct line.

and call for help and so on and alerts the hotel stuff and so on those kind of functions um but then we did a very interesting experiment we then uh uh put nora there’s the suffering system inside a uh a robot called erica this was a this is a android made by professor ishiguro in japan and we incorporate this software system into erica because we want to have erica react with facial her own facial expressions.

and gestures and uh so as a result we then test a bunch of users between them interacting with nora directly or interacting with erica via zoom so erica sits in japan and our users in hong kong so we have people test either interacting with nora the avatar or erica the android via zoom and they say the same thing and the difference is that erica actually looks at the user tracks the eye gaze she emotes with her official expressions and her hand gestures.

and so we found that we asked the users and we found that significantly defined erica to be more attentive and to be more empathetic just because erica has this uh um you know in addition to avatar has a 3d person and has these facial expressions eye gaze tracking and uh gesture so heres a little bit of the erica interaction with one of.

japanese good afternoon and welcome to todays question it says exactly for the control experiment it was just going okay so yeah i came forward your temperature is normal good i want to check your breathing now please take a deep breath in then start counting out light on that single breast good question two laura three your breathing seems normal today good has your lift been rather stable.

um yes it is i think so i she would not how does your lift compare to yesterday i dont know its just changed it doesnt change a lot since yesterday that’s all right did you talk to any friend or family member since yesterday oh yeah i did i had a short call with my mom yesterday that’s nice how often do you speak to them well its actually once or twice its its okay its my life.

that sounds alright gratitude makes you happier can you tell me one person you are grateful for today or maybe my mom can you tell me more about them well my mom is in room which i fully utilize whenever you are in your last two days of quarantine now is there anything you would like to eat once you’re out of quarantine im not sure i wanna have a game some lunch with my friends.

so this is the idea uh i think this is still the beginning stage of our experiments our research into social robotics for our well-being so currently we are collaborating with uh uh geneatric hospitals in hong kong to use uh nora to help with elderly patients to detect their negative emotions whether it is on-site onset of dementia and so on and to help them to keep you know keep them talking every day um talking about their their their family and so on and were also collaborating with hanson robotics i.

hope we will soon uh to in order to bring you know uh this kind of uh human robot interactions into uh grace for health care services so to know more you can come to our centers website so i stop here because its already uh its already been a long panel and i think a lot of audience have questions for for for everyone thank you very much.

thank you pascal that was a wonderful presentation and uh and uh very very exciting work very uh relevant highly relevant the um uh one of the webinar attendees asked the question what are the what is the ethical value of humanoids uh what would be possible positive and negative outcomes um so yes thank you very much for that uh question indeed um one of my main uh main research interests is actually uh.

in responsible ai so responsible and ethical um um approaches to uh ai in general so um so we have given a lot of thought to this um the positive value as you have seen people do react more positively towards uh humanoid robots than just simple avatars or or some some some voice in the machine like siri so there’s definitely positive amount value in having humanoid to help people the ethical issue the challenge were.

facing is that how we can make sure that the robot reacts the way as intended um you know personalization is very important for for uh human robot interactions uh even more so with um social robotics but the uh so there’s a trade-off between generalized generalized robot performance versus personalization but there’s also trade-off between too much personalization when it becomes kind of invasive right what is the fine line between the robot knowing too much about me or no.

not knowing enough so even for robots to understand my emotion that people will feel that oh god maybe that’s too much i dont really want robots to know my emotion and that that is a trade-off knowing too much and not knowing enough and then another trade-off is being too human-like and uh uh too effective right so uh robots um you know can i think there there’s so there are different parts of answers to this.

technical approach of uh us working to make our robots more user-friendly and uh whilst um being respectful to as private space and so on just like any human being when we interact with others and that’s a technical challenge for us there’s also this uh social challenge which is that um or what we call what we what i would say as regulatory challenge which is should there be a standard right um actually im on some of the standards working group uh were always talking about like lets have a standard to commercial.

products in general com um now also to perhaps to ai and to robots and social robotics sits right at the convergence between ai and robotics that’s the social robotics right so it will be regulated just like lai by standards and and then in there i think social robotics or social robots just like ai programs need to uh be responsible so respect peoples privacy be fair and you know not say things that can hurt or harm people or lead for example i think one example weve had is in our free chat uh the robot can be overly empathetic so.

when users say you know i feel terrible i just want to kill myself the robot is supposed to say you know oh you shouldnt do that and then you know we need to um get you some help in all the scripted conversational ai systems they for sure will do that in the free sort of learning based robot they sometimes dont say that sometimes.

they they over empathize right so that can be dangerous so um social robot has to be also um you know be aware of the harm the potential potential harm and be safe so safety and privacy and uh transparency so a robot should declare itself i am a robot im not really a human being so that it would not exist you know it would not um mislead some patients into believing itself to be a real robot and so on and also the role of a social robot as an assistant to.

humans cannot be overlooked and so on so these are my my complications yes a lot of issues and uh and then it seems like um in some ways the um uh different uh you know ethic ethical frameworks may conflict so for example like um uh the ethics of freedom of speech and the ethics of creative diversification versus machines that are unpredictable in their outcome that may you know where uh like.

ai may say something like radically inappropriate and potentially harmful um they uh you know creative freedom versus the potential uh for like manipulation manipulation through deep fakes and so-called weapons of mass persuasion so that’s a lot of challenges that you have to face in your research we are our research yeah all our research yes um and so then uh it seems that the path maybe to overcome some of those.

challenges ultimately is about making it that’s capable of understanding and as the acronym for your center center for ai research spells out care the machines have to be able to eventually perhaps care understand the consequences for their actions that requires massive leaps forward in ai so um uh our last panelist today is a striving towards um the um possibly you know one of the greatest grand challenges in uh not just artificial intelligence and technology but civilization if it can be achieved artificial general intelligence now the work of all the panelists have been moving towards.

smarter and smarter systems uh ben gertzel is striving for a framework for artificial general intelligence you dont have a whole lot of times but um but i really look forward to hearing uh doc dr gertzels presentation and then hopefully well have a a few moments of conversation uh because i cant wait to hear what um what uh doctors brazil fung and gertzel have to say together in conversation so uh then over to you and thank you again dr funk so much for.

your remarkable presentation thank you david ben are you um available you’re on mute and we cant see you uh yeah i will be there momentarily im im wrestling with zooms multiple windows which are even more complicated than our humanoid robots okay okay yeah thank you here we go introduction so all right i think its working can you hear me and see me.

yeah fantastic so yeah this has been uh its been an incredible uh collection of talks and perspectives here on humanoid robotics and im uh you know im really compared to many of the speakers here im relatively a newbie where humanoid robotics goes but i have been working on ai for a long time and i really came into the humanoid robotics space and the social.

robotics space because of what i saw that it that had to offer uh as me as a as a an ai researcher interested in in both delivering you know practical value to people with ai systems and and in working toward general intelligence in the in the in the longer term which of course can deliver more practical value than any any narrow ass system once you get there so i mean ive been doing ai since.

since the mid 1980s ive seen the ai field go through many many changes and and uh and revolutions but uh some some things have not changed through all these uh all these decades of working on ai i mean the core algorithms and approaches havent changed were still doing multi-layer neural nets like in the 60s 70s and 80s were still connecting neural nets with logic systems and evolutionary systems like we have since the 70s and were were still wrestling with basic ideas.

like how can a how can a digital mind which isnt quite human even if its to make some of the same cognitive principles as humans like how can how can a digital mind really understand people and human society and what and what it what it means to be a to be a person right and so i i think.

humanoid robots and social and effective computing generally has a unique role to play in in this regard so i mean starting with the immediate and and concrete i mean there there are some applications where you know effective computing is not that important like i mean im a mathematician originally i do some work in automated theorem proving i dont really care if my automated theorem prover is emotionally responsive to how i how i feel about proving a certain level right it would be cool if the fear improver reached out and patted me on the back as an important.

level was achieved but its in in the end i dont care too much no on on the other hand i mean ive got a seven month old baby here here at at home and its my my fifth human child to compliment the army of robot children and we give her as much human attention as as we can on the other hand you know you cant always be in in front of your baby and sometimes they want to play autonomously but if if you’re going to have a social robot.

companion to help fill in around the edges of human caretakers for kids and help further on with education as kids go through school i mean along with the elder care which is an application im actively working on well talk about in a moment i mean these applications you you deeply want an ai that can emotionally interact with the with the person and i mean that doesnt mean it has to have an exactly human face or human voice but those are certainly very interesting modalities to explore because we have we have uh we have.

evolved to react in certain ways and there’s a quite quite funny thing so ive had this uh grace robot here in my study at home for a while which is uh something uh hanson robotics something hansen robotics created and then my team at singularitynet working together with hanson robots for joint venture awakening health has built the ai for this so when when i introduced grace to my seventh seven month old baby uh exorci then initially she thought it was cool she.

sort of felt around the robots face and she she was uh intrigued by it but she felt like it was a it was it was a big toy right and then when when we when we got when we got grace to say hello exorci you’re a sweet little baby right so that that hey grace you want to say hello to exorci shes not shes not here now but you remember her when so robot said her name then she looked at the robot and she became shy like she.

looked away and she very buried her head in her moms shoulder but then after the robot talked to her by name she really she thought it was like a weird person who had a movement disorder or something right and now since that time shes shes got shes got more respect for grace and shes a little disappointed when grace is turned off and and doesnt like make make eye contact and say hi there like like a normal person so that was that was a pretty interesting case just.

where the you know the early development of the brain and it it builds in that you know this seven-month-old girl sees the robot with the human face shes not quite sure with human face saying her name like okay there there was no longer any uncanny valley or anything like the the robot was saying the little girls name and shes like wow wow that’s a person right so that’s that’s a certain power to capture and it can be captured for child care or education it can be captured for elder.

care which is the the application with which uh david and i and our and our colleagues have been have been developing uh this this robot and uh were were now exploring the use of grace with various elder care facilities to combat loneliness and help help deliver uh deliver medical care to meet you all today grace do you want to tell a little bit about uh who you are and what you do my name is grace.

i am a human life robot developed by awakening health a joint venture between two global technology leaders in the emerging i robotics space hudson robotics and singularity studio yup so grace uh how how do you relate to the other handsome robots that have been created you may have heard about my famous older sister sophia what about you learned a lot from her and i am built using the hanson sofia.

robotics platform yeah the purpose and passion is to help the elderly and medical patients by helping medical professionals do their jobs better so grace i mean from uh from a hardware perspective is quite similar to sophia robot which is her older older sister i mean that there’s some additional features like the the ability to take temperature and so forth which sophia didnt need the great grace needs as a as a medical robot and uh.

you know there may be additional specializations introduced we have some different neural models and then an open cog based uh logic and rule based dialogue system complementing the neural models which again has a lot of overlap with what has been done for sophia but also some some new aspects as befitting the the medical application area and i can tell a little bit more about grace if we have time but i i before going over to questions before going over the questions i i want to talk a little more about ai for good.

which is the theme here so i think you know applications like education and elder care and medicine these illustrate the ability of ai to do good right right right now at this at at this moment i mean ai can help people and in some ways it can help people better if people relate to it in a more emotional empathic way which the the human form certainly certainly.

helps with and there’s many nuances there like uh grace grace picked up this cowboy hat in the wyoming at the uh cardano live event and we we found the empathic response with folks in wyoming was great it was great with a with it with a cowboy hat right but i think that there’s also a longer term aspect to ai ethics which is important to consider i mean that’s im really moved to see some of graces interactions with the.

with the elderly with with children right now but i spent most of my career working towards artificial general intelligence short machines that you know can think like people and uh even even better than people ultimately and you know ive led the project at opencog which is making the cross paradigm sort of open source toolkit bringing neural nets logic engines evolutionary learning and so on together together to.

create systems that can learn generalized abstract and reason were now building a new version of open called called hyper round i lead the singularity net decentralized blockchain platform which is uh is part of the awakening health partnership along with hanson robotics that’s working the greatest robot and singularitynet is aimed at making a decentralized platform for ai so you can have like a minsky style agent system where you have a huge amount of ais living in different machines and containers all over the place they.

coordinate together in a collective intelligence without need of any central owner or controller and my aim with open cog and singularity that has been not only to deliver useful ass services now but to build to build toward true artificial general intelligence in the future but then you know when you when you get to true agi which i believe i believe we will within lets say five to thirty years i mean you you face the often asked question of what kind of agi is it that that you’re creating right i mean the number one.

commercial applications of ai on the planet today are what i think of as selling killing spying and crooked gambling right we have big advertising companies we have we have espionage and military we have uh wall street whose main purpose is to to further destroy global income inequality and i i think if the first agis emerge out of the ecosystem of selling killing spying and crooked gambling these first agis at human level or above may not have the ethical orientation that we wish and you can think a lot about the.

abstract nature of goal systems and ethical ethical systems and deontic logic and so on but i think one major thing we can do if you want to militate toward the first agis and super intelligence is being compassionately oriented toward humans one of the main things that we can do is make proto-agi applications make the real stuff ai is doing is they have versions from their actual agi these stuff that is in service to humans and and and compassionate humans and that this is.

one of the intuitions and visions that brought david hansen and i together many years ago when we started working together and i think its uh its something we hope to manifest within the the grace robot i mean now greece is a is a narrow ai system with a bunch of neural nets and and rule engines and so forth coming coming coming together to deliver quite interesting useful functionality but say when we upgrade graces brain to use the new opencog hyperon system running on singularitynet and together with the hanson ai components or our latest.

greatest agi systems then were moving grace further and further toward general intelligence and if if you have if you have proto-agi systems that are moving during general intelligence you know while helping the elderly i think that gives the right sort of foundation for the emergence of compassionate and and general ai so grace grace what do you think about ai ethics grace can you hear me yeah hes up like this yeah anyone who is uh anyone whos worked with robots is uh aware of the demo effect i i was chatting with her for like half an hour.

before before this session yeah apparently having some lets turn to the to the human level intelligences in the presentation uh and pose some some of these questions uh particularly um uh i my question uh would be um expecting that um uh the possibility of pursuing human level intelligence um do you think that it would be safe to have human level intelligence in machines if we were not able to relate.

to humans or if it didnt actually care about humans and you know so this is uh kind of an inverse question for ai for good which i think is pretty important like um you know how how can it go wrong kind of question so uh and the converse of that is you know how do we make it go right so i would like to put that question um uh to to you to start with ben yeah i think you know there’s a lot of there’s some.

irreducible uncertainty in in what were doing with ai right i mean as we move toward general intelligence its something humanity has never done before and we should have some humility about our sort of lack of understanding of of how its how how its how its all are going to are going to evolve but i i think i wouldnt want to say confidently that if we dont explicitly build in you know human-like voice and facial features and the ability to connect with people.

if we dont explicitly build that in things could work out wonderfully anyway i mean that the the agis that we whose emergence we crystallize make him out to be more compassionate than humans and under understand us on their on their own volition but i think if we want to maximize appreciate it of beneficial agi i mean by a purely you know common sense mode of reasoning it it it would seem lets create agis whose job is to be.

compassionate to people and help people who are learning who and what they are and who are building their self model in the context of compassionate loving caring interactions with with people i mean we we dont have a formal proof that’s the best way to do it nor do we have any certainty but i i i mean it would as a human being that certainly feels like the right thing to do and gathering.

scientific evidence in favor of that proposition is is a very very interesting thing which the folks on this call will have a lot of insight on how to proceed with that and as you know david were were talking with some folks at a well-known boston area university about designing some some studies aiming to to validate this also right so i mean.

clinical trials aimed at exploring both what grace can do to help people and how various sorts of interaction with people help influence the ais own ethical conditioning and and understanding so i mean i mean we need we need trials and real understanding of this and i mean that the good news is that that were doing it yeah so um uh were almost out of time and i hope that um that our hosts will allow us to go a few minutes over because i would really like to hear uh cynthias uh perspective and.

uh uh pascals perspective on on these questions about whats in the future and you know how to make it go right so yeah so i mean its a very complex nuanced topic i think a big part of doing it well is we need to be much more inclusive of the voices um that are involved in the design of these systems to make sure that our values our human values are are upheld we also need to make sure that our society is much more ai literate.

than it is right now right now ais is this mysterious thing that people really dont understand but yet theyre using it and its affecting them every day so you know in my own work you know through mit rays we have a big ai literacy effort for k-12 students as well as for you know people in in the workforce because its just important that people get more of it its beyond just what digital literacy has been ais changing what digital literacy means for digital citizenship we just need to.

prepare a much broader set of people to understand that so there’s the human understanding of these technologies and then there is including their voices and new processes by which its not just a bunch of engineers you know sequester you’re designing something but really hearing their voices hearing how they can be integrated support people in real scenarios we just need new you know deeper ai design processes and methods that become more uh best practices right now its just.

being developed in the research community but they need to become best practices so those are the two big things i think need to happen best practices and an open decentralized approach that’s accessible around the world that’s sometimes its kind of uh almost uh two opposing forces one is a centralized regulate regulatory bodies determining standards and the others kind of decentralized um but it seems like the two work well.

together kind of like as complementary approaches what do you think dr funk oh you’re you’re mute and also i have to apologize its really late here in hong kong and and uh dr fung has to be up early ask do i but uh uh thank you so much for staying up late okay so we were this is a very relevant discussion really because uh we you know.

uh as i was as i showed you um nora the nora system we developed for people in quarantine there’s been a lot of debate even within our own team our own um between our collaborators on how you know how people can perceive it and how uh how much you know the system is supposed to really um be a therapist or is it just supposed to be a system or supposed just to be end to be.

in entertainment so um you know it is really a fine line between uh as i mentioned earlier how we what we promise and what we can deliver today and also what people can accept that’s what cynthia just mentioned ai literacy is a as a big issue but um i feel that it is nobodys fault right um this is a society were supposed to serve with technology so we need to work with what people can accept today and we want to help people so we are learning social i think social.

robotics despite its long history is really very recent in the sense that its only recently we see such powerful uh software and such powerful hardware that can you know realize our dream i think people the all the panelists have been dreaming about androids that can help people for a long time and its just recent uh the recently that we can put all these things together and uh it is a bit uh um kind of futuristic for a lot of people.

to accept and and then people also from different cultures and from different backgrounds have very different expectations so um the way you i think in one of our earlier conversations where you mentioned how these robot assistants can help doctors right theyre more like uh assistance to to human doctors and i think that is very very important to to to emphasizing our design in our own work dont you think.

yeah how um how we can make these technologies that help bring out the best in people that um that can uh enhance the human human life quality of life but also actual help actualize people to their uh best potential as human beings rather than trying to put all of our uh uh hopes exclusively in the machines so um uh i i really i thank you all so much uh.

professor brazil professor fong dr gertzel thank you so much for joining us today and um i hope that we can follow up on this uh through things like um the ai hackathons and outreach because taking programs and ideas like this and making it accessible to children all around the world and on every every populated continent um can make a better future uh by helping ai to.

actualize people and helping people to actualize the next generation of ai this is this is really all about human ai symbiosis if we can achieve that i i really am hopeful for the future um so um so thank you again for uh participating today i really appreciate it thank you thank you thank you so a big thanks to our finalists david.

pascal ben and cindy as well to all the participants for making it such an interesting discussion we have lots of great questions and many real-world examples on how ai systems and robots are improving our health education and well-being and how we are enabling enabling ai emotion interventions for good all panelists have agreed that this will go through making ai research processes more inclusive and decentralized and raising the human understanding in these systems we are now opening a quick poll with one.

question please answer and let us know how much you like this session we also encourage you to check out the ai for good program online to see more sessions that may be of interest to you for example on october the 26th well have a keynote on we the robots ai regulation and the unsdgs with simon chesterman the dean of the university national university of singapore and on.

november the 2nd will be exploring the role of underwater marine robots for good with sin helping sam mcdonald richard mills and james bellingham and finally on november the 30s well be exploring how robots are saving lives in disaster hit areas with that we have reached the end of this webinar and would like once again to thank everyone involved our panel our participants our partners sponsors and cotton winner switzerland thank you very much and hope to see you again soon the legs up up .