We recently chatted with Susanna Dillenbeck, Commercial Partnerships Manager at Furhat Robotics, to learn more about the impact Conversational AI and Social Robots are having on our lives, and their potential for the future. Furhat Robotics is a Stockholm-based startup building the world’s most advanced social robotics and conversational AI platform.
Topics explored in the discussion include:
- The Meaning of Conversational AI and Social Robotics
- The Positive Impact of Social Robots for Businesses and Society
- The Challenges Faced With Social Robotics
- Gender and Social Robotics
- The issue of being Emotionally Connected to Machines
- The Future for Emotion AI
- Women in AI Challenges
- Advice for Women in AI coming from a non-technical background
🎧 Listen to the podcast here
▶️ Watch the video here
Nikita RE•WORK [00:08]
Hi Susanna, thank you so much for joining us for the RE•WORK Woman in AI Podcast today. I wanted just to firstly ask if you could tell us a bit about the company that you work for, which is Furhat Robotics. It was founded in 2014 I believe.
Yes, exactly, so we are a social robotics startup from Sweden originating from the Royal Institute of Technology here in Stockholm, and the company was actually never meant to be a company. So we are the result of the research of our four founders and our current CEO, this was also his PhD project, and it just turned out that the research that we made within social robotics was quite groundbreaking and people from all over the world started reaching out to the group and saying, OK, where can we buy your robot or your prototype? And then the company was born as a demand out of that. So we've been around for almost seven years and we are still based here in Stockholm, Sweden.
Why the name Furhat Robotics? What does that mean and where does it come from?
So, Furhat, it is implying the hat, fur hat, as you can see here behind me. And it is a very simple, but quite a funny reason. When the founders that I mentioned, built their first prototype, which was really hackie, their cords sticking out all on the back, they were invited to London, to the London Science Museum, to a big exhibition there. And the night before they left, they felt like, oh, our prototype looks really, really ugly. What are we going to do? Are there going to be a lot of journalists there? And someone found a fur hat lying around in the office because it was winter in Sweden and it is really cold. So they covered the prototype with the fur hat and brought it to London. And ever since then, it was remembered as the fur hat robot. So then when eventually the company was founded, they decided to call it Furhat Robotics. So we still call our robots a Furhat robot. So there's a very simple explanation to it.
It is a simple explanation, but I wouldn't have guessed it. And for those of our listeners and viewers today, for the YouTube edition that we're also doing, that might be a bit less familiar with the terminology of conversational AI and social robotics, can you just tell us a bit more about that?
So typically when we talk about conversational AI, we mean technology that enables conversation between a human and the machine and that can be speaking conversation, it can be written conversation, Siri on our iPhones is one example of that, Google home device, or the chatbots that you type with on several websites that can all be grouped into the very broad term of conversational AI. And then when we are looking at social robotics, we are talking about physical robots for social context. So we are using the conversational aspects, but it is more than that. It is also a physical embodiment. As you can see here beside me, these are turned off, but these are some of our old robots. So we don't only have the chatbots online, but we give it a physical representation that can look different in many ways, and we have the conversational aspect included in that. Does that make sense?
Nikita RE •WORK [03:22]
Yes, thank you. And why do we need them? Why do we need social robots? And what positive impact can they have on businesses and also just on society at a broader level?
That is a good question, and I think in many situations in life we don't need a physical robot as that would just make things more complicated. It's enough to have Siri on the phone when we're driving in the car or when we're out running, for example. But if you imagine the difference between having your mother read a bedtime story for you over the phone or being there physically present, which would you prefer? I think you would prefer the physically present. Or imagine the difference between doing an interview over the phone, whereas when you're sitting with the interviewer at the same table and he or she is looking at you in the eyes and being in that physical space. So we strongly believe, and there is also some evidence, that the physical presence is very important for certain social contexts. But we also know, especially for covid, that sometimes it's impossible to have a human there for logistical, financial or other reasons, safety reasons. And then we think that a robot can be a very good complement to that. So when the social and physical aspect is very important, but we cannot have humans there, there we see that these social robots could be a complement.
Nikita RE•WORK [04:48]
And what about the challenges that you face with social robotics and in particular, perhaps the issue of trust. Is that something that kind of comes into your work a lot?
It does, and I think one of the biggest challenges that we see is that when people who have never interacted with a robot before and they see the physical embodiment that looks a bit like a human, it has eyes, it follows you with its gaze, then the expectations of what that robot can do is very high. It's almost the same expectations as you would have on a human. Compared to if you just see a smart speaker, then you typically understand that there are limitations to what the smart speaker can do. So one of our biggest challenges is for people to understand that technology has only come so far. And even if it's advanced, you cannot just walk up to a robot and say whatever you're thinking about and hope that that robot can return that answer. So that would be one of the challenges that we face. But coming back to your question of trust, I think what we see in a lot of places in society with new technology coming, is that it is increasingly important to understand what happens to the data. Is there a visual recording that is saved somewhere? Is there a speaking recording file that is saved somewhere? What is happening with all this data that this robot could be collecting about me? And that's, of course, up to each and every use case or application to make sure how do we build that trust into the users and for different applications it is different levels of importance, I would say.
Nikita RE•WORK [06:26]
And what about the challenge of gender and social robotics?
I think we can look at that from different perspectives. I should have turned a robot on here in the background. But one aspect, I think, from the gender perspective is to make sure that we have robots where the user doesn't look into designing a specific persona so that we as platform developers make sure that there are faces that can be male and female and non binary. There are voices that can be male and female and non binary, gestures can be made and differentiated. So that would be one perspective to allow the user to customize the robot in different ways that is locked into certain genders. And related to that is also to make sure that when we're creating personas for chatbots or for robots, that we really think about, are we confirming any prejudice that is related to this gender or are we being even more open minded here? So that comes more into the design phase. And lastly, I think one important challenge that we also see as a company is to build diversified teams and make sure that we don't only have a team of programmers and developers and designers here that are, for example, male between 30 and 35 and come from Sweden because that would give us a very narrow perspective. So I think also in the team that is developing something we need to make sure to have different perspectives included to make it more interesting and diversified.
Nikita RE•WORK [08:09]
And what about the importance of emotionally connecting to machines? Why is that important?
Well, I think it does not always apply that we need to connect emotionally to machines, if I start on that end. I mean if we think about going to the train station, buying a ticket in the vendor machine and running to the train, we don't need an emotional connection there, we just want to get that ticket so that we can run, right. And I think sometimes when we talk about emotional connection, it feels or it sounds like we need to fall in love with the robots or have very strong feelings towards it and I'm not sure if that's necessarily the case. So I think we need to look at what different use cases do we have and what is the purpose of the use case. So if we think about keeping lonely elderly people company, the purpose of such a use case, if we would use a robot in that case, would be to make this elderly feel safe and feel listened to and seen. So in that context and that specific use case, yes, then the emotional connection becomes very important to create a feeling of safety, which is the entire purpose of the robot being there. And if we look at an interview situation, for example, if we have a robot interviewing people, then the purpose there is to get as honest and transparent answers as possible, so there we need the user to connect emotionally, to feel very open and transparent and honest. So in different cases, the emotional connection is very important so that we can reach the purpose or the desired outcome of the thing that we're trying to achieve, whereas for other use cases, it's not really important, then it's more transactional. And we just need to get from point A to point B as fast as possible with as few errors as possible. Do you get my point? Do you agree?
Nikita RE•WORK [10:14]
Yes, it's really interesting to have more of a split between those different areas because, as you said, not every area where social robots will be applied requires the same things. That kind of touches on my next question, which is more looking at where we're heading, where are we now with Emotion AI and also where can we expect to be in the next two or three years or perhaps even longer term?
I think an important thing to say first is that when we're talking about AI detecting our emotions, it's rather technology that is looking for clues of how we could potentially be feeling, because just as between humans, it's super difficult to really know what someone else is feeling and thinking. We're only looking for clues if that person looks sad and tired or jumpy and happy, and it's the same with technology, we can look at it from three different aspects. One would be the visual aspect, such as if we're smiling, that can be detected by a camera, so typically that means that we're happy. If we're looking really angry or our eyebrows are going down then that could be a sign that we're not really happy. If we're turning our attention away, then we're probably not very interested. So the visual aspect is one way of detecting these clues. Secondly, we can look for spoken aspects. So if I speak loud and angry or if I use certain words that can be filtered or seen as either angry words or happy words, that can also be a clue to how I'm feeling. And lastly, we can have physical measurements such as measuring the pulse or sweating of the palms etc. to see ok, am I nervous, am I going up in speed or am I very relaxed?
So these three groups of clues, the visual clues, the spoken clues and the more physical measurements are ways that technology is used to detect how we as humans are feeling and our emotions. And over time, I think the base will become more sophisticated, more integrated to create a fuller picture of how a user is feeling and reacting at a certain point in time and also connecting to a specific individual so that we can have a better tracking over time. For example, if you're speaking to a robot, that robot will remember who you are. And also remember the last time you spoke, maybe you were feeling a bit sad and down, or you were angry and you showed signs or clues of anger. And how are you greeted the next time as compared to a person that was maybe showing happy signs the last time they talked? So combining these, making them more sophisticated, adding them to more software and then tracking that over time would be my prediction of how we can see this evolve over the coming years.
Nikita RE•WORK [13:13]
That's really useful to be broken down like that. Moving a bit closer to the heart of this podcast, I guess, looking at Women in AI. So what are there or have there been any challenges that you faced as a woman working within this field?
Well if I take a step back, AI for me is a really broad word that can mean so many things. And if I'm just looking at my own company and also women that I know that are in this field, they can be programmers, they can be founders of a startup, they can be researchers, designers. It's very broad. And I think these different groups face different kinds of challenges. One thing that's been important for me personally is that I don't have a deep tech background. So initially I felt that oh, this area is not for me, AI wow that feels very advanced and you probably need to program for 20 years, and then as I went, I realized that, oh, wait a second, this is a new area for everyone. So it means that it should mean that we're starting on an equal level. Of course, there are people that have fantastic experience, but many of us are starting at the same level. And there shouldn't be a difference between men and women. Something that has helped me a lot is to find these different communities, whether it's women in AI, women in tech. There are fantastic communities that help girls how to code and to find these groups and feel that I'm not alone in this quest, there are others and we can exchange experiences and knowledge and really help each other on that journey. So that has been the most important for me to feel that it's not a one woman race, and it's not only one that can succeed. We need more women in tech and in these new types of technology and we need to work as a group to get there, at least that's what I think. And how can we become part of these communities and really boost them and have a very positive energy and help each other and recommend each other for speaking slots and all of that. And seeing that movement has made me very, very happy because that's not something I've seen before.
Nikita RE•WORK [15:37]
I'm not sure what it's like in Sweden, but in the UK, we've definitely seen quite a growth in the number of women in AI, women in data science, women in tech groups emerging, and especially this past 12 months or so. I think because we've had the opportunity to make them virtual so that they're more accessible and not just perhaps in your own local or wide area, it can be global as well. So it's great to be able to tap into that expertise and to connect with peers from anywhere now.
Definitely, I agree and I think it's the importance, and I'm speaking out of my own experience as well, just to feel that, ok this is something that no one has done before and no one knows how to do it, so maybe I'm actually a good candidate for that. What would make me a less good candidate than anyone else because if no one has experience of it then we should all be equal in that. And now I think it's time for us. as women, to also demand that opportunity because we are all trying something that no one has done before.
Nikita RE•WORK [16:48]
Definitely. And for anybody that's looking to get started in AI or to progress their career, that perhaps is also coming from a non-technical background. What advice would you give to them for the best first steps for them?
Ok this is going to sound so cliche, but if I'm just reflecting on myself. As I said, I don't have a deep technical background and for me to take the step to become a programmer, that would have been huge and not really something that I'm passionate about, then I would only be doing it for the sake of saying yay I'm working in AI. So I needed to find an area where my background and my strengths made sense. For me it was coming from a background as a management consultant and being good in project leadership and mapping processes and seeing the bigger picture. So finding that strength and also seeing ok which area, I mean AI is so big, but which area am I passionate about. And to me that evolved around the conversational aspect. I'm really interested in communication both between humans and also humans and machines. So once I narrowed down both what are my strengths and what do I think is really interesting in my personal capabilities and what area within new technology do I find interesting then I could merge the two together then at least I had narrowed down the different ways to go a little bit.
I was lucky enough to work at a company where I could explore the conversational AI area a lot so that was a lucky thing for me. But if you don't have that possibility in your everyday work then I think find some of these communities like I said, women in tech, women in AI, there's women in robotics, there's women in voice. Of course you can join groups where there aren't women either but these groups I find very good. There's a community called The Voice... where they discuss different topics of conversational AI for example. So try to Google and find these different groups and just go in there and say who you are, ask for help and I think the community will provide a lot of help, especially if you've narrowed down a little bit what you're good at and what you're interested in.
Nikita RE•WORK [19:06]
Yes that is definitely a great place to start with, what you're interested in and passionate in and then digging down a bit more to become an expert in those fields, because, like you said, it is quite broad ranging.
And sometimes I find that these words become a bit of buzzwords. All of a sudden everyone feels that, oh yay, I need to work in AI. But first of all, what does that mean? And is that an end goal in itself, or is there actually something else that you're passionate about? Maybe you're passionate about horseback riding and you start a barn, but then you realize oh I can automate something here and then I need to deep dive into the new technology. So let's start with the passion, and I think don't go for AI just for the sake of writing on LinkedIn that you're working in AI. I would see what are you super passionate about and how can the advancement in technology help you reach that passion? And sometimes that might be through ways that you had never imagined yourself when you started.
Nikita RE•WORK [20:10]
And that passion always comes across and it certainly has today. So thank you so much for taking the time to record the podcast with us today, Susanna. It's been great to learn a bit more about robotics. We had the pleasure of learning a bit more about yourselves and the rest of the team at an event that we had on Conversational AI just a couple of months ago, so we'll definitely be keeping up to date with your latest trends and developments and progressions. And I will also share some links for the website for anybody that wants to learn a bit more about you in the podcast notes below. So thank you again, Susanna. It's been great to chat with you today.
Thank you so much and take care.
Check out all upcoming Conversational AI events here, including the Conversational AI + NLP Summit taking place in April.