the ConversAItion: Season 1 Episode 2

Social Robots & the Uncanny Valley

Jim and Gabriel talk social robots, AI-powered robots that look and communicate like humans. They discuss their development, applications and how people building them can navigate the "creep" factor. Gabriel is a Professor in Speech Communication and Technology in Stockholm and also the co-founder at Furhat Robotics, a company that builds human-like social robots.
Listen on Apple Podcasts badge
“Social robots—it’s something people have dreamed of for a very long time. But it’s only recently been possible to build something that’s actually useful, of course much thanks to advancements in speech recognition and computer vision.”
Gabriel Skantze headshot

About Gabriel Skantze

Gabriel is a Professor in Speech Communication and Technology at KTH Royal Institute of Technology in Stockholm, Sweden. He is also the co-founder and Chief Scientist at Furhat robotics, a company that creates social robots—AI-powered robots that look and communicate like humans. Follow Furhat on Twitter @furhatrobotics.

Short on time? Here are 4 quick takeaways:

  1. Social robots are autonomous, AI-powered robots designed to interact and communicate with humans. Furhat is a social robot with human-like expressions and advanced conversational AI capabilities.

    Autonomous robots have been featured in popular culture for decades, appearing in films from The Iron Giant to Terminator 2. Now, thanks to technological advancements, social robots are a reality—and advanced to the point of being useful. 

    Furhat Robotics, a company that develops human-like social robots, has disrupted the industry’s traditional focus on mechatronics (how robots move) by instead prioritizing the quality of the interaction with humans.

  2. A disconnect between appearance and movement is the root cause of people’s uneasiness with human-like robots.

    The “uncanny valley” is a concept that describes a human encounter with a robot or a computer interface that looks real—but not quite real enough. 

    Often, social robots have a super realistic appearance, so you expect to see the same lifelike accuracy in its microexpressions. But microexpressions are very difficult to achieve, so usually there’s a disconnect. They might, as Gabriel noted: “look stiff and sort of smile in an awkward way.” (Yikes.)

    To combat creepiness, Furhat aims to match a robot’s appearance and behavior. Their robots are human-like, but they’re not super realistic. The movement and expressions are sophisticated, but the face is a little cartoonish. The effect is similar to Pixar computer animation.

  3. Right now, typical applications for social robots are found in the service industry–but use cases may soon expand.

    Furhat is developing social robots to act as service agents in airports and train stations, to address travelers’ questions. 

    They are also working with a recruitment company to create a robot that can conduct early-stage interviews, the goal being to reduce bias against age, ethnicity, weight and height. Candidates would of course ultimately meet with a human interviewer, but the robot would eliminate the chance someone is filtered out due to bias early on. 

  4. Social robots are intended to provide better service—not to replace the human service agents we already have.

    Social robots aim to complement human workers, not replace them. In schools, for instance, Furhat could help teachers by facilitating lessons in small groups of students separated by learning style.

    The Furhat robot is something you have to see to believe, see it in action on the Furhat YouTube channel, here.

Read the transcript

Jim Freeze  Hi, and welcome. I’m Jim Freeze and this is The ConversAItion, a podcast airing viewpoints on the impact of artificial intelligence on business and society. 

[UPBEAT MUSIC]

The ConversAItion is presented by Interactions, a conversational AI company that builds Intelligent Virtual Assistants capable of human-level communication and understanding.

In this episode, we’ll discuss the emergence and evolution of social robots. They’re autonomous robots designed to interact and communicate with humans. 

Join us as we dive into the development and status of this futuristic technology. We’re going to uncover how folks in the social robotics space address the fact that their technology is, to some people—a little jarring, to some people, maybe even creepy.

To help us understand the ins and outs of social robots, we’re joined by Gabriel Skantze, a Professor in Speech Communication and Technology at KTH Royal Institute of Technology in Stockholm. Gabriel is also the co-founder and Chief Scientist at Furhat robotics, a company that creates social robots. 

Gabriel, welcome to The ConversAItion.

Gabriel Skantze Thank you very much. 

Jim Freeze Furhat is a social robot that displays human-like gestures and emotions—even complex expressions. It comes with a selection of pre-built expressions and gestures which can be further customized to fit any character, and its face can be configured in several ways to create unique identities with their own personalities and quirks. 

Could you start off and just give us a quick history of social robots?

Gabriel Skantze Yes, of course. Social robots in general, it’s something people have dreamed of for a very long time. People have been trying to build autonomous and social robots, but it’s only sort of recently been possible to build something that’s actually useful, I would say. And that’s of course much thanks to the advancement in speech recognition and computer vision and so on that makes this possible. 

Jim Freeze How did you initially get involved in the space?

Gabriel Skantze Yeah so actually my background is in conversational systems, more like speech-based systems. I’ve done research on this for almost 20 years now. And of course we also worked with animated agents on screens that you could interact with. But there was always sort of something lacking there, and I think it was this feeling of presence that you have with another person. So, if you’re talking to an animated agent on a screen it’s kind of like talking to someone over Skype, which is obviously not the same thing as talking to someone sitting next to you in a room. So taking that leap from this animated agent on a screen to a physical robot is like going from a Skype or video interaction to meeting someone physically. And of course people often prefer that. They often might be willing to travel long distances and pay money and so on to actually meet physically in person. And I think the same thing goes for virtual agents or agents that we interact with. 

Jim Freeze So it sounds like your background really kind of had a direct impact on the creation of Furhat?

Gabriel Skantze Absolutely. And also I think most people working on robotics or that comes into the social robotics area comes with a robotics background. So they are thinking about mechatronics and how to make this robot move. So that’s kind of a different angle than me and my colleagues had when we were thinking about the actual interaction—that that was sort of the focus. And then we think that the physical appearance of the robot helps the interaction, it sort of enhances the interaction. It’s the interaction with the human that’s in focus. 

Jim Freeze Interesting. So I have to ask, because I’m a marketing guy I’m always interested in how companies get their name. Furhat, quite literally it means “a hat made of fur”? Could you tell us about the origins of the Furhat brand?

Gabriel Skantze So we started this and we were researchers eight years ago and we were working on this first prototype of what we thought a robot head should be like. So we were using the animation technology, projecting it from behind on a transparent mask so you could actually project this animation of a face instead of using mechatronics. 

When you’re doing that of course there were a lot of things sticking out from the head with a little projector in there and so on. To sort of cover it all up, we found this fur hat lying around that some master’s thesis student had forgotten. So we were invited to exhibit the robot at the local science museum—people really liked the fact that it was wearing a fur hat. So we sort of thought that that was a really nice thing that made our robot different from other robots from a marketing perspective, so to speak. Also, a robot is something that you associate with something cold and hard, and a fur hat is furry, so you get this sort of soft feeling, a fun feeling about it also. 

Jim Freeze How did you and your co-founders go about deciding the particular appearance and design and how has it evolved?

Gabriel Skantze Yeah so actually, it was kind of a side effect. Since we thought that we wanted this to base it on animation instead of mechatronics. We thought that we could do much more powerful facial gestures and expressions and we could do more accurate lip movements and so on that you don’t find in mechatronic robot faces really. Things moving in the face all the time, that is very hard to do. But it sort of had this side effect, since you’re projecting the face anyway, you can actually project any face, and you can change it in the middle of an interaction if you would like to. You can make the robot look male or female or even a dog or some character. 

We sort of realized that this is actually extremely powerful. Most social platforms they have decided on a persona for the robot, a fixed persona which comes with the physical design. Whereas we can actually make it have any persona that would fit the interaction. Because also Furhat is a platform that you’re supposed to use for different types of interactions. So if you imagine a robot working in a bank, for example, you might want to have a more serious impression. Whereas if it’s working in a theme park with kids, you might want to have something more fun.

Jim Freeze Well you know it’s interesting, I can totally relate to that when I think about what we do at Interactions with our Intelligent Virtual Assistants. Because one of the things we do working with our customers is we talk about the importance of that virtual assistant having a persona. Now, along with that comes some issues that that we see and I’m curious if you see them as well.

I recently watched a video produced by Bloomberg that features your robots. In the video, they talk about the “uncanny valley”—the idea that when humans encounter a robot or a computer interface that looks real but not quite real enough, we as humans get a little queasy, might even think of it as a little creepy.

We’re familiar with that concept at Interactions. We’re an AI company and we recognize the importance of figuring out exactly where the “creepy” line is, and when AI crosses it. Towards that point, in 2018, we actually worked with The Harris Poll to survey over 2,000 consumers and found that 40 percent of them find it more creepy than helpful when AI-powered customer service agent sounds or interacts like a human but doesn’t notify the caller that it’s a virtual assistant. For that reason we believe that companies using AI-powered virtual assistants should never attempt to mislead the consumer into thinking its an actual person. 

So with social robots—AI-powered robots that are purposefully designed to mimic human behavior—how do you address and navigate the “uncanny valley”? 

Gabriel Skantze Yeah that’s a very important issue of course. There are different types of problems involved here. One is of course, if you only have a voice—because speech synthesis is so good nowadays—you might mistake it for a human, or even if it’s like prerecorded. And obviously you don’t have that problem with robots. I mean nobody mistakes a robot for a human, now at least. So you don’t have to tell them, like yeah, I’m a robot actually.

Jim Freeze I don’t know have you seen the Terminator movies? 

[LAUGHTER]

Gabriel Skantze Yeah, thankfully we’re not there yet. But we have these super realistic looking robots, I won’t mention any names, but with this kind of rubbery skin, that look very human-like if you see them on a picture. But once they start to move they don’t look very human like. So there you have the problem that the appearance doesn’t really match the movement or the behavior of the robot. You have sort of this super realistic appearance and then you would expect these micromovements and so on—microexpressions—and you don’t find that, so they look stiff and sort of smile in an awkward way. It looks like the living dead or something. 

Whereas I think that in Furhat what we are doing is trying to match the appearance with the behavior. In the sense that it’s still human-like, it has a human face, but it’s not super realistic since it has this back projection and often the face is often a bit cartoonish and then the movement is actually quite sophisticated with micro movements and gestures and lip movements and so on.

So I think there is a much better match there in terms of appearance and behavior. More like if you watch a cartoon movie like Toy Story. It’s not very creepy. And I think that’s what makes Furhat not so unnerving. Of course, there are some people who think that still, but most of the time when people walking up to Furhat they don’t seem to have that reaction. And even if we try it with kids and so on, they seem to be attracted to the robot. They want to walk up to it and interact with it and maybe touch it and so on. 

Jim Freeze Actually you’re kind of hitting on my next question, which is really, how do people react when they first see it? How have those reactions informed what you’re doing with the product?

Gabriel Skantze So one reaction. Many who have seen it in picture get sort of surprised when they see it in real life because it’s an animated face, so they sort of expect it to feel more like animated face. The physical appearance, the feeling of presence is a big difference. What I can say is it’s typically not a good idea to try and make it super photorealistic, but something sort of more cartoonish I would say. 

Jim Freeze So what are some of the commercial applications of the Furhat social robot, for example. What are some customers doing with it?

Gabriel Skantze We are for example developing a social robots for airports and train stations. We’ve had it at Frankfurt airport, we’ve had it in Tokyo, and soon I think Berlin train station. If you go into a train station or an airport you might have questions like is my fight delayed, or where can I find the bathroom or is there wifi around here and so on. So then it’s kind of a service agent that you would find in these places. That’s kind of a typical application, and you could imagine something similar in hotel reception, for example—checking in customers.

But there are sort of nontypical applications for example, we’re working with a recruitment company in Stockholm developing a robot that can do interviews—the first stage of the recruitment process. Their main motivation is to reduce bias in recruitment. So the idea being that the robot would behave in the same way towards everyone.

Jim Freeze That’s really interesting. I hadn’t really thought about that. I was trying to think about how someone coming into a job interview might react to that. It’s an interesting idea in that you’re right, it’s not going to be biased. 

Gabriel Skantze We can do is to make sure that the robot doesn’t have access to certain information that you typically might be biased against—maybe age, ethnicity, weight, height and so on. So it will basically conduct the interview in the same way with everyone and then that interview will result in a transcript that then a human recruiter can read and make an assessment without even seeing the candidate. Of course I should stress this is early stage in the recruitment process so they will eventually have to meet someone but they won’t be filtered out due to bias early on. 

Jim Freeze That’s a fascinating use case. Do you predict social robots are going to become fairly ubiquitous at some point in time?

Gabriel Skantze I think so, definitely. I think there are so many places where you’re looking for service and you don’t find it. If you walk into a store and you want to ask someone a question and you’re waiting in line and you want to get help. There are so many of these kinds of situations where I think social robots could be truly helpful. 

Jim Freeze Given the use cases you’re talking about do you ever encounter individuals who say, “Oh, you’re taking jobs away from humans and giving them to social robots?” And if so, how do you deal with that?

Gabriel Skantze Absolutely, that’s of course a common reaction. And since the robots are human-looking, it’s even easier to make that conclusion—this human-looking robot is going to take the job of the human. But we don’t think that’s the case. Again we think there are so many places where we need more service than we can offer today. It’s more about giving more service than replacing the service we have. 

Another example that we’re working on is schools. It’s not to make teach unemployed teachers—on the contrary—we more so see robots as help. Maybe you split students or the kids up in different groups where they can practice with a robot for example. We see it more as complementing rather than replacing.

Jim Freeze No, I agree with that. What’s your vision for social robots into the future and specifically for Furhat? 

Gabriel Skantze We really want to become a platform for social robotics where people can experiment different use cases. People are coming up with very interesting cases all of the time—many more than we thought when we started the company. 

And we’re thinking a little bit like what Apple is doing in combining hardware and software to create a great user experience. That’s what we want to do for social robots, so we’re making both the hardware and the software. We want to create this great user experience, because that’s what it’s all about. 

Jim Freeze Well, I have to say. I have seen face to face one of the Furhat robots and I was amazed. It’s amazing technology. This has been fantastic, I’ve learned so much. Very much appreciate you taking the time to participate in The ConversAItion

Gabriel Skantze Thank you very much for inviting me. 

Jim Freeze Gabriel, thank you. 

[MUSIC PLAYS QUIETLY]

The Furhat robot is something you truly have to see to believe. You can check out a video of the robot on our website. 

On the next episode of The ConversAItion, we’ll learn more about voice interface design with “Seuss-uh” Jensen, she’s a senior voice designer at Adobe. She’s interested in how voice technology will change the way we interact with electronic devices – and our screens. 

This episode of The ConversAItion podcast was recorded at the PRX Podcast Garage in Allston, Massachusetts, and produced by Interactions, a Massachusetts-based conversational AI company. 

Well, that’s the end of today’s ConversAItion. I’m Jim Freeze, signing off, we’ll see you next time. 

[UPBEAT MUSIC]

Check out more episodes of The ConversAItion.