At the heart of an industrial revolution is an innovation that changes everything. Building AI Boston sees artificial intelligence as a renaissance. From the heart of innovation and the mecca of tech learning, we bring you AI for real people. A conversation for everyone. Today we are joined by Dr. Tom Williams, a professor at the Colorado School of Minds, who teaches about human robotic interaction and robot ethics. Tom's mission is the ethical and equitable design of language-capable robots. Today he'll share his AI expertise and the reality of how close we are to making robots a part of everyday life. Welcome to the show, Tom. Thank you for having me on. Oh, we are so excited to finally get around to talking about the subject of robotics. I think that when people think about artificial intelligence, they think of any number of sci-fi robots. So this begs the question, were you into sci-fi as a kid? How did you get to where you are today? Just in brief. Yeah, I mean, yes, I was into sci-fi as a kid, but I think that really had very little impact on my decision to go into robotics. In fact, when I was applying to grad schools, I was not looking at robotics programs at all. I thought I was not interested in robots at all and was looking just to get into the study of interaction and language and human cognition and sort of accidentally found my way into robotics. But yeah, I was into sci-fi as a kid and I still am. Okay, so who's your favorite robot in sci-fi terms? Chris and I ask this all the time. Well, I would say my favorite, the answer I give to that typically is my favorite robot would be fresh cut grass, which is a robot from a campaign three of Critical Role, which is a Dungeons and Dragons podcast. So maybe not the typical answer, but as an improviser, maybe not surprising. No, Chris, yeah, Chris can relate to the improviser. I'm going to get into your experience with Chris in a bit, but you also have a Massachusetts tie-in. So this is why we brought you here. I mean, you went to Tufts University. Yep. Can you talk a little bit about that academic background and how that Boston or Massachusetts, I should say Massachusetts ecosystem sort of shaped who you are now. Yeah, absolutely. I'm originally from upstate New York, from near Utica, and did my PhD at Tufts University from 2011 to 2017 in computer science and cognitive science. So yeah, when I applied to Tufts, I applied there because they had a cognitive science program with some of the world's top philosophers and linguists and psychologists. And it just so happened that the lab there that was really tied in with the cognitive science program from the computer science perspective was a human robot interaction lab. That was studying how people could, how we can design robots to interact with people more effectively through natural language. And it turns out, and I didn't really realize this before I applied to grad school, that, well, one, robotics is actually really cool, even if you're not into mechanical engineering. And two, that Boston is really one of the big hubs for robotics nationwide or worldwide, both because of everything going on, not just at Tufts, but in places like MIT and also UMass Lowell and other places that are outside of Boston proper that are doing some really, really exciting work in robotics. Yeah, that's so cool. And I'm glad you brought up all of that. We always give shout outs, of course, to our Massachusetts friends. And, and this is, you know, building AI Boston, but I wouldn't love it now. You know, Chris, maybe you can introduce this one how you met Tom, where you are in Denver in the really unique spin on, you know, human robot interaction. Yeah, so I do improv comedy. And I was part of a team called Not My Robot. That was just a team name that we came up with. Nothing to do with robotics. And one day we got contacted by, I had known Tom, but I didn't make the connection that it was the same person. And he contacted us and said, Hey, would you like to come talk to my robotics class? And the first, our first reaction was, you know, we know nothing about robots, right? That's just the name. And he said, No, no, no, that's fine. And had us come in, because apparently he had the brilliant idea to take improvisation skills and apply that to artificial intelligence and robots. And we went in and met his class and it was great. We got to see robots, one of one of which I think we're going to meet today. But yeah, it was pretty cool, Tom. I, I don't know how did you, how did you even make that connection to begin with? So it's sort of funny, the connection with improvisation came into my class long before I was personally interested in improvisation. So improv and role play is used a lot in human robot interaction as a design method. So imagine you're designing a robot for a hospital, right? And you know that the robot's going to have to interact with nurses and patients. You might say, Well, what should the robot say? How should it move? How should it act? Well, you try, you play the nurse, you play a patient, I'll play the robot. And we, we prototype out and role play out what what we think the interaction should look like. And by playing it out, we can say, Oh, no, wait, that's going to be really awkward, or Oh, no, the robot needs more room to, you know, to move into this part of the, this part of the room or things like that. And so actually for a couple of years, I'd been having improv days, like the one you came to in my class, but sort of on a smaller scale. And just getting one other person who wasn't actually performing a proud comedy to come into the class. And then this past year, after I had personally started getting into improv comedy as well. That created this network of people who I could who I could draw into to really like real experts to bring into the class to help you in exercises. And also open up these new new ideas and realizations about the sort of deep connections between the fields and the realization that, you know, in improv, if you're thinking about like, you know, taking an improv class, the purpose of those classes right is to take somebody who can talk, and has some knowledge of what to do. And so I think that's the same thing we do with robots that like if you have an LLM that's controlling a robot, you can speak, and it might say things that are sometimes accurate, right, but unless you give it the skills to control its eye gaze, its gesture, the prosody, the way it chooses to be polite or not the way it reasons about status all these different things we would talk about in the theater class. It's just not going to be a natural interaction. It's really funny that you put the way that you phrase that that teaches you to act like a real human. Yeah, because the reason I got back into improv was because I was trying to get into improv and I was trying to get into the way that you say that. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. And so I think that's the same thing we do with robots. Oh, that's really funny. And for the radio audience that doesn't have visual right now, Tom has got this mannequin head who oddly resembles Chris a little bit. So it's OK. Well, we'll just say it's a it's a plain robot face on a mannequin 3D head, which has features. OK, correct. I think it's a morph into that space and do a split screen for the radio audience. It's hilarious. If robots of the future are hilarious, we can thank Chris for that because obviously that will have impacted. You know, I'd like a funny robot in the hospital, but I don't know if I want you as my bedside manner, doctor, robot, but here we go. All right. Let's let's just like take it way back because I was really interested when I first talked to you, Tom, that robotics and the history of robotics goes way, way back to a very practical solution. You can talk a little bit about that pivotal moment in history. Yeah. So the idea of robots to some extent goes back like thousands of years, right, in the sense that people have been thinking about non-human but human like characters for a long time. Like in Greek myths or the myth of the golem in Jewish mythology. There's all these stories about non-human but human like beings. And then in the time of Napoleon, there was this famous, famous mechanical person that got toured around called the mechanical Turk. If you're a tech person and you've heard of Amazon's mechanical Turk, it's named after that platform. The idea of it was it, you know, people, all these famous people in history tried to play chess with this mechanical person and this tried to figure out the trick of it. And the trick of it was there was a small person inside of the machine controlling how it worked. And then at that point we had this mechanical person, but people still weren't really viewing it as a mechanical character that was doing labor. Et cetera that they could interact with. It was more of like a, you know, a crazy contraption. How does it work? What's the trick, right? And this really started to change in the 1800s in the immediate aftermath of the Civil War. And I discovered, I'm not the one who discovered this connection necessarily, but when discovering this for myself, it came about actually in part through studying the history of the Massachusetts. I am writing, I wrote a book. I just got the page proofs today. So it'll be published in a couple of months called Degrees of Freedom on Robotics and Social Justice. And as part of this, I was doing some research into the history of race and technology and how some of those connections in American society initially got made. And I kept reading these things about Lowell, Massachusetts, you know, and because Lowell was an epicenter of the abolitionist movement, Ralph Waldo Emerson writes in his diaries about Lowell, Massachusetts, about going to Lowell and about the connections that were being made there between race and technology. And I started to think to myself, I know Lowell, Massachusetts, I've been up there like half a dozen times to go to a big robotic center. And I started looking into the center and looking at what used to be there before it was a robotic center. And it turns out at that site where the robotic center is today, it used to be the foundry for the Lowell machine shop, which is where they were creating the technology that was used for the Lowell Mills, which in the 1800s were processing 33 million person hours of slave labor per year. So it was where the technology that was building that was facilitating the United States slave economy was being built. And that was like a really interesting and led me to look more into what was going on at that specific moment in time around labor, both in Massachusetts and more broadly, as it related to robotics. And it turns out, maybe the first robotic patent was in 1868, right after the end of the Civil War, in it was created by these two engineers from New Jersey for this mechanical steam powered man, pulling a cart. And if you look at the patent, the robot has pretty distinct physical features is wearing a fancy hat and coat is smoking a pipe has curly hair has it has these particular facial features and literary scholars have argued that actually that this robot patent was based off of a popular character from minstrel shows at the time. So it was essentially modeled after this caricature of black laborers. Oh, and the patent, you know, the patent didn't go anywhere. The I believe that. Oh, I should remember this but it's either for for that prototype, or the prototype or the the mechanical Turk ended up getting burned in a fire owned by PT Barnum in Boston. But the so the patent, you know, the prototype didn't catch on, but the way caught on was it inspired a book, which a couple months after this patent came out. This author Alice wrote this book called the steam man of the prairies that has the first appearance of a robot in literature. So it's the first appearance of a mechanical anthropomorphic, you know, non human person that is that is a real character that's being interacted with. The story is not great. It's about this young white boy who has this black mechanical man who carries him him westward, the kid ends up blowing up the the mechanical man in order to slaughter some Native Americans, they steal their horses return home. It's like it's not a great story, but inspired this whole early genre of science fiction literature one of the first one of the first science fiction sub genres of Edison AIDS, and that directly led to the literature, both the stories by the stick asimov and theater like the play Rossum's universal robots, which gave us the terms robotics and robots and shaped how we think about robots today what they look like where they're used and so forth. That's really that is really interesting. Now that you've given us that background. Because when I looked into the background of the word robot, it's from the check for forced labor, and it didn't really make that much sense. But now knowing the context of that that that really does it clicks into place. Yeah, and it's interesting, right? Because this is something that people in robotics sometimes talk about the fact that, oh, Rossum's universal robots, the play, which maybe, even if you haven't heard of it, maybe some people might have seen like Joss Whedon's dollhouse, which is all about the Rossum corporation, Rossum corporation. It's like a retelling of RUR in some ways. And this this play is all about labor. The center of it is labor, but it also touches on like there's an abolitionist character in it. It touches on race and gender and and these inter interacting sort of, you know, power systems, and people in robotics sometimes point this and say, Oh, isn't it interesting that he used the word robot which meant forced labor. But I think if you look at the history and say and see what he was building on, it's it's sort of a no brainer, because the past 50 years of wildly popular science fiction literature that he was building on all had this metaphor as robots as as a metaphor for for slaves. Got it. Got it. Um, I think this might be a good segue now to point out the robot that you've got with you today. We've come a long way. And again, for the audience, I can't see this robot is just a plain sort of mannequin had with with open it nods it said, Okay, this is not creepy at all. But and very human like features on this is interesting. Go ahead, Chris. So when I when I first encountered this is robot have a name, by the way. Yeah, so this is the fur hat robot fur hat. It's from a Swedish company called fur hat robotics. They'll put a fur hat because when it's called their US office is actually in Boulder, Colorado, though they bet they bought misty robotics. So the the US headquarters for fur hat is right up the road for me. Okay, so fur head. For it. Like a fur hat like a hat for over half I said for head for. Yeah, okay, got it. Sorry. Are you going to go over. For hat is when I met for hat for the first time. It does when you first see it it has sort of a vibe of the uncanny valley. But you pointed out to me if I'm not mistaken that it can change features right. Yeah, so one of the things that my lab really likes about this robot, even though I'm not sure in practice, we always want robots to be this anthropomorphic, this human like in every application domain. One of the things that we really like about it as a scientific product is the way allows us to to to manipulate the racial and gender presentation of the robot. So the face that's on it right now, which is so the face for those who can't see is retro projected which means that there is a projector behind the face that projects an animated face on onto this, this mask. Wow. And right now it's got this white man's face on it, but you can change it. So it has different different different appearances so right I just changed it it has a sort of a black woman's face. You can change all sorts of different presentations, which is good for a lot of different reasons. One is because it allows us to scientifically vary the appearance so it's very interesting studying how robots can push back on sexism, for example, right. Well that might work differently if the if the robot is presenting as a man versus a woman and we can change that. And of course the other reason is because you know I'm using this white man's face on the robot is because that's the Alex face which is the first alphabetical order so it's sort of the default. But of course there's no reason why that needs to be the default. And in fact if we're using robots in particular with particular populations. It might be advantageous for it to be more visually representative of the communities into which it's being embedded. That's amazing. So let's just kind of dive into an at the kind of an ethics conversation now what do you believe are some of the most important ethical questions that we need to consider like in your opinion when it comes to AI and robotics and how robotics play in the future. Yeah, so I teach robot ethics at Colorado School of Minds. So I could literally talk about this for a semester. But I think that if we have to limit ourselves down the key ethical concerns that I'm really concerned that that I think are really important, all have to do with power in different ways. When we talk about power. Well in robotics when people talk about power sometimes they're talking about power systems electronics right but if we're talking about power from a social perspective, then there's this theory from black feminist thought called the matrix of domination, which helps us to think about the different types of power and power systems within our society that that interact with each other. And I think it's really helpful for thinking about the different ethical concerns that we're facing today. So the first domain that Patricia Hill Collins talks about her work is interpersonal power. And this is what a lot of people in social robotics are typically very focused on the concern about can robots persuade you to do things can they influence you to do things right for good or for ill right in for good like robots persuading kids to eat their vegetables I don't know for right is is the robot going to persuade you to do things you shouldn't is the tech company that that sells the robot going to use it to persuade you to buy their products is the robot by not saying anything at all, going to persuade you that something that is happening for the robot is okay when it really isn't. And so there's a lot of these interlocking concerns here in terms of just social interactions about how robots can can persuade and influence us during interactions and then how that can shape how we interact with each other. The second domain that that she talks about is cultural power and that that relates to like what we're seeing here in terms of these decisions about how to present the robot right. How do race and gender manifest in a robot design. Some people tend to say well let's avoid this let's just not design robots that have race or gender but of course if you're building human robots that's not possible. And then even if you're designing a robot that is doesn't look human right at all, if it talks, as soon as it starts talking, people will ascribe race and gender characteristics to it, or just they'll look at where the robot is deployed, and if it's vaguely humanoid they might make assumptions. The third domain she talks about is disciplinary power, which a lot of that relates to surveillance. And I think this is a big challenge today because robots are mobile sensor suites right they are not all but many robots right in order to operate or constantly are constantly observing the people around them. You can't see it but right now on my on my screen, I've got the control of your face for for the fur hat up and it's tracking all of your faces and trying to identify them. Right and so there's this concern is there around privacy that come into play when we have these mobile sensor systems. And then the final domain is the structural domain, which focuses on the institutions, organizations, laws and so forth. And this is a domain that I've talked about a lot in some of my writing around the risks about robots and policing, where we're increasingly seeing robots being acquired and used by police departments, primarily for surveillance, primarily for primarily specifically for surveillance of working class and communities of color surveillance of protests and things like that, but then also occasionally for for violence. There have been a couple instances of police killing people with robots, or blowing up people's homes with robots or things like that. And so I think across all these different levels of power. It helps us to think about these decisions we're making about what robots look like, how they are designed, how they're deployed, and then the influences that has on us, both interpersonally and at a societal level. Yeah, that is a firestorm a hornet's nest if you will have topics and I can see why you want to stretch that out over a semester. Yeah, yeah. Yeah, there's a lot to process there. That's, um, yeah, I think the interesting thing about this is like we we typically talk about AI on this show. The unique thing about this is that it's when you marry it with robotics now you've got not just like it's not just like a, I'm going to put it in as simple terms as possible, like a voice in a box, it's no voice in a box that moves around that can interact with you. So that being said, like how I know that they're using them, like governmental organizations, things like that are using robots. How close are everyday people to having robots, part of their everyday lives? Well, I think that's an interesting question. I think the on one level, they're already here. Okay, on one level. People, many people already have like a Roomba in their house, right. And so at that level, you know, having robots in our everyday lives, if we want them, they're there. If we're talking about interactive robots and socially interactive robots, I think we're a little farther away from everybody having them in their in their everyday life. But it's not clear to me whether everybody needs a robot in their everyday life, right. This everybody need a robot in their in their home to socially interact with, not necessarily, right. And so we are already seeing robots deployed in the wild, consistently in certain domains, but they're very specific domains. So one of my favorite robots is the Moxie robot from Diligent, which is a hospital robot that's been deployed in a wide range of hospitals to help bring things across the hospitals for nurses. So nurses can spend more time with patients and don't have to spend their time running back and forth across the hospital. In other words, really, it's not replacing anybody. It's not replacing the nurses is it's helping to make their lives easier and helping them to focus on the parts of their job that they want to focus on. And so if you are somebody who is at one of those hospitals, then that type of interactive robot is there. It's in the world already. We don't really see interactive robots so much in other domains yet. But I think we're really getting close in certain other domains. We're starting to see a lot of advances in robots in education, therapy, other domains like that that have real potential benefit to people, but it's more localized, right. Not everybody is going to have a tutoring robot or a therapy robot or something like that. But in certain contexts, when it's really, really helpful, sort of as diagnosed, then I think we're getting closer and closer to having that be a real solution. Wow. I need to ask you then, is there one advance or one area of AI and possibly robotics that you're most excited about or where you see the most potential in the future? Yeah, I think that, so I'm, of course, I'm going to advocate for socially assistive robotics, right, robots that are able to assist people through social interactions, both because I think it's overlooked and undervalued, those types of benefits, and also because it tends not to be the type of robot that is replacing people's jobs or using or is being used to make people work harder and faster, right. The very first industrial robots when they're deployed in, you know, 40s and 50s were deployed in order to force factory workers to work harder and faster. And we still see robots in factories, you know, having that same effect today. And so while I think with those robots, there is a, there is a large potential for impact. They are being deployed out into the real world. Those are not the types of robots I'm most personally excited about because I don't think the effect of them are uniformly positive. I'm hesitant to choose any particular socially assistive robotics domain because I don't want to have, you know, have to choose which of my students pieces of work is, you know, is most exciting. But I think the example I'll give as the newest project that we're working in our lab is we have a student in our lab, Naomi Morris, who is doing work with designing robots for survivors of domestic abuse in order to help serve as a conversational part of the project. We have a partner with them to help them to practice talking about previously traumatic incidents. We've seen a big benefit of these types of social robots is that people will be more willing to interact with them when talking about certain topics, then they would be to talk to a human, because they feel that the robot isn't judgmental. It's not going to record what they're saying. It's going to listen and talk, but then the conversation is gone. So in domains like those types of therapeutic domains or talking about traumatic incidents or disclosing health conditions or things like that, I think there's a real benefit to robots that they can uniquely provide and in a way where there are a tool that therapists can use as opposed to replacing human labor. That is fascinating. And I'm so glad that you ended it on that positive note because that is not even close to anything I would consider. And yeah, taking the competition away from human robot interaction, I never considered how beneficial that would be. Any final words, Chris? I just hope they don't come for the comedy. I hope the robot, we don't have comedy robots. I hope there's never a Robocop that looks like YouTube, Chris. That would be a little bit scary for a lot of people for many reasons. Tom, you're fascinating. I can see that we're going to need a part two. Thank you particularly for your shout out to your student for your vast knowledge of history. I mean, I think ethics is a very important conversation that I don't think many people can really get their head around until they're looking at history and perspective. And we certainly like to do that here on Building AI Boston, but please come back and share more. And I think this is the fun stuff. I can see why you're everybody's favorite professor. So thanks for sharing it. Yeah, thank you, Tom. This was awesome. Thank you for joining us on Building AI Boston. Stay tuned for more enlightening episodes that put you at the forefront of the conversations shaping our future.
