Automatic for the people: How the rise of robots could change everything

This past June, we released “Degrees of Freedom,” an episode that explored how human dexterity is the holy grail of roboticists. In that discussion, we spoke to Jason Millar, a professor at the University of Ottawa and Canada Research Chair in the ethical engineering of robotics and artificial intelligence. Millar’s conversation with host Manjula Selvarajah and producer Ellen Payne Smith was so compelling and wide-ranging — it touched on everything from AI safety to the uncanny delight of robot fails — that we’re sharing more of it in this special bonus episode. Who’s your favourite robot? Tell us at media@marsdd.com.

Subscribe to Solve for X

Narration: Hello, I’m Sarah Liss, one of the editors here at Solve for X. This past summer, we released an episode that homed in on one of the crucial things that separates humans from robots: Our ability to pinch and grab and fumble around for a specific thing in the bottom of a knapsack. That is our dexterity. In the episode, we spoke to Jason Millar, a professor at the University of Ottawa, who also happens to be the Canada Research Chair in the ethical engineering of robotics and artificial intelligence. Professor Millar isn’t working on the problem of dexterity per se, but he does spend a lot of time thinking about the complicated philosophical questions that come along with our current attempts to build better, more capable robots. Our conversation about AI, labour and robot fails was so fascinating that we wanted to share the whole thing with you. So, without further ado, please enjoy this special bonus episode of Solve for X featuring robot ethicist, Jason Millar.

Jason Millar: So, if we’re talking about a ChatGPT moment for the Amazon warehouse robots, we’re probably already in the middle of it. But if we’re talking about humanoid robotics, probably a little far off.

It’s not that technology comes into the world and just adapts to people’s expectations and the way that people do their work. It’s not at all. What happens is the technology enters society and it pushes back on society. So it’s meant to solve certain problems but it also forces people to reorganize the way they work. And you can see this all over the place. I mean, if you think about, you know, the change to a work environment that shifts from one where there are typewriters everywhere, big, noisy machines, and then like, you know, the modern office, like with computers and printers and things like that.

And that’s like a kind of a mundane example, but there are some great histories of tomato harvesters. The story that was told about the tomato harvester was about being able to automate a job that took many, many labourers to accomplish and it was inefficient. And so when the new tomato harvester came out, it was going to pick the tomatoes fast and get them to market quickly and really free up and automate and industrialize an industry that could only benefit from it. Of course, when you dig a little bit deeper in the story, the way it’s told is that the tomato harvester, although it did free up some of that dull, I guess, I suppose dull and maybe expensive harvesting, did so at a cost, right? So prior to the tomato harvester, the story goes that tomatoes were just kind of juicier off the vine. You could harvest them later in their ripening cycle. I’m not a farmer, but — so, you know, when you went to the market to buy tomatoes, you tended to get these varieties of tomatoes that were sweeter, juicier, more tender, and just more delicious, I suppose. They’d make better pasta sauce, right, than what we get today. But the tomato harvesters would smush them.

Manjula Selvarajah: OK.

Jason Millar: So they had to start planting different kinds of tomatoes.

Manjula Selvarajah: And probably picked them up earlier.

Jason Millar: Pick them up earlier and get them to market when they’re kind of green, right? And that was largely to accommodate a shift in power from organized labour, and it served to concentrate power in the hands of certain people, farmers, and the people who produce these tomato harvesters. So there’s always a more complicated story that we need to get into when we’re talking about automating the mundane tasks that people just don’t kind of like to do. I think we should think carefully about that before we say yes to it.

Manjula Selvarajah: It’s interesting, the story that you’ve shared. I mean, you would think that the machine was going to adapt to the process or to the product, but what you actually, to the tomato, but you actually had is that the tomato had to adapt to the machine. That I find really strange because then when I think of robots in other spaces, does that mean that here I’m walking along thinking, “Oh well, the robots have to adapt to our lives,” but potentially what could happen is that we have to actually adapt in some ways to the coming robots.

Jason Millar: This is a critical point you’re making, right? The fact that there’s politics in the story is no surprise. The article that story comes from is a chapter out of Langdon Winner’s book, The Whale and the Reactor, and the title of the chapter is “Do Artifacts Have Politics?” So the whole point he’s trying to make is that there is a politics inherent to every technology and we just need to understand what it is and study it in order to get the full picture of how technology emerges and how it behaves in society. A whole, like a whole series of changes come in to adapt to the tomato harvester. Because that’s the thing that’s available to us, right?

Manjula Selvarajah: But in a way, it’s unintended consequences. Is that your fear here as a robot ethicist, that we actually don’t completely understand what possible negative implications may be? I might be putting words into your mouth here.

Jason Millar: No, I think you’re getting it exactly. So, unintended consequences are funny things. They’re only unintended to the extent that you don’t study or understand them, right? So what we’ve seen with AI, social media, is that what they would call unintended consequences are often actually consequences that a great number of people who study these things in detail have identified. Privacy is one of them, perfect example. I’m talking about, you know, 2006, 2007, when Facebook was just emerging. The people who studied — who spent their days studying privacy or the impact of these technologies on kids could have told you very clearly they’re going to have a negative impact on kids’ mental health. And only when they’re kind of dragged into the courts or dragged in front of parliamentary or House or Senate committees in the U.S., then the facts start to emerge: “Oh, well, we never knew that this was going to happen.” So, an unintended consequence, I always ask people to —

Manjula Selvarajah: You’re actually pushing back at me at that word.

Jason Millar: Push back a little bit, yeah. So how unintended or like unintended or willful ignorance or some version of the two.

Manjula Selvarajah: I want to ask you a bit of a strange question. Why does a robot in a factory need to look like a human?

Jason Millar: If you’ve got a face on a robot, it’s almost always because you’re trying to get that robot to interact with other humans. The face creates the opportunity for getting another person’s attention, for forming a type of relationship. Having a smile when you want to have them approach you and appear trustworthy as a robot. Having eyes so that people can tell where you’re looking and kind of anticipate where the robot might be moving.

Manjula Selvarajah: Isn’t that blurring the boundaries a little between the human and the machine?

Jason Millar: Well, it certainly is meant to smooth the relations, right, and the interactions. So when you talk to human-robot interaction experts, they’ll say that that type of anthropomorphic quality, giving a face, giving a voice, gendering the voice, gendering the face, the body, everything about it is meant to smooth interactions between robots and people. So —

Manjula Selvarajah: Make adoption easier in some sense.

Jason Millar: Yes, it’s meant to make the adoption easier, less scary, more intentional. So this is one of those things that can be used quite beneficially if you need a robot and a person working together side by side, this kind of co-robotics notion, very important to have that kind of — to have some feedback so that you can build trust as long as the is calibrated, right? You’ll hear a lot of people saying we need people to trust AI more. We need people to trust robotics more. And I always try to correct them and say, “You know, shouldn’t we be saying we need more trustworthy AI and more trustworthy robotics so that when we’re asking people to trust those things we’re doing so in a way that’s justified?” Like you would not want to ask somebody to trust a technology that is not inherently trustworthy.

Manjula Selvarajah: So when I think about our lives now and our lives going into the future, I feel like parts of us are already enabled by virtual robots and soon parts of our lives, if not already to some degree, will be enabled in some way or will be interacting more with physical robots. How then do you think we maintain our humanity?

Jason Millar: Oh, I think we need to make very clear decisions as individuals and as a society. So, through creating rules, like rulemaking, laws, regulations, we need to decide where we stand in the hierarchy here, right? What are we willing to give up to technologies like robots and AI? And what do we just want to maintain as being uniquely human, right? We need to have conversations about where these technologies fit in society and have those collective conversations. I think we are about to have a conversation of that sort with artificial intelligence that I don’t know that the companies were expecting. So, when I look at the latest public polls on people’s attitudes towards AI, it has so radically shifted in the last couple of years, from largely ambivalent to completely negative. I mean people really don’t like AI right now. I think that is largely because of the way that the technology has just been unleashed on society with it seems like very little consideration of what the impacts of that technology would be.

When Alan Turing wrote his paper in the ’50s about the future of computer — like where he invented the computer and thought about what would it mean to have intelligent machines, you know, the responses that he got were things like, well, you’ll never have a machine that can create art. You’ll never have a machine that can play the game of chess. You’ll never have a machine that can write a sonnet and things like this, right? Pointing to those things that were just uniquely human and that brought a lot of meaning, human meaning, into the world and into our lives, the things we love and cherish the most. And I think that with AI and robotics, we are more and more encroaching on those territories with these technologies. Certainly AI has created almost a crisis in education where we’re trying to figure out, what does it mean to learn? What does it mean to develop knowledge? Like, what does it mean to write a simple essay about your opinion on politics, right? If you’re getting a ChatGPT or a Claude to do that. And so this, you know, I think when it comes to labour, we’ve seen a bit of pushback. Artisanal everything is a pushback on technology. And I think we need to think carefully about what are we willing to value? Are the philosophers valueless? Because ChatGPT can write something that looks like an essay, and it’s a pretty good essay.

Ellen Payne Smith: I have a few more to just really throw at you quickly.

Jason Millar: Sure.

Ellen Payne Smith: One thing that struck me in the episode so far, the way that robot demonstrations and robot videos, for me, robot fails, those are the ones that really I love —

Jason Millar: They’re the best, aren’t they?

Ellen Payne Smith: Do you have a favourite robot fail?

Jason Millar: I mean, I just, there’s something about watching a robot trip and fall on its face that I think it just gives us hope that we’re still (laughs) — but I think that’s the canary in the coal mine. If you have that reaction, you should be asking these questions because it’s a signal that we are happy that these things are not encroaching on our lives and when they start, I think people are going to be more and more bothered by it.

Ellen Payne Smith: One thing I’ve been learning about is maybe that tripping is actually a really important step that’s actually it’s very human to fail and learn from it. And we could be reaching a point where actually those fails end up biting us in the ass because that’s a robot learning how to do what we do.

Jason Millar: You know, I have nothing — I work with roboticists all the time. I really like, I love, robots. Like this is the thing I got into, right? But let’s be clear moving forward that robots have their place. If to fail is uniquely human, then why do we care if robots fail? Like I think it’s just funny. I don’t think it’s funny when a human falls on their face. I think it is hilarious when a robot does.

Ellen Payne Smith: I’ve been sending Manjula maybe like every week another weird robot thing and she puts up with it, you know? One week it’s a humanoid running a half-marathon.

Jason Millar: Sure, I saw those videos (laughs).

Ellen Payne Smith: Kung-fu robots.

Jason Millar: Oh, I saw those videos too (laughs).

Ellen Payne Smith: Oh yeah, what else have I shown you? Anyway, the point is that there’s this amazing excitement that robot demonstrations gather, and yet through this episode, one thing I’ve noticed is how invisible automation has been and how it’s transformed our lives. Actually getting into the nitty gritty of how difficult it is to harvest a mushroom kind of blew my mind, because I didn’t realize just how difficult that was, and yet how much automation already exists in farming and how much technology and robots already exist there. And surgery and robotics have already fused.

Jason Millar: Yeah, amazing advances, the Da Vinci.

Ellen Payne Smith: And so I guess, could you speak to that sort of, I don’t know, the visibility of the humanoids getting attention and the invisibility of automation that actually runs our world and makes things, like amazing things, happen?

Jason Millar: I don’t know that I’ve thought deeply about that. I don’t know. I mean, the robots, they put them in your face, right? Like, there’s certain design where it’s like, we want things to happen in the background. We don’t want people to have to think about it. And it’s stuff that can just happen in the background. Displacing massive workforces is not something you can just quietly do in the background, right? It’s not something that happens.

Ellen Payne Smith: Maybe it just speaks to what we’re interested in. When something looks like a human, we have more interest in it. Frankly, the robots —

Jason Millar: Oh, for sure. Oh, definitely.

Ellen Payne Smith: The Da Vinci one doesn’t look like a human. It looks like — you could never imagine a surgeon would be able to operate that. I mean, it’s incredible the way it looks.

Jason Millar: I mean, if you’re talking about, sure. So, again, I think this has something to do with faces and hands and legs, right? The Da Vinci is an incredible machine that has revolutionized surgery. I don’t watch YouTube videos about that every day. But like, in terms of impact and benefit, maybe, I mean that’s the robot that deserves the accolades, right. And my time. I mean everybody should know about the Da Vinci. But yeah, of course we watch the robots fall on their face. Or even just like there was a robot Neo about six or eight months ago. Do you remember this? There was this like insane hype cycle over this robot. I looked it up actually right before logging in for this recording because Neo is touted as a humanoid robot that will be living in your house. And they’re sort of taking pre-orders for it. They’re saying things like, “Oh, right now Neo is completely remote-controlled by a human in the background and it can’t do anything on its own.” So it’s like, OK, well, this is a horror story now. Like you’re talking about a human robot in the house with cameras and recording videos and somebody’s sitting in some studio somewhere like manipulating this thing in your home?

Ellen Payne Smith: Yeah, it’s like right now what’s happening in a sort of surgery with a piloted robot with an advanced surgeon is actually like sort of a video game for some tele-operated robot cleaning your dishes.

Jason Millar: Right and it was like it was weirdly incapable of doing anything as well. It couldn’t quite manipulate anything. It took it like 15 minutes to put a plate in the dishwasher. I mean, what are we talking about here?

Ellen Payne Smith: Manjula, you should come — I see you laughing but like it looked like — someone described it to me as looking drunk.

Jason Millar: You need to watch these videos because this was a huge story in October. I mean, it came across all of my feeds for about two weeks. My students were asking me about it. Friends at dinner parties were asking, “Oh, did you see that robot that’s coming out?” And I was just shaking my head, like, I can’t. So this is the hype cycle. It is the epitome of the hype cycle. But somehow people’s desire to see this kind of thing is just like, it’’ insatiable, it seems.

Ellen Payne Smith: Totally. Two last rapid-fire — favourite robot?

Jason Millar: Oh, favourite robot. Oh, there was —

Ellen Payne Smith: Mine’s the Roomba.

Jason Millar: You know, can you give me two seconds? (Laughs)

Manjula Selvarajah: This is important.

Jason Millar: Hold on.

Manjula Selvarajah: Is it a movie robot?

Ellen Payne Smith: What’s your favourite robot, Manjulah? I’m curious.

Jason Millar: Okay, so it’s a robot called Keepon. So Keepon was the most basic robot. It was just like a little dancing robot. And it danced, they put out a video of this robot dancing to a Spoon song. And I could watch that all day and all night. I used to play it at the beginning of classes and just get my students to just kind of dance. You have to watch this video. It’s Keepon dancing to Spoon’s “I Turn My Camera On.” That’s the video title.

Ellen Payne Smith: Fantastic. Manjula, do you have a favourite? I just want to add it to the —

Manjula Selvarajah: I don’t think I have a favourite. I would have to say that I have a most annoying robot and a robot I like and they’re both in the same movie. So it’s R2-D2, who is, I think I’ve mentioned this to you before, Ellie, R2-D2 is like this kind of sensible, dependent robot. And then you have that other, super annoying robot —

Jason Millar: C-3PO.

Manjula Selvarajah: Yeah, so annoying.

Jason Millar: His sidekick.

Manjula Selvarajah: It’s so funny that the sidekick is so human-like because it’s annoyingly — it’s annoying as a human.

Ellen Payne Smith: It’s a know-it-all.

Manjula Selvarajah: So it’s funny that I would say that, like, the most annoying robot I can think of and kind of the robot that I would want are in like the same scenes. But I can’t think of a real robot that I have come across that I thought, oh, you know.

Ellen Payne Smith: One thing that is so crazy. It just came to me, though, because you’re an ethicist, Jason. Bear with me. In this episode, specifically, we’re talking about force and torque sensors. I don’t even want to begin to explain to you how long it took me to understand how that works, but I figured it out. And then there are people developing artificial robot skin and soft robotics, which are just amazing. And that made me think about these equivalents to our bodies and how we work and because we see robots in the world, should we be giving them a sense of pain as well?

Jason Millar: You know, look, I mean, we have to be careful when we talk about a sense of pain because these robots don’t feel anything. And this is for a philosopher, this is very important. And people will talk about conscious robots. Oh, if robots had consciousness. Oh, we’re just around the corner from conscious AI, these kind of things. I mean again, this the stuff of horror movies to me because what seems like a fascinating engineering problem is dangerous. I mean like, why would we want to create something that could experience pain? I’ll call it experience pain. If you wanna call it sensing pain in order to improve the digital feedback loop that prevents it from doing certain things and that makes it more effective in its function, OK, fine, but if you’re talking about giving it a sense and experience of pain, like the way that you experience pain as a negative sensory experience, right? Something that you would not want in your life, well, that’s something different, right? I mean, why would we want to do that? Why would we want to create a being that could experience that and own it and have — and then we shouldn’t be creating those things or even talking about creating those things, the moral status of a thing changes if it can experience pain. I mean that’s one of the criteria that we use as ethicists to determine what types of beings and life forms are deserving of moral consideration. So plants, as far as we know, don’t feel pain. Chop a tree down, right? So I just want to be careful because I think some people conflate the two. It’s like, there’s the engineering concept of like feedback loops and training algorithms and reward mechanisms and things like that. We do that with computers right now, other and other technologies.

Ellen Payne Smith: And there’s value in teaching something like, oh, don’t exceed this value of temperature, which —

Jason Millar: Exactly. But it’s not feeling like it got burned. We don’t want to do that. Why would we want to do that — it sounds horrible.

Ellen Payne Smith: Well, yeah, that’s true (laughs).

Jason Millar: (Laughs) Right? Am I right? Can we agree on this?

Manjula Selvarajah: Yes, we can agree. He’s going to walk away thinking we are evil geniuses. Absolutely.

Ellen Payne Smith: Should there be like things built into the design of things that show that they’re not human? Like there’s a moment where you see a humanoid put itself into a box and then you realize it makes this kind of movement where you’re like, this is really not human. And what I see it doing, it has this sort of like affect. It’s like this isn’t human. But as we reach to make things more and more human, I wonder is there an obligation on having some like a big stamp on its head saying not human or actually maintaining that less human movement into something so that people never worry about that.

Jason Millar: I think that as a general rule, we should be building things and minimizing the anthropomorphic qualities unless absolutely necessary for that exact reason. So I think it’s the latter that you just pointed to that’s important to keep in mind. I think you should have to justify the inclusion of anthropomorphic qualities because it triggers these emotional responses in humans. So there’s a famous video of the early Boston Dynamics robots. One of them was called Big Dog. It was like a big dog. They call it Big Dog because it looked like a big dog. Now they’ve got their little Spots that run around and they’re pretty amazing. But there’s a famous video of one of the roboticists testing Big Dog and Big Dog was kind of hopping around on its legs and the roboticist like raised his foot and just kicked it hard to try to knock it right over. And look, this is a robot. It doesn’t feel pain, but the reaction that people had to that video was horror. So people watched that and they were horrified. I mean, so the point is that we need to be very careful about the anthropomorphic qualities we build into these robots in order to not trigger those types of reactions. Why would we want somebody to have that reaction to a warehouse robot or something like that? You know, if there’s a reason to do that, let’s do it. I think we should have a rule that, like, we don’t just put robot faces on robots. You know? You can make them look more boxy and robotic and less human-like if it’ll get the job done. And so I think we need to be careful about anthropomorphizing for that reason.

Ellen Payne Smith: I’ll let you go. Thank you so much, Jason. And for putting up with all my extra questions at the end, I really appreciate it.

Jason Millar: Yeah, no problem.

Manjula Selvarajah: Jason, you’re fantastic. Thank you so much. Oh my gosh, what a fun chat. Thank you.

Jason Millar: Yeah, what do you want me to do with this audio recording? Like, was the sound OK? I can just keep it on my phone if you need it or I can just send it to you.

Narration: Solve for X is brought to you by MaRS, North America’s largest urban innovation hub. Gab Harpelle is our audio editor and mix engineer. Jason McBride is the senior editor. Mack Swain composed our theme song and all of the music in this episode. Kathryn Hayward is the executive producer. Our associate producers are Lara Torvi, Sana Maqbool and me. I’m Sarah Liss, filling in this week for our host Manjula Selvarajah. Thanks for listening.

llustration by Kelvin Li; Image source: iStock