242 – Reflections on Humanoid Robots

[00:00:06] Michele: Hello, everybody. Welcome to the IDEMS podcast. I am Michele Pancera, an Impact Activation fellow, and I’m here with David Stern, one of the founding directors of IDEMS.

Hello, David.

[00:00:21] David: Hi, Michele. I’m looking forward to another discussion. What are you wanting to discuss today?

[00:00:25] Michele: Yeah, it’s always fun to chat. And today we have quite a peculiar topic. We are talking about humanoid robots, which only in the last few years has become an actual topic. It was not in the radar just a few years ago, but now there are so many companies really pushing, both on the physical side of it, the production of the actual robot, and on the intelligence.

So there’s a lot to think about or to talk about here. Maybe we can start by thinking about how realistic that is. Do you have an opinion on when and if we are going to have capable humanoid robots, and by capable I mean capable of working some simple jobs that would substitute humans.

[00:01:29] David: I don’t know. I mean, this is the thing, the way you’ve defined that, some simple jobs that would replace humans, why do you want a humanoid robot to do that? It’s plenty easy to get robots that can replace humans for specific jobs. So why do you need it to be a humanoid robot to do that? And I’m very interested in the idea of humanoid robots, and, as you say, there’ve been incredible advances recently on both fronts, not just on the idea of humanoid robots, but actually artificial intelligence friends.

You know, there’s articles about not the physical side of it, but the interaction side. And there’s some really interesting elements of that where these exist and they play a role and they’re gonna be part of our societies in one way or another in the future. So I think the “when” is irrelevant to me, the “if” I don’t believe it’s an “if”, I believe it is “when”. But the question of how these are going to integrate into society and what roles they’ll play, and how this would be used, I think really depends on the society. And that’s what I think is most interesting.

The most interesting part to me is not about the humanoid robots themselves, but it’s about the relationship of the humanoid robots to the different societies. And I believe that’s going to be very different in different societies.

Humanoid robots in Japan are going to be very different to humanoid robots in the US, or in the UK, or in Italy, or in Kenya. They are going to interact in society, with society, in very different ways in different contexts. And that’s what I’m most interested in.

[00:03:15] Michele: Yes, this is definitely an interesting question. But let me first try my best to answer the first question, which was why, correct me if I’m wrong, why do we need a humanoid, like a human-shape of a robot?

[00:03:33] David: No, I was not posing that as a question, but I’m happy with that question. What I was saying is that I am not against us having humanoid robots as playing certain roles, I am against defining them as humanoid robots when playing the role of taking away work, or certain types of work, because I think the work aspect is not what is interesting.

I feel it’s the social aspect, which is the interesting part of the humanoid because it interacts with society in a different way. That’s what I was trying to say. I’m very happy for you to still say why you think the humanoid shape is important, but I want to clarify my statement or my question. It was not a question of why an humanoid shape. It was the fact that the definition of a humanoid shape as being related to work is much less interesting to me as the definition of humanoid as actually interacting in society and the way that that plays a role.

[00:04:33] Michele: I see. So the reason I was mainly thinking about the relation with work, even though I agree that the most interesting part is exactly what you’re mentioning, is because that is at the moment, and I guess it will be for a long time, the business case. If companies that are building humanoid robots can replace workers, then the cost of the humanoid robot can be pretty high. And so I guess that’s going to probably be the first use case for humanoid robots for a while.

[00:05:13] David: But, as I say, I believe that that may be the example, that may be an argument for it. But I’m not sure how valid that is, I can’t think of a context where the humanoid aspect of the robot is the defining feature that means it can replace work.

I mean, one of the interesting cases for this is care of the elderly. Care robots in different ways are a very interesting place, and Italy and Japan are very interesting cases with aging populations, which have a lot of longevity in all sorts of ways, and so the need for potentially humanoid or some form of robotic carers is an interesting one.

And the question I have for you is, is humanoid actually important in that? And if it is important, is it important because of the work, or is it important because of its interaction with the social fabric? That’s the question that I don’t know and that’s an interesting question.

[00:06:23] Michele: Well, for what concerns the interaction, being a humanoid robot, as in the interaction with humans, I’m sure that plays a role but I am very hesitant on giving opinions. What I have an opinion on is on a more practical level. You know, if a robot has to perform a very specific task, there are plenty of ways, depending on the application, to optimise the shape, to just do the task in the most efficient way. Now, that is, I guess, also the case for all of the devices that our mobile phones have substituted now.

For example, we don’t have the need for calculators anymore because they are embedded in our phone. We don’t have the need for maps because you have maps in your phone, and so on and so forth. And you could make the case that, as a single object, the mobile phone is not as good or efficient as some of the others, but the power of the mobile phone, or the smartphone I should say, is that it is versatile and you can just download an app and you can use it for so many things.

And the reason why I’m talking about this is because if you have a humanoid robot that lives, well, lives may not be the correct word, but that exists in our houses, in our factories, in our roads, in our environment, well, we shape the environment in such a way that it’s usable by humans for the most part. So what if you have your worker, your humanoid robot that stays with you in your house, what if that robot takes care of you, but then while you sleep, it cleans the windows or mows the lawn?

Well, if it has to be versatile, you want to have a shape that can adapt to any task you could want it to do, and the environment that we have is adapted to the humanoid shape.

[00:08:47] David: I hear what you’re saying, but it’s equally adapted to dogs, so why wouldn’t a natural shape be more canine than human? Because actually there are other advantages, the human shape is not the only shape our environments are adapted to.

[00:09:06] Michele: Yes, so two things here. Yes, dogs can live in our environment, but they have a hard time turning on the lights or using a sink or so on. So, I don’t disagree, but I think that the human shape is the best for the environments that we build for ourselves. But then there’s another interesting point, which is training. How do you train a humanoid robot?

And the main way, at least at the moment, is that an actual human with sensors all over his or her body does the tasks, and those motions are then registered by the humanoid robot, or by the intelligence of the humanoid robot. It is a big advantage for training. And then you can also, I don’t think this is doable now, but one of the visions is that once the intelligence becomes good enough, then videos should be sufficient for training. And I don’t think you can do the training at the same level with other shapes.

[00:10:19] David: I don’t know that I agree with that. I think that there is a nice pattern to recognising that our shape is the shape which works for the tasks that we want to do. I’m not saying that there’s no sense in that, but I don’t think that, in terms of that training, that’s necessarily the case. And you say that a dog can’t turn on the lights, well, that’s not true actually. There are particular cases where you could train dogs to do that, and if you can train with a video, then actually, that doglike shape that is able to do, there’s real advantages to, especially in robotics, to not just living in imitation, to recognising that you have advantages and disadvantages that come in what you’re creating.

Now, don’t get me wrong, this is not something I’ve thought deeply about, so I’m not saying that I have any answers here, and I do understand the attraction of humanoid robots, and one of the real questions is the emotive side, you know, emotional attachment. Forget about the actual robotic part and just get to this sort of sensor, which is your AI companions, which are out there.

There’s a whole set of debates that are worth having just about that. And there is that fact of what interactions could look like with such a humanoid robot, if those features were able to be emotional as well as just work related. Being able to assign tasks is one thing, and then you get into the questions of, well, okay, so now we all have robotic slaves. Is that really the future that we envisage for society? Is that what we want? Is this actually the best use of the limited resources we have to use them in this way?

These are really interesting questions, I don’t know the answers. You know, there’s plenty of films which address this in fun ways and engage in some of the potential implications of such things. These are very interesting questions for our society at this point in time. But the more interesting questions I want to come back to are: does the value or the potential of a humanoid robot depend on your society? And will they be received and integrated into societies differently in different societies?

That’s the question that I think is so much more interesting, because it comes down to culture, it comes down to expectations, it comes down to roles and responsibilities, it comes down to demographics. You know, in an aging population where you have not enough young people coming through, the incentives are very different from in a population where you have a population pyramid with lots of young people. The incentives for these are very different, and so I’m really interested in actually questioning how such ideas engage and interact with our societies themselves.

[00:13:38] Michele: Yeah, I really like the fact that you’re bringing this up. I didn’t tell you yet about my worries, my main worries, connected to humanoid robots – and I have quite a few – but I think for the case of Italy specifically, of course I occasionally think about Italy, as you can guess by my pronunciation, it is worrying that the population of Italy, or the age curve is shaped as it is.

So many workers are exiting the workforce and not being replaced by enough young workers. And I mean, it is conceivable that we could increase the natality rate now, but that wouldn’t solve the problem really, because we already have that all of the ages between zero and let’s say 30 to 40 years old, there are just not as many as they should be. So how do we solve that issue? Because it’s going to get worse and worse while time passes. 

[00:14:49] David: Sorry, I think I’d like to come in because I know a bit about the Italian context, but not all of our listeners might. There are many villages, there are regions in Italy where villages are broadly abandoned because of these problems where it’s not just people moving to urban centers, it is because of the low population rates, there’s sort of aging villages where the population is not being replaced. And so you are losing villages, this is really real in certain Italian contexts. It’s not the same as in, even in the UK, we don’t feel this as acutely as it’s felt in Italy. It’s one of the reasons I mentioned this before.

[00:15:26] Michele: That’s exactly right. Italy has had the worst, or maybe not worst, lowest natality rate, probably just as much as Japan. I think they’re more or less even for the last three decades or so. It is a very apparent problem in Italy. And, obviously, another strictly connected issue is people who retire, they have to have their retirement, the pension, and who’s paying for it?

[00:15:56] David: And they’re living longer. And this is again, in Italy as in Japan, where some of the oldest communities in the world exist.

[00:16:05] Michele: Yes, exactly. And unfortunately that’s a problem.

[00:16:09] David: And this is a wonderful thing, isn’t it? It is both a problem from a societal perspective, but it’s also an amazing sign of progress, it’s a wonderful problem to have.

[00:16:18] Michele: Yes, exactly. I completely agree. And I’ve been thinking about what the solution could be, again, not increasing the natality rate that would solve a future problem. And I guess one theoretically possible solution would be a sensible high scale immigration. I don’t see that happening, at least with current politics. What’s the alternative? And in a futuristic imaginary world, maybe humanoid robots could play a role here.

Again, I have many worries connected to humanoid robots, but for what concerns Italy, I am slightly more hopeful. 

[00:17:05] David: But I think this is a really interesting question, because what you are coming back to is exactly what I’m interested in, which is tying it back to a particular society in a particular context. And I find it very interesting if you take this idea of a society or a context with very low birth rates where, as a population, you are becoming sparse within the land, the resources, and so on. That’s a very interesting context to be conceiving the role of something like humanoid robots.

And I think there are many other broader discussions, which I think would be useful and sensible to think about around this. But I do like the idea that if you are going to consider something like this, look at it in that context. Consider what it means in this context, and we come back to the example, which I’ve used quite a lot, humanoid robots for care, for human care in different ways as being an obvious application area, lots of people are thinking about this.

And it is one of the questions in previous episodes we’ve discussed, what are the jobs of the future? Well, the jobs of the future are probably in care. This is one area where we hope, you know, related to healthcare, related to care in general, this is something where I hope there are at least as many jobs in the future as there are now. Do I want humanoid robots to be there to replace those carers? No. Do I want them to be there to enhance those carers? Yes. And there’s really interesting questions around this.

Now, if you take that question of the role they’re playing, and if we think about the role now of enhancing a carer rather than replacing a carer, is humanoid important? And I come back to my original question. I think it’s about the social fabric. So, if we think about the role that we would like for the robotic element to enhance carers abilities in an aging population, I don’t know that I want the humanoid to be the characterising feature, because I don’t want the robots to be trying to replace the humans and replicate the humans.

I want them to be complimenting the humans. I want them to be able to do tasks that the humans struggle to do. Lifting somebody in and out of the bath, that’s a horrible task for humans to do as a carer. It’s really backbreaking work, which is needed, it’s necessary. But the humanoid shape is not the best shape for that task.

And so actually, I come back to the fact of asking the question, if we are looking at this for work, but even if we are looking for this as a place in society, are we right to be thinking about humanoid as being important? And I don’t know, I don’t have an answer to that. I see there are pluses and minuses, and I think what you’ve articulated very well is the value and the importance of multipurpose.

There’s wonderful equipment which enables people to lift people in and out of baths, which looks nothing like a human, which looks more like a swing, it sort of lifts you up, you sit in it, you get moved into the bath, and these work very well, but they’re single use. I love the parallel with the mobile phone. What does that look like when it becomes multi-use?

So, I think distinguishing and actually thinking carefully about what is it that we’re wanting to solve with these and not getting caught up in the science fiction element of “isn’t it amazing to have humanoid type robots and that we have the technology to build these now?” Wonderful, and science fiction, this is great, this is stuff films are made of.

But going beyond that to actually saying “well, okay, in a context, let’s say the context of a community with an aging population, what is the balance we want for jobs for humans, the ability to actually offer good care services for the elderly, and the potential for technological robotic solutions?” And we look at that in a whole, and I come back to your context, and I say in the context of Italy in an aging population, I’m really interested in robotic solutions, but I don’t know that humanoid is the characteristic that I care about.

Multipurpose is much more important. And complimenting the human component is what I am more interested in because I don’t want to lose the human carers, I want the robotic carers to actually compliment and enhance the human carers. That would be one of the ways of framing this, and that’s why I love to think about these things in context.

[00:22:02] Michele: Yes, I really like the way you’re framing it. Maybe before we finish the episode or we conclude the episode, let me just mention that while you were talking I realised that probably my biggest concern is that these humanoids may become way too cost efficient compared to a human. I think that’s my concern, because then there would be a forcing function to substitute humans exactly where you and I don’t want, ideally, humans to be substituted. But a human worker is expensive.

[00:22:42] David: Absolutely. Now you take another context where, actually, the population pyramid is totally the other way round and you struggle to get teachers, and you replace your teacher with a humanoid robot. I get very nervous. That’s exactly a context where, again, I think there’s possibilities for this, but replacing human teachers with humanoid robots, that’s the sort of thing which is sort of a nightmare to me. These are not dreams, these are nightmares.

Having technology-enhanced teaching, which enables teaching to happen more effectively, but keeping humans in the loop, of course. But I feel the same is true, and I think we should be thinking similarly, even in the healthcare and the care facilities. So thinking about this as being something where your fears stem exactly from the fact that people are focusing on the wrong thing. They’re exactly focusing on, well, okay, we could replace, we could be more cost efficient, but that’s not the only optimising function that we should be driving.

That is a societal question that we need to ask. Are we needing and wanting to optimise on cost efficiency only? Or are there other forces and we can actually deal with higher complexity than simple financial optimisation? And that’s a really interesting question, and when it comes particularly to thinking about humans, the reason to think of humanoid is because you’re thinking about humans. I don’t want you simply optimising on cost efficiency, because that feels wrong.

This is exactly the parallel to slavery. And the interesting parallel to slavery is that slavery was terrible for the slaves, but it also eroded certain social fabric elements. And so it was actually, in many ways, it was very detrimental to many societies. Yes, many individuals got wealthy on the back of it, but societal structures, there’s a lot of issues around that. So trying to think of a, well, a new form of slavery in terms of humanoid robots, there’s wonderful movies that look into this.

[00:25:02] Michele: Yes, science fiction has done a really good job. There’s so much of it.

Well, thank you David for this conversation again. I really hope we are going to talk again about humanoid robots in the future. I really hope they will never become very cost efficient and that we, as a society, stop thinking so much about being as cost efficient as possible.

[00:25:29] David: This is something that is outside the normal realms of the podcast that we do, and the reason is that it relates to hardware rather than just software. Actually, there’s enough challenges and problems with the software piece of this. You know, the hardware piece, there’s so many other issues about how we use our resources, what we’re using resources for and so on, that, yes, this has been a fun excursion, but it’s not something I feel equipped to really engage in, because hardware is hard.

There’s a whole set of other complications that come in, and maybe one day we’ll actually start engaging with the hardware side of things. But at the moment, our focus is really on the software and the data side, because that’s something where it doesn’t require the same level, well, it just has a different set of constraints on it.

[00:26:22] Michele: Yes, it’s not that software is easy, but hardware is hard.

[00:26:28] David: Well, that’s not exactly true, but it’s a good place to finish.

[00:26:33] Michele: All right. Thank you, David.