Interview: David Gunkel on AGI Ethics and ‘the relational turn’ – Part Three

In an exclusive interview for AGI Ethics News, KB Miller sat down with philosopher David Gunkel to discuss the ethical implications of treating artificial general intelligence systems as “entities” and to explore a radical shift in how we approach the moral standing of artificial intelligence: the “relational turn.”

David J. Gunkel (PhD – DePaul University) is an award-winning author, educator, and researcher, specializing in the philosophy of technology with a focus on the moral and legal challenges of artificial intelligence and robots. He is a Professor of Media Studies and Department Chair at Northern Illinois University. He’s written numerous books, including Person, Thing, Robot and Robot Rights. 

Due to the length of the interview, we’re splitting it into four parts.

KB: Yes, that could be rewarding … However, good design reaches the instinctive rather than the conscious, and I wonder if the human beast — being — is self-aware enough to do what you just said … Shouldn’t good design somehow work at an instinctive level so that … so that people who are less self-aware, perhaps, or simply distracted, can still operate in a safe environment?

DG: Possibly. For myself, I’m really interested in how things get messed up once they’re out in the world. And get messed up in ways that sometimes are challenging, but also often open up new opportunities. 

I’ll just give you one example from something that has nothing to do with AI. At one point, you had these kids in the Bronx who wanted to make music and didn’t have musical instruments, so what did they do? They took turntables and records and invented hip-hop. When the turntable was invented, no one thought that it was going to be a musical instrument. No one thought vinyl records were going to become the raw material for remix. And an entire new kind of way of expressing yourself — making music, creating culture — was born of that sort of misuse of the designed object. That’s where I see really interesting things possible.

Now, in order to get something like that, there’s a lot of mistakes made on the way. And there’s a lot of things that happen that don’t necessarily deliver the same kind of outcomes. But what I’m really interested in is ways in which we can design technology that invites that kind of engagement by users, that doesn’t foreclose it — that doesn’t put so many guardrails in place that we stop things from that happening. And those guardrails are often not just technological, but legal, and institutional. 

If we want to really allow for people to invent and be creative with these new technologies, we’ve got to create a situation in which we foster this kind of shared discovery that needs to take place, not just from manufacturers, developers, and designers, but also users, experimenters, and kids, like the kids in the Bronx.

KB: It’s happening so fast that … we have a lot of moral behavior in the systems already that seems questionable, if not downright dangerous and harmful. How do we balance them? What would you, if you were on the building team, if you were on the design team for an AGI, how would you set out to balance [harm and innovation]? Do we need, for instance, some system of maturity the way we have with humans. I mean, children, adolescents, and adults are sort of three different moral categories … we have different laws around how you respond to the three of them [when they act]. Should we have some kind of graded ethical system for AGI, and what would that look like if we did?

DG: Right, yes, so … There’s been a lot of work on this, in advance of the technology, becoming even close to feasible. Already back in the early 2000s, Michael and Susan Anderson were already asking this question under something they called “machine ethics.” And the question was not how do we control machines, or how do we take responsibility for what is done with these tools that we’re creating, but how do we, in the operations of these decision-making systems, that they were working with, create ethical protocols. And ethical decision-making that would be part of the ethical community and be judged morally.

Susan Anderson [University of Connecticut], who’s now deceased, she was a philosopher, and Michael Anderson [University of Hartford], her husband, who’s still alive, is a computer scientist, and they actually tried to make ethics computable, so that they could program that into some sort of application. and unfortunately, their work has sort of been forgotten, or at least sidelined. You don’t hear a lot of people talking about it currently, but I think there’s something in what they discovered in that process that is important for us.

[We have] the opportunities and the challenges of approaching these questions, because these are not brand new questions — they’re questions that we’ve encountered historically in all kinds of different forms. But now, we are seeing, as you said, with this rapid advancement in the models that we’re creating and their capabilities, this kind of panic that, oh my god, we’ve got to do something really fast.

And I think that’s probably the wrong inclination to have. You want to be deliberative. And you want to make sure that you have a way of approaching this that is informed by the history of moral reasoning. And not just Western moral reasoning, but moral reasoning more broadly construed, because these things are global technologies, and they’re not going to just belong to one particular part of the world.

And that’s the challenge before us. And I have students who will say to me often, “oh my god, that sounds terrible.” My response to them is, yes, it’s a big ask, and it’s a big task.

But it does mean that the world needs us, that there’s room for what we do to have value, and to make a difference.

KB: How many people are killed by cars…? How many people are we going to sacrifice before that happens? Is that a project for centuries or a project for the near future.

DG: You’re asking me to predict here. It’s not something I do often.

Yes … every innovation, it doesn’t matter what you’re talking about in human history, it has costs, not only in terms of resources, in terms of attention, in terms of effort, but also in human lives. As you know, the automobile, right now, is still a piece of technology that takes countless numbers of human lives every year. And yet, we want to get places, and we somehow excuse the loss of life for the convenience of the transportation options that the car offers us.

So, none of these things are without risk, and I think the real challenge, in the face of any of these innovations, is not to come at it and say, well, how do we get rid of all the risk so that there is none? But rather, how is it that we manage the risk, and how is it that we decide what risk is acceptable? 

And again, this comes down to a question of power. Because someone is going to make these decisions. Is it going to be manufacturers? Is it going to be governments? Is it going to be you and I? Is it going to be experts in the field?

I think the biggest worry for me right now, as we’re confronting these epoch-defining questions, is the way that we are limiting who is in the conversation. In the United States, for example, when Congress is beginning to think about any kind of regulatory framework, or principles by which these technologies will be deployed and developed, who they bring to Washington to talk to the lawmakers tend to be the CEOs of major corporations. And you want to hear from more than those people.

And this is something that affects all of us. There’s got to be a better way to include a wider range of stakeholders in that conversation. Because in the past, when we’ve limited who is able to speak to these challenges, the results have oftentimes disadvantaged many parts of society in ways that are rather unequally distributed.

KB: One question that friends ask is why should we build AGI at all? Is there a role for AGI in our world? And what is that role? Or is that even too utilitarian an approach? Should we even go there — [though] it seems inevitable? My pragmatism [speaks]… 

DG: So, here’s the really interesting thing: You can’t know if you should go there or not until you try. In other words, if you could know that in advance, you would have this kind of epistemological certitude that is only possessed by the divine … if there is one. We don’t know. We live in finitude. We live in a temporally limited framework.

And whether or not something is possible, we can sort of guess, we can extrapolate and make predictions, but you’re not going to know one way or the other unless you actually put in the effort to try to do so. And the history of technological innovation is full of miserable failures, but also, incredible successes. And most of those were not able to be predicted until they actually were allowed to play out and transpire. 

So I think you’re right. I mean, It isn’t like you’re going to be able to say, don’t build this, because that involves a certain certitude. It says, we know this is something that you shouldn’t do, but that is a guess. It’s as much of a guess as saying that we should build it.

And you’re not going to be able to get the certitude one way or the other before you engage in the activity. So I think this is a very pragmatic kind of answer, but also I think it’s pretty true to how we experience previous efforts like this.

KB: What would you do if a robotic AI came into your home and demanded moral standing? What’s the process you think that we need to go through? What would I say … because as you know if you read Benso, it’s not about granting [standing], it’s about acknowledging it.

DG: Right. So, the way you asked the question, I think, is really crucial, because it really is grounded in the trajectory that we’ve created for ourselves out of science fiction. Already in Capek’s RUR, the robots rise up and demand recognition.

And this is a very standard way in which we from the Enlightenment, think about the granting or recognition of the rights of others — that they have to ask for it. They have to demand it, right? And you can see, with the struggles over human history, women got the vote because they demanded it. Enslaved peoples got freedom, because it was demanded that they be recognized as human beings. Etcetera.

So, we have this sort of prototype that comes not only out of human history, but also out of the robot stories that we’ve told ourselves. I think the recognition is going to evolve much more bottom-up, than top-down. And it’s going to be a much slower incursion, rather than a revolutionary event like we’re expecting from the science fiction narratives that we’ve created for ourselves.

So, let me just give you an example of how I think this is already happening. Right now, in 12 U.S. states, robots have rights. And that is something that, when I mention it, people are like, wait a minute, how come I didn’t hear about this? Because if robots have rights, there must have been an uprising, and they must have demanded their rights, and as a result, I missed it. Why did CNN not cover this?

And this is because of what rights are. Rights are not these kinds of things that we think are God-given privileges, bestowed on us by our Creator. That again is an Enlightenment-era narrative about how rights occur.

In actual practice, rights are about how we sort out social reality. In social reality, we have responsibilities and we have rights. And they’re coupled to each other. And so, if you have a responsibility to something, that other thing has a right. Or if something has a right, you also have a correlative responsibility to respect that right. So it’s a way of basically negotiating conflict and coming up with ways to get along with Others, and other kinds of Others that surround us.

So, in these 12 states, where robots have rights, the robots that have the rights are these wheeled devices that operate on the streets and deliver things: packages, coffee, sandwiches, things like that. And, we have them in Chicago, other cities have them now, but they’re becoming more available across not just the U.S. but elsewhere in the world.

And the company that builds these things realized that when the robot is in the crosswalk, there is a potential for an accident, because a car could blow through a red light or a stop sign, hit the robot, and we’d have to figure out liability and accountability for the accident.

So, they lobbied these legislators to make laws that now recognize the robot as having all the rights and responsibilities of a pedestrian when it’s operating on the street.

Now, does this mean that the robots are pedestrians like persons? No. Does it mean the robots are persons. No. It just means we’re taking an existing law and fitting it to a novel entity. And so, the rights of the robot will evolve in this way. As we fit various challenges to social reality — to the laws that we currently have available to us and try to make them accommodate this new reality that we’re building alongside these intelligent or semi-intelligent machines.

As a result of that, we will slowly but surely widen the circle of who has rights. And why they have rights in the social world that we inhabit. Now, does this mean that robots will have the right to free speech? That seems like a big ask right now. But if you would have said to someone three decades ago, “Do you think corporations will have a right to free speech?” they would have most likely said no. But one court decision later, and they do.

So, it’s this kind of case-by-case process that I think we’re looking at.

AGI Ethics News

AGI Ethics News covers the ethical and societal implications of artificial general intelligence (AGI). We bring expert insights and thoughtful analysis to help readers understand the challenges and responsibilities of emerging AI technologies.