Plinth

I was curious about what kind of games you grew up with?

I would say the games that I really remember are the Will Wright games. I spent so much time in SimCity. I just remember the pleasure of terraforming the landscape, and then eventually getting into the details like what zoning goes where, industrial or commercial, why you have to have this road here and this fire station there. What child thinks about that stuff? No other game has provoked me to think about its construction, or provoked me again as an adult to consider how he pulled off this magic trick of making a child care about abstract city planning. So for me, that was a hugely influential game.

I think of the Emissaries series of simulations as your first major public moment. I know you studied cognitive science and worked at Industrial Light and Magic a while before that. What were you making before you made Emissaries?

Between 2012 and 2015 — 2015 was when I first made the first Emissaries — I had been making a bunch of simulations. I made eight to ten different simulations, and I showed them various places. I was very proud of them. The first one is called Entropy Wrangler, and is simply a collection of stuff I liked, and I just dumped them in this void space. They had some physics attached to them, so some got sticky with each other, while others—in a very stupid and autonomous way—would attack and break the other ones. It was kind of like a virtual version of sticking a scorpion and a tarantula and a rattlesnake in one basket.

From there, I started to learn about the possibilities of making work within these parameters. You know, Tesla or Boeing, they do simulations all the time to simulate better wind flow on the car, and they publish the best one; but if you're not concerned with that, you just want to see the process of the simulation iterations, well then, you have a potentially infinite work.

This for me is very alive and interesting. But I also felt like I didn't know where I was going with this. There's no north star for why I was doing what I was doing. And I realised that something narrative would really help here to orient all this decision making. Simulation seems so open, right? But what if you paired it with a story and they fought each other, like they were different forces. So you have this tension in the artwork. And for me, this was an important thing for Emissaries. It also unlocked this very basic idea—which maybe I wish was taught more in art school—that you need some tension in an artwork for it to be interesting. You can't just make a romantic comedy where everything's purely happy. You need a little darkness in there. And you can't make a super dark post-apocalyptic thing that's completely humourless. It sounds like an obvious thing about artwork but no one taught me that and so I thought: oh, I finally have some tension between two sculpting forces. Now I have something interesting to work with.

With this new work, Life After Bob, it seems like your work bleeds into the world more than has in the past. It’s not a simulation, for starters, and it’s much more allegorical and speculative about narratives taking shape at the moment, around AI. I wonder where you see yourself on this boundary between your role as a designer of hermetic worlds, a sort of SimCity player with a god’s-eye-view of the system, versus more of a storyteller and social commentator?

Life after Bob came after making Bob, this AI creature, a super Tamagotchi thing that started at The Serpentine. And then I started to work more on an AI model where he's able to make inferences from very, very few examples. Typically, you hear about deep learning that needs thousands and thousands of examples for it to make an inference, like recognising an unlabeled picture. With my neuro-symbolic approach I was able to allow Bob to make distinctions between things with one or two examples. By creating these abstract rules, I could stupidly apply them to new situations and see if they fit: like when you go to a new city. For instance, when I went to Seoul, I didn't really know the rules of what Korean culture was like, but I applied all my inference from living in a city and I was half-wrong, but I was also not totally disoriented. So I tried to apply a similar approach to Bob.

My partner and I were having a kid, Eden, my first daughter. I was expecting her, so I thought, I’m going to do something easy. I'm going make a narrative, like a cartoon, for kids. So easy! No, it’s way harder than I thought it would be. I thought the production process would be a known quantity: animation, sound design, you need voice actors, and you need a story. ‘You need a story’ is like an ICBM missile to figure out. You need a story? That's the number one problem everyone has. Working on that was a very humbling experience.

But I would say, in retrospect, making Life After Bob was a way to speculate. What if the AI Bob could be something that's not just an artwork that lives in a Serpentine, but something that starts to learn more about you, and maybe it’s integrated into children's media – not to sound creepy, but I just think this is inevitable, so it's something I'm interested in exploring. Imagine if your parents introduced you to any movie or children's book, and the character could just talk back to you. And then over time, that character remembers things about you. Or remember certain patterns of behaviour, the little phrases you'd liked or didn't like. And then imagine you get sick of that character as a child. You know, my daughter was into animals, and now she's into princesses, and that phase went pretty fast. Imagine now that what the animal has learned about the child is now transferred to the princess. I'd started to wonder about this while I was writing Life After Bob and I thought, maybe Life After Bob can be a way to speculate about things I just can't make yet. And maybe in some other parallel track, in my life, I can try to work on some actualisation of that.

This was a way to speculate about an AI agent that understood aspects of your own desire, your life path, your life script, your hang-ups, your parentage, some matrix of all those things that shape who a person is. And it could just be with a person and guide them. In the movie, we make it hyperbolic for the narrative. This is a way of automating introspection on a mass level that I couldn't do without speculating about it first. So Life After Bob was my initial way to think about it for myself, and hopefully, for viewers. It’s something I'm interested in: how do you automate introspection?

I love that it’s also a film about anxiety. I mean, Bob the guide embodies a kind of anxiety about choice and life path. You kind of have a reverse Pygmalion with Life After Bob, where Bob the AI learns everything about Chalice and becomes a kind of father, sibling, or superego figure. We fall into this sort of Freudian dynamic pretty quickly, right? So it kind of makes sense that you’re having all these feelings while having a kid. With Life After Bob—like with Miyazaki—it’s about an adolescent, so it’s a story of hope set against the opening up of an expansive and volatile world, with this quite traditional architecture of self at the centre. Part of me wonders, do you think that this is always the endpoint for AI, to mirror and exacerbate our psychoanalytic self-narratives? Or do you think that there is a space for transformation of the self?

Interesting question, I mean, it's certainly something I wrestled with in making this first episode. How far out do you go? It's like okay: “Ian, is this a story about where human evolution might radically go in the presence of AI, in a way that we can't imagine yet—here's an opportunity to imagine it—or is it a coming of age story set in a world that includes AI that provokes the coming of age in a new way?” And I landed on the latter, but I was haunted by the former question.

I thought: I have to start somewhere that I personally had time and energy to care about as a new father. I’m really concerned about how much influence I have as a parent, how that shapes my child's life, how that shapes her sense of trust of the world, her sense of options, her sense of risk, her sense of what's true, her sense of how to clear the fog, when things are very confusing. And for me, as a parent, how to let go as a dad versus when to be protective, give her the right tools. And when the time comes, how to not be a parent and just be a friend. All these concerns were like overwhelming. They're very pedestrian concerns, and within the scope of the possibilities of human life. They were very ancient ones, it’s almost like a repetition or rite of passage just to go through them.

I'm not done with this question of where AI and humanity go. I realised about halfway into production, while we were animating everything: it would be beautiful to make an eight-episode series where the end of it you actually feel you have new archetypes of human beings, not just a repeat of an archetype in sci-fi dressing.

For instance, let’s just say for Chalice, maybe she is an archetype of someone with unearned life experience, because Bob lived the life for her. That would be quite new, and in a world of AIs like Bob, you could start to have people with unearned life experience living as if that was their experience. Chalice maybe dirt poor, we don't know, but she accrued all this experience that she didn't have to live. What would that do to you?