Parag Kumar Mital: I think this has always been a problem with the media. I think it's especially exacerbated and difficult with artificial intelligence because it's much easier to craft narrative. It's much easier to create media. We need to effectively have more criticality as a society.
[Music]
Ana-Maria Carabelea: Welcome to Art Is Not a Thing, the podcast series about art as a practice of critical inquiry, knowledge production and world-building. My name is Ana Carabelea, part-time host and producer of the podcast. For the new season of Art is Not a Thing, we’ve partnered with Radio O1–one of Austria’s leading cultural voices to bring you the next six episodes. Over the next months, my colleague and radio journalist Hannah Balber will discuss with artists and researchers whose work questions and reimagines our practices in and of the world. In today’s episode, Hannah discusses rumors, power, and the way narratives are shaped, manipulated, and distorted: from the political intrigues of the Mughal court to today’s AI-driven media environment. Our guest is Parag Mital, an Indian-American artist and researcher working at the intersection of AI, simulation, and storytelling. His project The Game of Whispers is an interactive and generative artwork that creates a society of autonomous AI characters whose speech, alliances, and conflicts evolve through their actions and interactions. Enjoy the conversation!
[Music]
Hannah Balber: Welcome to the Art Is Not a Thing podcast, Parag. It's a pleasure to have you here.
Parag: Thank you so much for having me.
Hannah: Your audiovisual artwork The Game of Whispers brings audiences into a world inspired by the Mughal Empire, which ruled over much of what is today India, Pakistan and Afghanistan from the 16th to the 18th century. And in the work, we encounter beautiful gardens, palaces and characters from the Royal Court. To begin, what is The Game of Whispers and what is the project about?
Parag: Oh yeah. So, thank you again so much for having me. The Game of Whispers is telling the story of the Mughal Empire and a very particular time period of Shah Jahan, who has built the Taj Mahal, the Red Fort. This is the fifth Mughal emperor, 1600 present day India, and it's telling the story of his succession. He has these two sons; they're rivalling the throne. And it's also a very contentious point in time in India's history. At that time, you could say, and this is very contentious, to say at least, that there was some level of mysticism and pluralism, and society saw active translations from different faiths. There was music and art. And then there's this other side of history where it's become very orthodox and, you know, this sort of very orthodox Islamic rule. And present-day India kind of has this growing tension again, but through maybe the Orthodox Hindu rule. And so, we see this playing out all over the world: these growing tensions between various communities and this sort of constant back and forth of conflict. And growing up with that, you know, it's hard to understand, you know, like why is this a constant factor of life? And anyways, so The Game of Whispers is meant to be this sort of revisionist narrative, kind of letting the situation unfold and understand other narratives here? Are there other ways to even imagine a better future, let's say? It's also intimately a commentary on artificial intelligence because at its core, each character is piloted by a large language model. And what that means is that there's no script. Each of the characters are effectively generating their actions, their languages. Everything is unscripted moment to moment. And that's interesting for a number of reasons. One is, I'm very curious just what these language models even interpret as history. You know, there are competing narratives of what is the historical record and history itself is this process of abduction, of trying to posit, you know, what may have been through the various sources that are available to us. And so, as these language models proliferate and become more and more ingrained into daily life, what is it that they're encoding in us? That, I think, is a very important question. And how does Empire, for instance, utilize that? In the case of, for instance, spreading narratives, rewriting history, and things like that.
Hannah: What does it look like? Could you describe that? What does the audience encounter when they enter the installation?
Parag: So, there's a five-channel video installation. So, there's five vertically oriented screens, quite large televisions, 70 inches. And they're oriented vertically to try and give this feeling of the miniature painting form, which, if you're not familiar, has a rich history going back to Persian art. And it's a very meticulous craft, very finely detailed. And so, we were inspired by that. Me and my collaborators, we wanted to pay homage to that form but also invoke this sort of game-like universe as well. And so, it's kind of a mixture of both worlds, something that felt for me like Final Fantasy meets Mughal miniature painting in some way. And so, there's five of these screens. Each screen centers on a particular character in this historical period of time. You know, we have Shah Jahan, his wife, Mumtaz Mahal, who's interned in the Taj Mahal. We have two of his sons and a gardener. And these are the kind of central figures of the story, although there's 20 characters in total. And then on the periphery, there are the two speakers with music that's been composed by Sanaya Ardeshir, an Indian musician. And each of the televisions is also communicating what each character is experiencing. So, as they talk to other characters, we hear that come from that individual television as well.
Hannah: What initially inspired this project? How did it start?
Parag: I had been reading a lot of William Dalrymple. If you're not familiar with his work, like The Golden Road. Also listening to a lot of Empire–an incredible podcasts that just talks about, you know, starts off with the birth of corporation, East India Company, and just kind of goes through the role of empire through the ages and focused on this sort of region, in particular. And I found it fascinating because, one, this is a history that's so intimately tied to my own heritage, but also it's a history that I don't really know about through my own parents. It's not a time period that is easy to talk about because there is a lot of pain there, I think. You know, my parents grew up through partition. India is only 75 years old or so. And at that time, this was a very, you know, very violent place, where they grew up. And there's a lot of migrants, there's genocide, just like a lot of things happening in this time period. And it's hard to understand my own history and heritage as a result. So part of the fascination is just, you know, how did I get here? How did I form my beliefs? Part of the fascination is also like, why are my beliefs so different to that of my previous generation? And some of that is embodied in this sort of dynamic between, you know, a fairly orthodox character and a much more pluralist mystic character. And so those tensions are playing out very much like they play out in my own family dynamic.
Hannah: And what drew you to the Mughal Empire and specifically the reign of Shah Jahan as a point of departure for this work? So, what was so fascinating about that period?
Parag: I think it's a period of time in present-day India's history that has so many pivotal time periods. In particular, this moment in time, I feel like is a great inflection point in some way in history. And, you know, listening to William Dalrymple talk about it, or these other historians, it kind of adds more and more weight–oh, this was a really important time period. And how did this company, you know, effectively a company came and colonized this country on behalf of an Empire. And, you know, there's just so many ways that this history has played out. And it's really fascinating, I think, to understand, given present day empires, the present-day Silicon Valley mold of empire and what this looks like. So, it's, in particular, a very fascinating point in time in history. I think just focusing on, for instance, Shah Jahan's own narrative and how he also rewrote his own history multiple times through the form of miniature painting. This is kind of how our story was handed down. Today we have many other devices, of course, some of which is social media. And I find it fascinating to think about how history is being rewritten constantly through, you know, misinformation, through the spread of AI being able to write as much as it likes at any given moment. It's so easy to craft a narrative online. So I think there are a couple of reasons. You know, going into Shah Jahan's own capacity–as any empire would do–for rewriting narrative, but also the sort of present-day commentary on how we're able to rewrite narratives en masse using large networks of agents, for instance.
Hannah: You turn that historical knowledge, that kind of history lesson or also very turbulent period into a very beautiful, almost game-like environment where we see all the different characters walking through the gardens, talking to each other, interacting with each other. Could you walk me through the technical architecture of this piece? How did you build that world and what world did you build there?
Parag: Yeah. So the level design is done by my collaborator, Brianna Browning–brilliant artist as well. We took the Red Fort as the basis for the simulation. And this is the home that Shah Jahan has built for himself at that time. It still exists today. It's a really beautiful place. I highly recommend you go see it. But we took the actual map layout. We excluded the buildings that were outed by the British, of course. And we focused on a couple of areas within this. So, we have the Diwan-i-Khas, the mosque, the Charbagh garden–which is so emblematic of Mughal sites–, and the Lahore Gates–a really beautiful gate of the entrance. And the simulation plays out in this environment. There are lots of, you know, procedural facts as well, like wind and the subtle kind of animations, the sound just so pivotal to the experience. And Udit Duseja and Krishna Jhaveri also helped with this work adding things like peacock sounds, the kind of atmosphere that you might hear at that time.
Hannah: So, then you have the characters which are very central to the story, and you use AI for this project. How? How did you build that, those AI agents and the characters?
Parag: Sure. So our studio has been developing this internal tool called Emergentic, which is now publicly available, but it allows you to craft entire worlds comprised of non-player characters. And these characters are language agents, so they're effectively piloted by language models. It allows you to give them a persona, give them a custom kind of motivation, reasoning, and put them inside of this container or sandbox world effectively, and hit play. And from then on, everything that you see and hear is dictated by those models. So it's unlike traditional games where things are scripted and there's more, a lot more thought, I guess, gone into how dialogue will play out and the situations that are in relation to a player. But in this case, there's no player. You know, the experience isn't for the player. It is a simulation. You know, it's all of these different language agents interacting with one another.
Hannah: And what kind of simulation? Like what is the story? What are the characters talking about or even conspiring about?
Parag: Yeah. So we kept it fairly loose, in some sense. We wanted there to be tension and some resolution, as well. There are multiple sites on the map, in particular, there's this beautiful river behind some of the buildings. And we thought that would be a really fun site for some tension to arise. And so, part of the tooling with Emergentic is to allow you to have some orchestration of this chaotic system effectively. And so, this is kind of the sleight of hand, the authorial creative control that we get to have on the story. And in that, we've specified certain loose elements, like there's a plot to kill somebody, something happens by the river, there's a cover up, you know. And so like with this sort of really basic dynamic and tension and resolution, we start to see all the ways in which this could play out. And we ran maybe, I think, 200+ simulations at that time–almost two years ago–before landing on this simulation as being the one that we would show to our audience in the festival. For a number of reasons, one being there's a very poetic ending, I find. Both Dara and Aurangzeb die. And it's by the gardener's hand that Aurangzeb is ultimately killed, which I just found so beautiful and poetic. And if you know my studio's name is also The Garden in the Machine, and I feel like there's this sort of provocation that I give myself at least, you know, to remember the return to the pastoral in some sense.
[Music]
Ana-Maria Carabelea: Parag Mital is a computational artist and interdisciplinary researcher based in Los Angeles, California. His work explores how meaning forms and how computational systems increasingly shape it. With a scientific background in arts and computational Technologies, he develops living narrative worlds using AI-driven agents and real-time environments, where stories emerge through interaction rather than linear authorship. His work has been shown internationally at the Victoria and Albert Museum, British Film Institute, and Walt Disney Concert Hall. He has collaborated with artists including Massive Attack, Sigur Rós, and David Lynch. His recent work The Game of Whispers was commissioned by the Los Angeles County Museum of Art and premiered in December 2024 at the Serendipity Arts Festival in Goa, India. The technology powering The Game of Whispers is the first of its kind. Mital’s creative technical studio, The Garden in the Machine, has created custom technology for crafting and bringing autonomous characters to life inside games.
[Music]
Hannah: And while you were running those simulations, you said you found one version which was truly poetic to you or made a lot of sense. But were the simulations all that different, or were there any surprising outcomes?
Parag: Yeah. Yeah. I mean, they were very different. You know, ultimately, it's a process of iteration and building the worlds and the story and the backstories and like, you know, having it all resolved in a way that feels poetic and distilled in some way. The final simulation is 15 minutes, you know. And it's not until eight minutes into it that the first murder occurs, and then 12 minutes the second murder occurs. And in previous simulations, you know, maybe the murder happened right away. And we wanted to build some more tension. Or for some reason somebody killed Mumtaz and we're like: Wait a minute, I don't feel like we should be killing the mother right now. It doesn't feel pivotal to the story. And so, there's definitely, you know, authorial control in this process, and playing out the dynamics so that there's the artistic kind of sleight of hand there and just learning as well what the dynamics of the system are. It's a very, I would say, a very novel form of storytelling in a lot of ways. You know, there's no script. You know, we're really designing a world, a container and then hitting play and letting this sort of thing unfold. The closest things that we see like this are in games, you know, which have done this very successfully for 20+ years, but they're often in relation to a player. So what does it mean to create this sort of story without a player? And I think that's kind of what's novel about it.
Hannah: And how were you thinking about big questions like historical narratives and truth and how truth is also maybe something unstable or negotiated, vulnerable to manipulation? How were you thinking about those big issues while you were working on this project?
Parag: Yeah, it was a really big question. You know, working with these language models and building simulations it's really shifted my understanding of truth in a lot of ways. You know, I feel that I grew up as a scientist. I went through academia, you know, like publishing papers. And I, for a long time, really had the single worldview that there is an objective reality, a single truth that we can measure. And this is it. And more and more, I realize that there are so many different realities, and you know, so much of reality is interpreted. Fundamentally, we are limited in our understanding of the world. And so we use society, we use these social structures to communicate and understand more about the world and function with it and survive. But it's such a fascinating idea to think about, you know, what is the kind of universal truth. And so we develop methods–the scientific method–to help us understand it. That method relates a hypothesis-driven way of understanding the world. It can't really prove anything, but it can disprove things. And so, we're kind of chipping away at our idea of what things might look like. And then on the other hand, we have the sort of ancestral, indigenous and faith-based systems of knowledge. And in a lot of ways, they're kind of these creative versions of reality. You know, it's like: Oh, I think the world is like this. You know, so said somebody thousands of years ago. And you know, who's to say, you know, we can't really prove or disprove it. You know, there's so many ways to conceptualize the system, but we're inside it.
Hannah: That's on the question of truth and science. And one aspect that the work, of course, touches is that of how narratives are spread or how false narratives are spread, especially in today's media environment where we see the rise of AI and all of those consequences. What parallels do you see between how narratives are spread in The Game of Whispers’ world and the way narratives circulate now?
Parag : Yeah. You know, these tools are incredibly great at finding a narrative between one scenario and another, and they can just craft an entire narrative, and it's very convincing. Turns out that AI is really good at spreading misinformation. It's employed widely online. We already know this. It's utilized for personalization–when you read your Facebook newsfeed, this is a kind of silo that's been adapted for you and that helps advertisers target narratives that hopefully push you towards purchasing their product, for instance, or believing what they want you to believe. And misinformation is kind of that push, sort of like, you know, if I just give you a little nugget here and a little nugget there, I can kind of push you towards believing or thinking or buying a certain thing. But with artificial intelligence, it's incredibly easy to create content, it's incredibly easy to craft narratives, and it's very hard to validate them. It's very hard to understand even who wrote them, if anybody wrote them. Let's say, you know, when I read my Instagram feed, I have no idea sometimes if this was written by a person. And increasingly, that line is getting blurrier and blurrier. And I think that we're all wrestling with this balance of agency and automation. You know, it's great to utilize these tools for increased efficiency sometimes. I really don't want to use my brain sometimes, and so, great, I'm going to use this tool to help me, whether that's pen and paper or it's this tool that’s just going to auto suggest my entire email for me. But at the same time, we have to maintain our agency. We have to remember that we still have to function in this world. And the people pulling the strings behind those automated systems maybe don't have our best interests, or at least it's not very transparent what their interests are. So, a large part of simulation, I feel like, especially utilizing language models, is to understand: what are the in-built biases in these models? Can we see that as we start to form larger and larger simulations? Can we start to see how they organize? Are they competitive? Are they suggesting that we eat more McDonald's? You know, like what are the in-built biases in these models effectively? And there's a handful of people working on this, especially here in the Bay Area, from alignment and safety researchers. And yes, so I'm finding a lot of research in that world that is inspiring to me these days. And that's fundamentally what we're also working with, both as an artist and a researcher.
Hannah: I feel like many people already are very much aware of this problem or that many people I think, nowadays even tend to mistrust information regarding the source or I feel like many people say: Oh, this might be AI if something is like slightly strange or extraordinary. Do you see this as also a big problem that nowadays there's this mistrust? How would you say has AI already altered our relationship to truth?
Parag : Yeah. I mean, I guess deep fakes are kind of part of this as well. And like we started to see, you know, these mash ups of political campaigns created by generative video. And the writing was on the wall: Hey, at some point it's going to be really hard to tell the difference between a generated piece of media and something that had been created by a human. And depending on your familiarity with these tools as well, it's especially hard to tell the difference. I think this has always been a problem with the media. I think it's especially exacerbated and difficult with artificial intelligence because it's much easier to craft narrative; it's much easier to create media. In a lot of ways, media creation has become democratized in the sense that you don't need a professional studio to create something that looks like a professional video. And that's both a blessing for young creators that may not have the funds to tell their story, but also a curse in the sense that, well, anybody can craft a narrative that can try to convince us, persuade us, misinform us, and we need to effectively have more criticality as a society. We need more education on these tools, on what they're capable of, on how to think critically. And I think fundamentally that's probably the most important issue of our time, is how do we ensure people are able to think critically of the information that they read online?
Hannah: Why did you want to incorporate AI if you're so critical about it, even in your piece? What did you want to show with it?
Parag: Yeah. So throughout my work, you know, 15-20 years working as an artist, utilizing computation and artificial intelligence, AI is both the form–as in I'm using artificial intelligence to create the work– but it's also very often the subject of the work. I am critical of it. It doesn't mean that I don't think that we should be utilizing it, but it is important to be aware of the harms and dangers of it. It is effectively a military technology. We are in an information war for our own minds between societies, between nations and AI is very much the tool of choice for that. So there is a lot of motivation for me to communicate to younger generations, to generations abroad: What is the capacity for AI to form story, to misinform in the context of empire, in the context of present-day narratives?
Hannah: Are you planning on expanding the project or transforming it for the future?
Parag: Yes, I am. We've already expanded it into an updated map. We've really enhanced all the visual details. We've added a player in the end, which allows you to pilot any of the characters, and so then you can spread your own misinformation and see how stories unfold. And yeah, we've also expanded the technical apparatus, Emergentic, allowing for richer kind of orchestration of narrative through that.
Hannah: And what do you hope visitors take away from this project? A feeling, a reflection, a thought?
Parag: I hope you come away with a sense of criticality, a sense of wonder, a sense of inspiration, even, on the capacity for story and art to provoke questions, to also ask more questions, to read about history, to be fascinated by Empire in some sense, and how this is something that's happened before and we've survived empires in the past and will survive current ones too.
Hannah: Thank you very much for this conversation.
Parag: Thank you very much.
[Music]
Ana-Maria Carabelea That’s it for today. Thank you so much for tuning in, and hope you enjoyed it! Today’s episode was brought to you by Radio Ö1 and Ars Electronica. Join us next month for a new episode, and in the meantime, follow us or share the show with someone you think might like it.