I spoke with Ryat Yezbick & Milo Talwani about The Innocence of Unknowing at Tribeca Immersive 2025. See more context in the rough transcript below. (Photo by Mikhail Mishin courtesy of Onassis ONX)
This is a listener-supported podcast through the Voices of VR Patreon.
Music: Fatality
Podcast: Play in new window | Download
Rough Transcript
[00:00:05.458] Kent Bye: The Voices of VR Podcast. Hello, my name is Kent Bye, and welcome to the Voices of VR Podcast. It's a podcast that looks at the structures and forms of immersive storytelling and the future of spatial computing. You can support the podcast at patreon.com slash voicesofvr. So continuing on my coverage of looking at different projects at Tribeca Immersive 2025, today's episode is about a piece called The Innocence of Unknowing by Raya Yazbek and Milo Talwani. So this is a video essay type of piece that is taking a look at the issue of shootings and the choreography of retreat and all sorts of different topics, actually. It's covering white supremacy, racism, police violence, and all sorts of sociological aspects of shootings. And they're taking this footage and they're removing out all the shooters and also then using artificial intelligence and this Socratic dialogue to have conversations with the AI to ask questions. And they're making an argument over the course of what ends up being like three different sections of 40 minutes. And so It's quite an extensive piece that I had a chance to see it performed, almost all of it live. I had to watch some of it pre-recorded version. They would have it running ongoing through all these different sections. There's like three main sections. And in two of those sections, there's a personal memoir section for each of the creators where they talk about their own personal experiences around this topic. But they're kind of using artificial intelligence in a way that they're doing all this kind of prompt engineering on the back end. They're asking questions, but they're asking even more detailed questions through the prompting to be able to have like more of a critical theory, deconstructing all the different systems of power. And they're using it in some ways as a proxy for their own voice, because being transgender creators, they're looking at these other issues in terms of to what degree are you going to be trusting this disembodied voice more than hearing it directly from their own voice. And so there's other aspects that they're exploring there as well. So we're covering all that and more on today's episode of the Voices of VR podcast. So this interview with Riot and Milo happened on Friday, June 6th, 2025 at Tribeca Immersive in New York City, New York. So with that, let's go ahead and dive right in.
[00:02:13.817] Ryat Yezbick: My name is Riot Yazbek. I'm a filmmaker, visual artist, do performance, work with new media. And I'm also a faculty at the Narrative and Emerging Media Program at Arizona State University in L.A. And I, yeah, my work's really informed by my background in cultural anthropology. I was in a PhD program for a little while doing that work. So I think a lot about societal structural analysis of issues, particularly as they pertain to how surveillance impacts our collective being and understanding of each other and everything from home, family, friends, intimacy, all of that.
[00:02:50.658] Milo Talwani: I'm Milo Tolwani, and I work in a lot of different ways. The last year has been spent wrestling with AI and trying to figure out how to, if at all, like work with it in ways that feel nice and non-destructive. I do development work for other people's VR and AR projects as well as my own intermittently. I do small amounts of software engineering as needed. I'm a composer. I'm a filmmaker. I'm a traditional filmmaker. I really like movies and the frame. And I used to... Yeah, I don't know. That's what I do.
[00:03:37.045] Kent Bye: Okay, and maybe each of you could give a bit more context as to your background and your journey into working in the space.
[00:03:42.308] Ryat Yezbick: Yeah, definitely. Yeah, I mentioned my background comes from cultural anthropology and contemporary art. And not to go too far back, but my initial PhD research was on the impacts of 9-11 on the Arab and Muslim American community. in Dearborn, Michigan in 2008. And I was really looking at and analyzing the costs of the need to be perceived as a full citizen by the US government at that time, because many people in the community were not perceived to have the same rights as everyone else. And that kind of line of work around understanding that sort of particular kind of state sponsored violence has been something I've been interrogating ever since as a third generation Lebanese American. So I then went into contemporary art and afterwards got a job at Sundance Institute working with Shari Freelo towards the new frontier programming. And it was at that point that I really started seeing the possibilities of what immersive media could be. and have since just really been continuing to make my work. And this is kind of the first foray into really playing with new media tools with Milo, where we also met at Sundance back in 2017. And I've been resistant to it over the years because I wasn't sure how to have the critical sort of engagement with the medium that I wanted and then with this AI project we've created the innocence of unknowing that you know looks at a news media archive of mass shootings through the lens co-created with an AI we trained as a humanities scholar was really able to find a way to do the same kind of social structural analysis that I've done but through looking at the underlying assumptions operating in the technology in the piece itself And so it's been a winding long road to figure out how do all these things come together in terms of what my interests are, how I work as an artist, and what lies in the immersive media space.
[00:05:43.416] Milo Talwani: My background is in contemporary classical composition. And when I was a child and teenager, that was far and away the most important thing to me in the entire world. I decided when I was 11, I was like, I can either do math or I can do music. And I know which side I come down on. Mostly because I, yeah. And then one of the nice things about like new music spaces, was exposed at a really young age to real-time music generation, algorithmic stuff, electroacoustic composition, and that was sort of my route into lots of other things. I dropped out of college when I was 19 or 20 because the university I was going to kicked out the composers that I wanted to study with. And I was sort of depressed and distressed at the state of music and stuff. And through, I don't know, just fate or whatever, I found myself volunteering for a couple different VR artists, like setting up their work at art galleries. And it was really cool to see... One of the really cool things about immersive media and real-time work that I felt a strong kinship with as a composer is you create the world and the rules of the world in your piece, and you're often doing that basically from scratch every time. And I think that's really compelling. When I was writing contemporary classical music, I was thinking about for each piece on a piece by piece basis, you know, inventing a musical language and grammar and thinking about ways that the audience might find their way into that with the musical languages and grammars that they were already exposed to. So I wanted to be building new stuff, but stuff that was familiar enough that there was a way in to explore. And I think really good immersive art is really intentional about that, is actively creating full thought out, fleshed out ways and structures of being or something. Shortly after, you know, I was like setting stuff up at galleries and stuff. And, you know, in 2016 and 2017, there were not a ton of people who were doing that. So I got a lot of experience and ended up working for a couple of really silly film festivals. And then, you know, because I was well-qualified and willing to work for $14 an hour or whatever, Sundance offered me a job facilitating the submission process for the New Frontier section of the festival, which was the section of the festival focusing on not VR and AR specifically, but that's where that stuff would go at the festival. It had a long history and showed all sorts of works. And that was something that I really loved about working there was it wasn't even immersive media specifically. It was part of this larger art context and art and film context. And I really liked that. Riot was my manager when I started and we worked together for, I don't know, seven or eight or a year, a year. Yeah, yeah, yeah. And then I stuck around for another four or five years and worked, you know, my way up from facilitating the submission process to sort of overseeing that and eventually under Shari Freelow helping to curate the festival. And I also did a fair amount of, like, technical... There's some stuff I'm quite proud about technically, about some of the... There's some stuff I worked really hard to set up and invent to show work while I was there, and that was sort of... Through just like being trusted to build stuff, that was sort of like the first place where I really started thinking about writing software, designing systems, and ways that I could not just use technology, but create technology for connecting people with art and stuff. I quit in 2022. For, I mean, among other reasons, my mom had stage four breast cancer and she's, that's now in remission. And my dad had dementia with Lewy bodies and he's now passed. But yeah, left in 2022 and then really needed, I was really tired and I had started to feel like My work as a curator and as someone who was organizing shows was, in a roundabout way, just doing free advertisement for Facebook. And that started to feel really bad. And so I responded to that by playing a lot of acoustic piano and writing a lot of acoustic piano music. and making several normal movies, normal-style films. And then I had kind of told myself that I was, like, done with that. And then in January of 2024, ASU asked if I would, like, adjunct in their, like, LA-based VRAR grad program, teaching students to build, like, prototype their, like, capstone projects in Unity. And I was like, You know, actually, like maybe working with students and helping people who have no institutional power or whatever, like build stuff like maybe this is like a nice way. Maybe this isn't horrible. And then it was and then it wasn't horrible. It was actually really cool. Two of my now former students are helping us produce this piece and we're sort of integral for production design of the installation. Yeah, and then, you know, Riot and I were, like, happy to, like, I mean, we'd stayed friends, but, like, we're happy to, like, reconnect and be, like, working with each other again. And so a couple months after I started at ASU, Riot asked about doing some engineering work on a previous iteration of the project that we have at the festival and spent a lot of last summer working on that earlier iteration. And that was sort of the beginning of this collaboration. Yeah.
[00:11:52.031] Kent Bye: Yeah, I noticed that we're kind of moving into a new phase of XR where it's artists looking at archives and archives of work and finding new ways of investigating that archive, sometimes using AI tools. And this piece in particular, it felt like we're set in a classroom and you're at lecterns and it felt like, you know, a video essay, but also lecture, a performance. There's a lot of ways that we're looking at footage from the past around shootings in America, but also deconstructing them and then using AI to train that on certain tests that were also creating this sort of like interesting Socratic dialogue where you're speaking with the AI, asking questions and really having them reinterpret what they're seeing and then breaking down what they may not be seeing. And so, yeah, it's kind of an interesting structure and form of digging into this archive. But I'd love it if you could give a bit more context for how the Innocence of Unknowing first started, like where did it begin? And this iterative process of trying to add all these components of the AI with the lecture, with the spatial volumetric capture that's in there, but also just working with like archival film. So it's kind of an interesting hybrid of like an immersive essay film, I guess is a good way to describe it.
[00:12:58.649] Ryat Yezbick: Yeah, I mean, it really kind of began for me about five years ago, actually, now, which is wild to think about how long sometimes art projects take to come to fruition, right? And it's strange because sometimes, you know, things just find you and then you're like, oh, I have to do this thing. But it started for me because I remember watching in 2018 news coverage of the shooting that happened at the YouTube headquarters and the way that that was covered. And I was just looking at the people coming out of the buildings with their arms raised. And I thought, this is such a strange choreography for such a horrific moment in someone's life because it marks both that you're not aggressive, but you're also incredibly vulnerable. You're putting your hands above your head. or on the shoulders of someone in front of you, your chest is, you know, your most vital organs are completely exposed, right? So, you know, and I also, I was also raised in a very charismatic Pentecostal Catholic church where there was a lot of praise and worship. in that position and so I was kind of sort of mulling around these things and thinking about that position of the body and the body specifically and under moments of duress and so it really began for me by examining choreographies of retreat as orchestrated by the police and military over time I started gathering a bunch of news clips of mass shootings for that reason and the piece started with this sort of quick clip you see in the beginning of the film as you start to move back in time to the University of Texas Tower shooting it was one of the first things I made out of the archive because I really wanted to see what does it look like to look at these choreographies over time like what is the physical somatic reality of that and This is also very much so related to immersion, right? When I think about the potential of immersion, it is about the body being somewhere and having an experience. This case, not a very great one, right? So really was sort of trying to pull those threads out. I was just gathering footage off of YouTube for years. I had to take a break in 2020 because I couldn't work on it at that time. It was too intense. So I say five years because I think I started in 2019, took a break, came back to it, was trying to get support for the project and really struggled to find financial backers for this work. You know, I think it's one thing to talk about gun violence prevention in society. But the minute you say you want to talk about state surveillance or police brutality, it's a whole nother thing. No surprise. So, you know, who knows? Maybe I'm missing some places and I need to rethink that. But so, yeah, it just took a while. But when I got the job at ASU working under Nani de la Pena and shortly after having worked with Kamal Sinclair, who I just love and appreciate so deeply and greatly, she's been such a good mentor to me over the years. You know, I was able to get some startup funds and Milo came on. It was like, oh my gosh, you know, Milo is, as you can see, such a brilliant, wonderful human being who is so multi-talented. I mean, they were so young when she started at Sundance and she's just incredible, incredibly smart. And so it was sort of an exciting moment, you know, it was so exciting to be like, oh my gosh, okay, Milo, what do you think? And I was so amazed that you were down to do this piece with me because I was like, I'd been showing it to other people and people were just like, oh, you know, it's a hard, it's a hard subject matter. It takes a really particular person to to want to take that on with you. And so it's just been such a beautiful collaboration, really, and being able to work with Milo on that. And we go back and forth and think about what's working, what's not working. It's a meticulous process, but it's been really fruitful and really gorgeous because we're both very politically aligned. We're both very aesthetically aligned. Actually, we share a lot of those same values and I think have very similar feelings about what makes a good artwork. But archives, I don't even know if I answered your question.
[00:17:04.604] Kent Bye: I was trying to get a sense of how the project began, so it definitely gave a good start. I want to pass it over to you, Myla, just to hear from your perspective coming onto this project and what got you excited about working on it.
[00:17:15.034] Milo Talwani: Yeah, well, so, like, initially I was posed, like, a juicy technical problem, which was the iteration of the project that I was brought on to build was, like, can I build a system through which, like, an LLM could, like, edit a non-narrative 70-minute essay film out of an archive? And specifically this archive, but that's sort of, like, a slightly more general problem. And I was noticing... at that time and to this day sort of trends in the way people are sort of marketing AI, which like don't reflect, I think what is one to the extent that it's good at stuff, what it's actually good at. And two, like people are interested in like AI agents or whatever, and like these like fully autonomous things that like go out and cook stuff up and then like report back. And that seems I'm really fairly bearish on That for a couple of reasons. I mean, I think like video editing, this is sort of a roundabout answer, but like video editing, for instance, is not this linear thing. It's a whole host of back and forth practices and revising. And you don't do one step and then another step and then another step. It's this like really complex workflow. And I hadn't really seen anyone attempting to, one, even with a fully autonomous thing, set up workflows that actually reflected how a human might edit and not a really sanitized, abstracted, false version of what that looks like. But I also think art is for humans. like I fully believe that AI can make good art for AI and maybe there's value in that, but like humans are, I mean, there's a lot of us and we feel a lot of different ways and there's not like a single utility function that defines like what effective art is. This is this like personal and very abstract thing. And I think, you know, you can like get some like optimized average, but no one actually really likes that. You know, I think about like even like humans who are like optimizing for that average or whatever, like fail. I mean, just like the recently low in stitch live action movie did quite well at the box office. And so far as I can tell, no one actually liked it and everyone was kind of upset about it. And so you can like optimize for like making money, but that doesn't mean that this is stuff that you can make stuff that people will go see, but you're not going to make stuff that like people will love. And so I was really interested in how can I, with this earlier iteration of this project, how can I, one, define video editing almost exclusively linguistically, and how can I You know, if I'm willing to throw, oh yeah, and then how can I insert humans back into the process at every single step? So if you're going to have like an LLM understand video or whatever, and there's like other architectures that might allow for different stuff, but like, I didn't have access to that stuff in 2024. I don't really now, you know, like you need to know what is happening in, if not every frame, you know, a lot of frames, a minimum on a second, probably a lot more. And that's a lot of data. That's like a massive amount of data and context windows are limited. And so you need to be able to decide for some types of footage, when do I want more data? When do I want less data? Humans watching video footage don't just see object, object, object, object, object. they see a bunch of stuff and they aggregate that into like what's important and that's useful because you know there's a massive amount of data and you can compress that down to what's useful but what's useful is variable and changes between different types of footage and the specifics of and then like when you're editing There's a lot of different things that might make a good edit. There's a lot of different things that might make a good structure. And so it was cool to build a system that was like, what are we looking for in footage? What kinds of structures are interesting? Like, what is this? And then sort of like to be able to like iterate and go back and forth. I did a lot of work on that and it didn't produce anything good. It could edit videos, you know, and it was responsive in some ways to human feedback. But at no point did I generate anything that I was like, this is beautiful. or moving or like moving beyond like the morbid curiosity of a system kind of failing to understand what it was working with. And so I think at the start of this year, after sort of taking a break for a while, we came up with this new form of the project, which is even more, I don't know, even more involved and sort of like even more back and forth between human and AI. I don't know if we've described it yet, but it's like a four channel, kind of live performed performance piece in which we look at and analyze the archive in like a variety of ways. And so building that, there's like a version of the piece that could run like fully in real time. And I would like to make that like no one's given me money to do that. And I mean, even just like buying a computer that I could do that on. But yeah, but, you know, and we knew that we were sort of like targeting festival and gallery. And so we wanted like stability and like predictableness that would like allow us to like deliver consistently good performances. And so, you know, we would think about what are the topics that we're interested in? like what are the examples in the archive that are related to this? And then I would ask the LLM about a whole bunch of different stuff about that footage. And I would... string together like a variety of computer vision models for a variety of different tasks, object detection, segmentation, pose detection, facial analysis and recognition, like emotion stuff. We have our analyses, we have the computer's analyses. It's like, what vision tasks are indicative of this? What vision tasks help inform these analyses? How can we make this legible to the audience? And so it's like a lot of back and forth of analyzing, generating footage using these computer vision, you know, et cetera, et cetera. And so it's really looping and roundabout.
[00:23:36.351] Ryat Yezbick: Yeah. And in relation to that, too, you mentioned Socratic dialogue. And I think that what you're pointing out as well, Milo, is when the A.I. would give us stuff that and you see this in the piece, it does an incorrect analysis. It thinks it's seeing something that's not happening. Right. So one instance, it thinks it sees a bunch of victims exiting a building and it thinks they're all being arrested by the police because they're being escorted by the police. And so, right, it's taking something like that, that moment of visual misinterpretation, and then going back and saying, ah, okay, why did you assume that? Or what did you see that made you think that? And then getting its analysis again on why it went wrong. And right, AI, as we can see when we use chat, GPT gets very apologetic sometimes. So, oh, I'm sorry, I didn't, you know, whoops, you know, so around such a big thing here. But I think as we drill down, right, the goal was really to get at what are those underlying assumptions at play in the LLM and even in the model we trained after feeding it all of these humanity texts. And what are those base values at the core of what we're looking at here and what assumptions are being made on our behalf, right, by an AI, by technology? Mm-hmm.
[00:24:44.417] Kent Bye: Yeah, I think that's an interesting trend that I'm seeing in terms of the archives is like being able to see these deeper patterns. And, you know, one of the recent interviews that I just did was with Emily Embender and Alex Hanna. They have a book that just came out called The AI Con. And they're really quite critical on AI on a number of fronts. And Emily likes to call large language models as more like synthetic text extruding machines. Part of the point that they're making in the book is that, you know, from the very beginning in 1956 conference at Dartmouth on artificial intelligence, where artificial intelligence was coined, it was funded by DARPA, the military in the context of the Cold War. And so it's always been a technology that is around both consolidating power and automating, but also there's an element of. artificial intelligence as a term is like very ill-defined in the way that it can mean so much. And so there's a bit of a magic trick that we currently have with AI, which is that we are projecting all these human qualities onto the AI to the point where we're almost like surrendering our agency and sovereignty over to it. So in this piece, it's very interesting that you are at the same time using AI to deconstruct the power, but sometimes you're also like as an example, saying that the LLM is thinking, which is something that Emily M. Bender and Alex Hanna say, well, these are not thinking. They're just probabilistically coming up with these associations. They're not thinking. But there's also this aspect of calling it a scholar. And then there's ways in which that we're creating replacements for nurses. And by calling this, OK, this is a nurse. I'm like, no, this is not a human quality. This is just a technology tool. So there's ways in which that you're kind of using that magic trick to kind of like engage in dialogue. But at the same time, yeah, so I'm just curious how you negotiate some of these tricks of how people may be projecting even more power onto the AI, but using the AI in a way that you're helping guide it to then at the same time deconstruct the power systems and structures of a society.
[00:26:34.777] Milo Talwani: Yeah, I mean, I think. You know, we actually sort of took out a portion where the AI system describes it as, like, not really a scholar and not really thinking, and in which it talks about using these words because they're, like, linguistically convenient. And if you have to say every single time, like, three sentences of caveats, like, it becomes really difficult to communicate. I do think it's, like, it's almost like... I don't have access to other people's brains. Like other people are a black box. And I think like a lot of my thought is taking in and recombining almost at random. I think it's almost like not interesting to me. I'm not super interested in like, this is me and maybe this is a cop out, but like, I'm not super interested in what thinking is. I'm glad other people are thinking about that, but I'm interested in the results of that thinking. Is it beautiful? Some people think real beautiful thoughts and I love to hear about them and some people don't. LLMs almost never generate stuff that I think is beautiful. And I think that that's what's relevant to me. I think that the incentives and costs, around who's funding this and why and like who's making it and how are likely to preclude it from becoming beautiful. I think the piece is about, I mean, it's fundamentally about the failures of all of this AI stuff to not just to solve, which is a problem that's too big for it, but to even really understand what's going on. And yeah. Yeah, I don't know if I answered the question exactly, but those are some thoughts.
[00:28:30.103] Ryat Yezbick: Yeah, I mean, I totally agree with the folks that you're referencing. And I'm curious about that book. You know, we did think a lot about this. And as Milo said, we did take some stuff out that did answer a bit of that. But we also were thinking about the visual representation. We didn't want this to be something where we magically appear behind a curtain and there's, you know, this sort of Ozian voice from nowhere. You know, we're like, this is a computer chip, right? And in that monologue too, that part that we did end up taking out for exactly brevity flow, you know, more time needed to make visuals there, et cetera. We took out that bit. So it was like us really talking about how we're telling it what to say. And, you know, I think the idea, we did make a conscientious decision for it to have a name, you know, and then we took out why it named itself what it named itself, and we're interrogating that too, you know. So it's like, it's a bit of a tongue-in-cheek thing to get audience attention. And then also, you know, we are both transgender people, and, you know, we were imagining this being for a general public audience. The reality, sadly, is people see you in a certain way, and they don't want to listen to you, and they tune you out. So we are playing with the possibility of here's a generic automated voice that people are now used to hearing. Are you going to hear what we have to say better if it comes from this AI? So the whole end monologue of the piece, you know, I wrote it initially and then Milo and I came in and edited it together. The AI didn't write that part. That's the only part actually I think that it's saying that it did not write. It was this interest and desire to really have the audience think critically about like, what are they hearing from this thing? Are they digesting it? Did they believe it? Did they hear it better because it came from this disembodied voice? So there's this invitation to think about that, even though we're using the tricks of the magic of anthropomorphizing this thing. We're trying to do it as critically as we can so that the audience actually hears what we're saying. And I mean, I don't know, be interesting study to see. Do people listen better to a disembodied voice? Maybe. I mean, their bias aren't implicated then. I mean, we all have biases, you know, so.
[00:30:42.008] Kent Bye: Yeah. And I think within the context of immersive art piece, there's like leeway to start to push the edges of some of these things, because there are some things that are coming out of the large language model. It's like this sounds written, but also you're training it on a corpus of like humanities texts that are also heavily tuning it towards these perspectives that I don't normally hear from large language models. And so I'm just curious if you can maybe elaborate on that process of. training it with what corpus of data, and then there's also a process of editing where if it's resulting in something that doesn't actually fit with what you agree with, you have the ability to edit in the Socratic dialogue format to create this essay format that is, in that sense, a little bit more contrived and controlled in the way that you're guiding it with that human hand, which based upon what I know from LLMs, there was certainly either the corpus or that guiding part that I was like, okay, this is not the normal experience that you get for most of these large language models.
[00:31:34.304] Milo Talwani: You know, in my experience, and this is, again, I'm not trying to do free advertisements for massive AI companies, but prompting does get you really far. The questions that we ask in the piece, the words that we use to say to ask it are not always the prompts that are required to get those answers. But if you explicitly ask for... I mean, it doesn't necessarily do a good or artful job, but in general, if you are asking communist questions, you will get communist answers. If you are asking abolitionist questions, you will get abolitionist answers. And if you just say, like, what's going on, you're going to get centrist, Marin, wine mom, core, neutral answers. But if you're asking really specific questions, even in my experience with fully off-the-shelf models, you'd be surprised at what you'd get. And then I don't know, like I was doing a lot of explorations last summer with using just like off the shelf models for screenwriting tasks. And a lot of the stuff that I do when I'm like making normal movies, they're often fairly rowdy, like politically, aesthetically, like whatever. And, you know, I would like be asking things that I thought like as a trans woman were like really, really, really funny bits, which it was like, you know, this is like offensive hate speech. And, you know, like one or two, you know, if you like, if you ask it like a couple of different ways, And it would say, like, no, sorry. And then eventually, like, you are being transphobic. Like, I am a trans woman, and I think this is hilarious. It will eventually be like, oh, yeah, sorry about that. Here's this, like, insane thing that, like, you know, they really should not be saying. Um, so yeah, I mean, I think, you know, the fine tuning is really important, but I do think, and again, like I'm not trying to do like advertisements for these companies, but like, it's like, because of these things are like, whatever, probabilistically, like spitting out the likely next word in a sequence, there's a lot of work you can do to set up what is likely gonna be the next word that will really radically change the kinds of answers it's willing to give you.
[00:33:49.671] Ryat Yezbick: Yeah, you're making me think, Milo, AI really does reflect racial politics in the United States, doesn't it? Like when you think about the fact that there is an assumed normalcy of whiteness, right, or assumed given whiteness at play, and then you have to make qualifiers, right? to become incorporated into a system that sees you as an outlier. And AI reflects that. I mean, it's the same thing with our political, the way we have to lobby for ourselves, right? We have to be LGBTQ incorporated. We can't just be a bunch of people who are marginalized. We have to segment ourselves off and we have to name ourselves and we have to like, you know, present ourselves in that specific fashion in order to be seen and heard within this political sphere. And AI asks us to do the same and we have to name and call that out because otherwise the default position is one of, you know, the supposed majority, you know, this patriarchal white neoliberal context in which all of these things have been created by mostly white men and this is not to vilify white men in any way it's just to say this is where it comes from and it's presumed neutral and it is not neutral at all so when we gave it those texts it's true it's like we aren't sure how much maybe they really influenced it i can i can
[00:35:05.914] Milo Talwani: I mean, I do know, and it's honestly, it's not a ton. There are ways that I could have done a better job fine-tuning, like, less naively, but that would have required either generating a bunch of synthetic data, like, off of those texts, and that basically would have required, like, a whole bunch of human intervention in ways that I was, like, I had other stuff to work on, but yeah, could either have, like... You know, and I don't want to, like, because I do think it's, like, important and beautiful to be retraining it, both, like, as a gesture and as, like, and as, like, meaningfully changing the output. But it's, like, it's a lot smaller than you'd think.
[00:35:41.914] Ryat Yezbick: Well, 10 texts compared to an LLM, right? You know, you've got to imagine that input, you know. But the texts are great. Yeah.
[00:35:52.395] Kent Bye: I wanted to comment a few things on my experience of the piece. One is that as I was watching the piece, there's a number of different iterations you have where you're giving a performative aspect of it, where you're giving these little segments where you're diving into deeper details to a whole wide range of different topics that Because it's an essay film, an immersive essay film, it feels the closest to like the tradition of literature where this could be like a book that people really dive in into with footnotes to really dig into it. And so when you're translating it into an immersive performance, then, you know, there's so many different ideas and topics that we don't have the time to really dive into every little nuanced part of that. But I do want to share some moments of revelation that I had watching that was just that. The section you had on the social media saying that a lot of the journalists will take social media footage shot by citizens. So like this kind of blend of citizen journalism turned into the spectacle. And the more the media does that, the less that they're looking at these larger systemic issues that are looking at the root causes of all these different topics. The thing that I really appreciated with Innocence of Unknowing is that how can you start to use AI as a technology and a tool to restore some of the deeper context and relational dynamics of this as a story and use the AI, even if it is kind of very much crafted through the prompting or hand-guided, that you're able to at least have this presumably neutral voice, although obviously we know it's not neutral, but at least this independent investigation of this huge archive of data that is able to find these patterns and then elucidate them in terms of stuff that you wouldn't normally see on a news reporting. I'd love to hear any reflections on that process because that seems to be at the core and the heart of this piece is to restore some of those deeper relational and contextual dynamics that are often being collapsed in these stories and to go back through history and to start to weave together a story that ties shootings back to Vietnam and capitalism and white supremacy and these other spheres of power and influence that are maybe invisible to most of the day-to-day of how news is covered.
[00:37:53.598] Ryat Yezbick: Yeah, I mean, I think it's interesting because when you said Socratic earlier, others have used that to describe previous projects I've done. And it's interesting, you know, seeing this piece live and do its work in this beautiful collaborative way. It's like, oh, right. Like, there's a part, I think, of both of our brains that likes to really look at things from a very big, high perspective, right? Well, I want the whole picture, right? and i think you know for me at least in my social science training it's always for me been about what is not being said what is being assumed and how can i get to the bottom of that and journalists often do the same thing my partner is actually an lgbtq journalist that looks at anti-trans legislation currently which is a heartbreaking endeavor but right when we just take things that face value or when we get pulled into the emotional reality of the horrid horrific reality that we live in and how violent our society can be we really sometimes miss exactly how things are being pulled out from underneath us constantly makes me think of naomi klein's shock doctrine a bit right you know it's like we are deliberately gaslit we are deliberately thrown off or put onto something else because the solution is actually really right in front of us. And people who have a lot of power, money and control don't want us to look at that. So I always just think things are often interrelated, you know, and it goes back to that hands up position. I mean, we wanted to include and have a bigger even conversation around police violence. Right. But it's like it is all there. It is all interconnected. Right. And when you look at something like mass shootings, I mean, it becomes possible to talk about trans rights in that. It becomes possible to talk about labor politics and class warfare. It becomes possible to talk about white supremacy and all kinds of things because it's all embedded in this. And so I think, again, going back to those assumptions, what are we missing? What is not being said? What are we being asked to look at instead as a deliberate tactic so we don't look at the thing that actually needs to be addressed? I think these are questions... Americans and people around the world need to be asking right now because fascism is here. And we need to get better at looking at when we're being lied to. But it's always been happening, frankly, so.
[00:40:15.271] Milo Talwani: I don't know, like earlier I said that I think about a lot of my thinking as taking stuff in and noticing patterns and recombining stuff. I don't know, like when I first started studying contemporary classical music as like 10 or 11, and you'd have to like, you don't just sit down and like, or the way that I was taught and the way that my teacher at the time insisted I do things was, you know, you don't just like sit down and write note, note, note, note, note, note, note, until you get to the end of the piece. You sketch out ideas. You think about how those ideas might relate to each other, if at all, if they belong in the same piece. You think about how you might develop those ideas. You think about the kinds of forms and structures that those ideas suggest or might serve them. And then you slowly sort of start to put this thing together. And I would like bring in sketches or early versions of my pieces to lessons and if I didn't have a good reason like why I had written something or like if I couldn't articulate like how it was related or justified, we'd cross it out. And I think that there's some silly stuff in that, you know, and I don't think that that's necessarily the best way to make compelling music. And I think plenty of music, most of the music I make now, you know, I sit down and I play at the piano until I get something that I love and then that's the thing. But I think that that's sort of central to how my brain works is I or I don't know how you don't think deeply about stuff or something like these are not perfect answers. But yeah, just if you're asking an audience. to look at something you know you should put a bunch of time and time into that and make it really good and you should be able to answer questions about that and if they're coming away with like really obvious stuff that you should have thought of like that reflects poorly on you in my opinion i think that there's like yeah there's like a respect for other people's intelligence there's a respect for your own intelligence i feel embarrassed about stuff when i do a bad job i think in like traditional cinema you know particularly for like real expensive stuff like you have hundreds of people working on every single aspect of it from lighting to sound to camera stuff and there's like often dozens of people working on just the electrical systems that like allow all the stuff to happen you'll shoot 20 takes of the same thing to get exactly right you know like i think that immersive media despite taking a lot of work or maybe because it takes a lot of work and there's not a ton of funding for it to do anything i think people often let themselves off the hook and i really i don't know i it's hard for me to do that and feel proud at the end and i just think that that's true of of everything of art of science of living in the world of being with people yeah
[00:43:15.120] Kent Bye: So I wanted to talk, I guess, the qualities of presence that I had in this experience and then ask a question specifically about some moments in the piece. And so like overall, like the installation is that I'm in a classroom. And so it felt very much like a lecture type of situation where I'm learning a lot around the stuff that you're presenting. And there's the four channel video installation that A lot of times it gets combined into one screen, but sometimes you do separate it out to four different channels. But within the context of there, you're showing a lot of this archival footage, but you're also having like this photogrammetry volumetric scan with all these artifacts that you're zooming around. And so it gives a spatial dimension of some of these different clips or, you know, using depth kit and to put live performance elements that you have overlaid onto things. And so there's a lot of kind of more XR type of visual aesthetic innovations that are happening with the context of you know, making it beyond just the 2D of the archival footage. But we're also in the context of a classroom that has like flowers because, you know, it's talking around school shootings and we're in a classroom and you're meta talking around that. And so there's a ending piece that you come back to again and again that is a little bit of a memorial that I feel like is really powerful with both the music and the flowers and like a montage sequence of all of the school shootings that have happened in the United States. And I think another, I guess, surprising emotional peak of this piece was each of you have like a monologue where you give a little bit more of a memoir or your personal story that is included in this piece. And I thought that, you know, tonally it's interesting because it stands out so separate from the rest of looking at and analyzing the archive. it felt like a real emotional core of like each of your stories and your relationship to the story. And so I'd love to have you maybe elaborate on this kind of memoir piece and monologue that you're able to give a little bit more about your own story and connection to this topic and why you're interested in sharing that. So yeah, just kind of have you elaborate on that.
[00:45:11.684] Milo Talwani: I mean, I think So the memoir part specifically, I mean, we're showing some of the worst moments of so many people's lives and we slow them down and we play them over and over again and we analyze them and talk about them. And I think there's like, not to be like stuff is comparable or whatever, but I do think there's like, Yeah, I just like, I felt like a certain responsibility to be like, I'm doing this to other people, and I should have the, it would be irresponsible to like, not do that. Or yeah, like, I don't quite exactly have the words for that. But it was like, I don't want to do things to people that I'm not willing to be done to myself. And I think that I think there's also like, I mean, one of the things that I feel really strongly about about this piece and this is, you know, like there's not a school shooting at every school, but like these days, like every school is doing school shooter drills. Everyone has in the back of their mind, like some level of fear about that. And maybe you don't think about it a ton. Both of my parents were teachers and I was aware of every single time there was an active shooter on campus or nearby or a drill, you know what I mean? And like, you know, we're doing all of this sometimes fairly cold and analytical stuff. And I feel like it was really important to remember, like, these are real people. These are real people's lives. These are not data points or like to to I'm there and people see my humanity. And, you know, to to sort of make myself another data point, maybe makes it clear that all of those data points correspond to real humans. I also just like, you know, like I said, when I started on this piece, this was an interesting technical challenge to me. And that's a lot of the pleasure of working on this piece. I am not going to lie. It is like it is fun to think about and design some of these systems. I don't really have to think about shootings as I'm just doing a lot of the technical work on the piece. But yeah, just like many Americans, Yeah, just like there's been a lot of guns in my life and people have died. And as we were working on this piece, I only wrote my monologue, I think last month, you know, I knew probably six months ago that I was going to do it. And I really was only like able to sort of like get that out last month. But even before I decided that I was going to was like, oh yeah, like I'm looking at all of this stuff and this has been this through lines throughout my life. And just like, it would be like, I could hide that, you know, but like that would be an intentional decision. And like, why would I, I mean, it was, yeah, it was just not exactly related to the rigor or, but it's just like another way of looking at all of this stuff. Um, yeah.
[00:48:03.365] Ryat Yezbick: Yeah, it's interesting because when we were thinking about what are the categories we want to discuss in this work, and I mentioned that some of my first research when I was, I think, 22 years old was the self-discovery of realizing as a Lebanese American within a greater community that I'm a part of, just the amount of government surveillance that this community's had to endure. And in the last year and a half, two years, since October 7th, It's just been frustrating for me as a Lebanese American person to go into spaces and to hear very well-intentioned people wanting to do something about the horrific violence happening against Palestinians and realizing that no one knows. about this long history in this country of surveillance of Muslim and Arab peoples. Arab and Muslim people have always kind of teetered on this like question of whiteness historically throughout the U.S. Census Bureau, throughout the last century and a half. And I think that's fascinating, not to try to be overly analytic, but to bring it back to the more personal, but of this saying of, are you, you know, this assimilationist kind of in and out, in and out, in and out. And I think a lot of, You know, minority groups in the U.S. have experienced that over time. When I look at Arab and Muslim people and the way Arab identity was not embraced in my own family, you know, I was raised to think of myself as white. We didn't celebrate. I mean, I didn't even know that my grandfather had spoken Arabic in the home until like a couple of years ago. He just wasn't talked about. You know, and so having this awakening, you know, after 9-11, you know, as well, realizing that my family was starting to go, oh, but, you know, I remember hearing my dad saying, I love my dad. He's a wonderful human. But I remember him saying, oh, why do they hate us? And I remember my mom and my sister and I looked at him and said, who's they? Who is they? That's, you know, your family is Lebanese. We're Lebanese. who's they who are we being told is the people that hate us right and so you know I don't even think I've fully gotten that into the piece yet because I been working full-time while editing this I haven't had enough time to reflect but I would like to work that more into the piece because it is about just this intense cultural erasure and the pain of that cultural erasure and the pain of The disavowal of one's heritage because you're told you're inherently violent and wrong and lecherous. I mean, you look at the stereotypes of Arab and Muslim people in Hollywood film over the last century. It's egregious. It's disgusting. And I know people in the black community have very specific examples of this. There's forms of dehumanization. And so I think sharing some of that history that people don't know that specifically happened to this very particular community that is an extension of my community after 9-11 was crucial to feel like it had to share that here because it relates, you know, it really does.
[00:51:15.376] Milo Talwani: I think the monologues serve like a purpose. I mean, you know, that's sort of like the inverse of like a question that we ask at the end, the AI asks at the end of the piece, like, are you more able to hear this because like the AI is this like objective, you know, unbiased robot thing. And I think, you know, those monologues sort of like ask like the opposite question. Like, are you more likely to believe us? You know, like... I have personal experience with this. Like, does that make you believe me? And I don't know what the answer to that is, but I felt like it was important to get both sides of that in there.
[00:51:50.964] Kent Bye: Yeah, I really appreciated it. And I felt like I said, it's another emotional core of the piece that kind of takes a turn that helped me understand the deeper context for how this all came about. So, yeah, I guess as we start to wrap up, I'd love to hear what each of you think the ultimate potential for immersive art and immersive storytelling might be and what it might be able to enable.
[00:52:11.724] Milo Talwani: I think that to me, art is about making good art and beauty in all of its forms. And I think good art is often working with and in response to new forms and new possibilities. And what is good or beautiful is often in relation to that. But that's what it's about to me. I think that art is awesome. And I think that that's like enough. I feel distrustful sometimes when people talk about political things or embodiment thing. Like I think that if you want to do politics, politics is a really effective venue for that. And just watching a piece of art is not inherently political and making art is not political. They might inform people. They might make it easier for people to do politics. good politics or whatever, but it's not a substitute for then going out and actually doing stuff. And that's a really core message of the piece. And I think, you know, people sometimes talk about like embodiment and stuff in ways that I really don't believe in and I think are intellectually dishonest. and yeah i mean i think like if you need to experience something from someone's perspective like in order to like treat them like a human being like you're probably a bad person like we do not get to know the entirety of everyone's experience and we need to be good people and like exist in society regardless of like how well we understand and i also think that like you know you like up until the point where we get like wires and your brains or whatever and like you're beaming in like whole lives or whatever like you're immersed only for as long as you want to be and in very limited ways and i think there's really beautiful art to be made with that technology and some people do a really good job at that and that makes me happy
[00:54:01.400] Ryat Yezbick: Yeah, it's interesting. I think this field has an obsession with the other, right? It's a good point, Milo. You know, coming from, again, cultural anthropology and the colonial history of cultural anthropology is steeped in this desire to understand, right, the other person in that history, obviously in terrible language and horrifically racist ways. But there is this fascination in the field of immersion about wanting to get inside someone else's brain, body, being. And I completely agree. Milo, you said it so well, right? We cannot know. We never know fully what someone else is thinking. And we should be okay with that. There's a beauty in life about not having total access. Gosh, I have been and fallen off my meditation train here, but I'm also a trained meditation teacher. Ultimate immersion is sitting silent with yourself, and that's the thing this society is obsessed with escaping. And so, you know, I see this in a lot of work in the field, people trying to bring people back to themselves, but no amount of entertainment or art form is ultimately going to do that long haul work for you, right? You have to do it for yourself. And, you know, someone, I'm not saying this as I've like completed, it is an everyday dedication to being a good person and to showing up for yourself. And that's where it starts. It starts by showing up for yourself, giving yourself what you need. And, um, So, yeah, I don't know. I don't really know what the edge. It's interesting to think about frontiers in this space because that is, again, that goes back to cultural anthropology and this desire to colonize, to know, you know, what are the edges of knowledge? What is ours and what really shouldn't be ours? And I think that's an interesting question for the field, maybe.
[00:55:54.721] Milo Talwani: There's like. There's technological innovation and stuff that I'm genuinely excited about, but I imagine it's not super dissimilar to the first time you got a new mixer for your oil paints or whoever used linseed oil for the first time or when you realize you can stretch can. It's like the painting or whatever that is the thing that matters, and there's going to be exciting new stuff to enable you to paint better, but I like the painting. And it's really fun to think about pigment. Yeah, but it's about painting.
[00:56:29.754] Kent Bye: Anything else left unsaid that you'd like to say to the broader immersive community?
[00:56:34.083] Ryat Yezbick: I don't think so. Not for me. I mean, I am grateful to this community for all the conversations over the years I've been able to have. I've always felt like I'm a little secret curmudgeon inside of the community. And I do genuinely love so many people here. I mean, I've had such great mentors and friends and did the MIT Open Doc Lab Fellowship the last few years and just met so many kindred spirits. Milo, another kindred spirit, right? So, yeah, I don't know. I also have a lot of love and gratitude for everyone. Mm-hmm.
[00:57:01.696] Milo Talwani: Yeah, I don't know. I don't think I have anything to add other than it's exciting to be here and it's exciting to be with people and see their work.
[00:57:12.856] Kent Bye: Awesome. Well, Riot and Myla, thanks so much for joining me here on the podcast. I really enjoyed the innocence of unknowing. I feel like it's this really interesting intersection of immersive essay film meets like expressions of archive and mixing in of memoir and all these spatial techniques and interrogating AI in a way that it's using the kind of magic of AI, but also deconstructing all these systems of power. It's like a really rich piece that I think could come back to again and again, because it is so rich and dense in terms of all the different insights and ideas that you're exploring. I feel like it could be a book. I feel like it could be like other multimedia formats, like a film could be a firm vision, immersive, interactive. Yeah. I feel like it's a kind of a rich topic that you could go in a lot of different directions. So I just really enjoyed having a chance to explore it here at Tribeca. So thanks again for joining me here on the podcast to help break it all down.
[00:57:58.120] Ryat Yezbick: Thank you, Kent. Appreciate this. It's been a delight to be in conversation with you.
[00:58:02.722] Milo Talwani: Thank you.
[00:58:04.517] Kent Bye: Thanks again for listening to this episode of the Voices of VR podcast. And if you enjoy the podcast, then please do spread the word, tell your friends, and consider becoming a member of the Patreon. This is a supported podcast, and so I do rely upon donations from people like yourself in order to continue to bring you this coverage. So you can become a member and donate today at patreon.com slash voicesofvr. Thanks for listening.
