Using AI to Reclaim and Preserve APIA Heritage
Cindy Olnick 0:00
Today on Save As:
Paul Kim 0:01
You can be scared or you can be critical and you can be skeptical of the technology, while at the same time being competent and understanding and knowledgeable about the tools.
Cindy Olnick 0:20
Welcome to Save As: NextGen Heritage Conservation, an award-winning podcast that glimpses the future of the field with graduate students at the University of Southern California. I’m Cindy Olnick.
Trudi Sandmeier 0:32
And I’m Trudi Sandmeier. So Cindy.
Cindy Olnick 0:36
Yes, Trudi, I’m here. What you got for me?
Trudi Sandmeier 0:39
Well, we’re going to venture into the world of technology today. Are you ready?
Cindy Olnick 0:42
I’m pretty ready. Yeah, yeah. My head’s popped off several times exploring this topic with new alum, Paul Kim. But he did a very creative and innovative master’s thesis for his dual degree in Heritage Conservation and Landscape Architecture and Urbanism in Artificial Intelligence, AI.
Trudi Sandmeier 0:44
Yeah, and not just one kind of AI, but two kinds of AI. Who knew?
Cindy Olnick 0:57
There’s AI and then there’s generative AI, right? There’s a lot of overlap, but, and as Paul describes, you know generative AI is about generating content, text, images, video, and it’s been around for a while, and we’re, you know, starting to see very clearly the benefits, the risks, the trade offs, especially in narratives in our field. Narratives about under recognized communities. And Paul specifically focused on Asian and Pacific Islander American, APIA communities in Los Angeles as his case study for how GenAI perpetuates false narratives, but also how communities can use it to reclaim those narratives and conserve their cultural heritage.
Trudi Sandmeier 1:59
So with great power comes great responsibility. What do you do when these new tools come online, and how do you think about it carefully before you just dive in and see what happens? There’s consequences. So this is an exploration of those issues specifically.
Cindy Olnick 2:19
Yes, and a good case for how we need to just get over it and start using it, because it’s happening with us or without us.
Trudi Sandmeier 2:28
That’s right.
Cindy Olnick 2:29
So our producer, Willa Seidenberg interviewed Paul at her house in the Mar Vista/Palms area of Los Angeles where Paul grew up. Here’s Willa Seidenberg’s interview with Paul Kim.
Willa Seidenberg 2:45
Welcome, Paul, to Save As. Please introduce yourself.
Paul Kim 2:49
My name is Paul Kim. Just recently graduated from the dual master’s program with Landscape Architecture and Heritage Conservation.
Willa Seidenberg 2:57
What led you into landscape architecture?
Paul Kim 3:00
Yeah, so I did a Fulbright in Thailand right after college, which led me to work with a development project. I was working with indigenous groups up in Northern Thailand, cultivating specialty grade coffee and understanding the whole sort of land dynamics, the conflict, the social spatial aspect of that region. I was sort of looking at the Heritage Conservation Program, because I was interested in this idea of cultural landscape, because really wanted to learn this sort of socio spatial dimensions, especially in like a city like Los Angeles. Very diverse, very cosmopolitan. And that sort of idea of also adding on the Heritage Conservation program really solidified when I was in Europe. There’s so many pockets, especially in the bigger cities like Paris, there’s so many different people from all over the world that, just like create that that’s what makes Paris Paris. Not just the buildings, but these sort of interstitial spaces, and then the communities, the fringes. I mean food. I mean like you name it.
Willa Seidenberg 4:09
The name of your thesis is Encoding Counter-Memories: Artificial Intelligence as a Tool for APIA Community Empowerment. What got you interested in writing about this?
Paul Kim 4:25
It was a culmination of like various interests I had in the past, and through training in the landscape architecture program, I realized there’s a real opportunity to use emerging technologies to see if there are ways to explore the things that I’m interested in, particularly cultural landscape. So I jumped on it, started investigating, and I think six months later, the technology with ChatGPT blew up, and that’s when I really thought there might be something there that I need to explore.
Willa Seidenberg 4:57
What’s your primary question that you were looking at in your thesis?
Paul Kim 5:03
Whether emerging technology could create an alternative to the non-dominant narrative, so counter-memory. The main idea is a non-dominant narrative. Non-dominant memory, when you look at like a history book that’s traditionally like what we consider a dominant memory. But let’s say oral history from your grandparents or your neighbor who experienced the L.A. riots, and they might be talking about their own versions of it, and that is, in itself, also a memory, a legitimate history. So the question was like, why is one type of memory or type of narrative considered more valuable or more important or more accurate than another form. And that’s sort of how I started investigating and then came across this idea of memoricide, and memoricide is erasure of memory, or erasure of culture. So I was really interested in that sort of idea of how, whether it’s possible to utilize technology in a way that we can highlight and reinforce this notion that there are multiple ways of looking at history, multiple ways of looking at culture, heritage, preserving it.
Willa Seidenberg 6:16
We lose a lot of history because those counter-memories don’t get saved, and don’t get put into the history books. And so it makes what we do very important, and oral histories, like you mentioned, so so important.
Paul Kim 6:34
What gets a lot of erase in that process, this sort of overemphasis on formal knowledge through textbooks, experts and agronomists coming in without really realizing that there are tacit knowledge that gets passed down from generation to the next generation. I didn’t really think about it and as I was sort of coming up with the whole thesis, but then as I was writing, I realized most of my background, professionally and academically, been really looking at this sort of data management, knowledge management, and activating the knowledge, right, the data that we collected. How do I package it so that we can use the knowledge in a way that’s interesting for other people to actually receive, rather than just purely just getting from atextbook?
Willa Seidenberg 7:18
You mentioned that GenAI is a redistributive technology. Could you explain what that means in practical terms as it relates to API a communities?
Paul Kim 7:33
Initially, I sort of framed it as a democratizing tool, until I spoke to my friend Andrew, who helped me edit this. And he really challenged me the idea it’s like, are you really saying democratizing, or are you trying to say something else, like you’re trying to redistribute the access and power? It’s not a term that I came up with, but I thought this would be a great tool if used appropriately, that it could, you know, almost redistribute the way we sort of talk about our own history. How can we provide access to development of this technology to people that should actually be using this in order to document, collect, analyze their own story.
Willa Seidenberg 8:16
So it’s more like not keeping the creation and the power in the hands of maybe a certain group of tech entrepreneurs who know how to use this, but really getting it out to all of us.
Paul Kim 1 8:29
Absolutely, absolutely and in a very mindful way. Because when we talk about data sovereignty, control of the flow of that information, creation of that information is maintained by select few. But the consumer, we ourselves become almost like a information source. We become like the commodity. Our identity, our attention span, became a commodity. What we’re realizing is that when the coder and the creator of these AI systems share the same sort of profile. What ends up happening is, then they’re gonna interpret things from their specific lens of the group. What ends up happening is, then people that are using it may not actually…
Willa Seidenberg 9:12
Fit that pattern. Maybe first talk about, how can generative AI be used in the heritage conservation world.
Paul Kim 9:24
Actually, it’s already been incorporated deeply into our field, both, as you know, landscape architect, but also historic preservation, heritage conservation, especially when we talk about like, you know, Building Information Management, or a lot of the LIDAR data, GIS mapping. There’s like, several different case studies done in Europe where they could predict the degradation of specific parts of a building based on the climate change pattern. And so there’s a lot that’s happening already, even like the spatial movement because of the earthquake. And there’s this, like much happening right now without us as users realizing that it’s actually being processed in the background with AI technology. I’m not trying to say that this is going to replace our work, but let’s re examine how it’s being already utilized so that we understand and sort of hopefully steer it towards more responsible way, rather than just, you know, being a little bit too late and playing catch up.
Remember that 80/20 rule, like, if we can do something and 80% is automated, and then we can focus on the 20% I think that’s a win. And let me give you an example of it, if I’m a community activist and there are things that I need to do in the back end. Let’s say, if I do a community survey, the actual amount of time it takes collecting their information, processing and analyzing it manually, it takes so long. Let’s say the campaign takes about 10 hours, eight hours of that data, processing, collecting, analyzing, and you only have that two hours to actually meet with people and follow up. Imagine if we reverse that, let the take care of that part, the tedious part. It’s not that I’m not going to be paying attention to the analysis. Of course I will. So imagine that reducing the amount of time takes for more of that menial task, and then freeing up more time where I can actually engage with the community members, and you can, in that process, collect more information, right, whether through sharing their own stories or presenting it with new evidence, like photos or other things, documents. Like that, to me, is worth the incorporation of technology, right, because it frees up that time where you can actually build trust, build relationship, and be part of that, engage with that community.
Willa Seidenberg 11:48
And focus on the things that a human definitely needs to do. What are the most urgent ethical concerns that you have about using GenAI for cultural investigation?
Paul Kim 12:03
We need to really understand that this is a tool that will assist you in the process. And then more of a deeper or technical question that I have. One is how to combat the biases that’s already been encoded, and then second would be the idea around data sovereignty, who owns the information? If I input the information, do I control that information?
Willa Seidenberg 12:27
And I would imagine that those kind of questions are going to be in in the courtroom in the years to come. Right?
Paul Kim 12:34
We’re already talking about that with like music and movies and studios, and also authors. There are a lot of publications, a lot of publishers, authors, editors, are all saying you haven’t credited any of that. So who really owns it? Then lastly will be misinformation, deep fakes. We really don’t know how to still, really figuring out how to combat this, or even just like understanding and what to do with misinformation and false information, as well as deep fakes. So deep fakes are, image that sort of a lot of people are creating. This is like where you overlay someone’s likeness onto an AI system so it recreates an image of someone that’s very similar to that person, if not exactly like that person. So there’s a case where classmates created pornography using their classmate’s image, and that got them into trouble for distributing. These are the extreme cases, obviously, but it’s really pushing our legal system to really understand, like, what are we going to do now that it becomes even much more easier and pervasive now that the technology is like pretty much free to everyone who wants to try it out.
Willa Seidenberg 14:05
So how do you envision that APIA communities, can use this innovative technology, but make sure that there is cultural authenticity?
Paul Kim 14:18
So that is a great question, because that’s something I’m still working on. AI does a great job of collecting and processing massive amounts of data, disparate information, and able to make these connections. So one way is like, give an example of, let’s say, the Japanese internment camps. And there are institutions that are sort of collecting, not just the oral histories, but really trying to present that in a way through the museum or other workshops and other activities. But I think this is a real moment where we can leverage, again, the generative AI technology to see what else could we create, let’s say new ways of presentation or workshop or other ways of engagement. And so I guess, in a way, that ideation of these engagement opportunities, maybe that was that’s something that we can think about and ask AI to help us. Second thing would be more identifying throughlines, patterns, connections that we normally would have missed if you’re manually sitting and going through, you know, hundreds, if not, thousands of documents, and it’s just impossible for a group of people to actually process all that information. So that’s something that we can utilize the technology to say, Are there any themes that you’re coming out that we have? What are the gaps or themes that we overlook?
And my personal interest is more interethnic experiences that often gets overlooked in our academic discourse, especially where, you know, API experiences didn’t happen in a vacuum. Even within the API groups, right, there’s, like, a lot of interethnic I mean, we’re in California, like, it’s impossible for us to say each group is just like their own thing. No, like it happened all together. So imagine it’s impossible for us to really include all that information, trying to analyze and process it. So can we actually use AI technology to help us sort of navigate that process of actually collecting, analyzing and disseminating the sheer volume of information that’s out there. And then again, the most importantly with that information is there a way to activate the knowledge that we’ve accumulated. And I think that, in itself, really becomes powerful tool.
Willa Seidenberg 16:35
Let’s talk about cultural landscapes. How do they serve as vehicles for counter-memory?
Paul Kim 16:40
I did start in the Landscape Architecture program, and then I added on the Heritage Contribution, and it’s precisely because of the aspect of cultural landscape. I always say land doesn’t exist without people. People don’t exist without land. It keeps a lot of the memory there.
Willa Seidenberg 17:03
I was struck by that with Manzanar. That sand blows over, and pretty soon there’s no evidence of the people that were there during the internment camps. And so yeah, it is a real realization that nothing is just blank. There’s a lot going on.
Paul Kim 17:21
And good example is like that when you say history of Manzanar, like, that’s memoricide, right? Literally, like, neglect is also part of memoricide. Another example that I can give is Catalina Island. There’s like, actual archeological evidence of people like, you know, Chinese and Japanese fishermen who actually live there to help with the whole fishery, we don’t really know about it. What are some ways we can activate it like to so that will actually get people to be interested.
Willa Seidenberg 17:53
What are ways that GenAI can help facilitate some of that?
Paul Kim 17:57
So are there ways to empower community members to really sort of take control of the actual cultural products, whether that’s cultural tourism, whether that’s the type of restaurants and the way they share their story, recipe, like cookbooks, right? Like there’s so many skilled cooks that may not be trained to get that title Chef, but they’re incredible. Language barrier is a real thing, but that’s sort of another thing that generative AI can really sort of level the playing field. It does a great job of translating. How do we make it so that it can help in that process?
Willa Seidenberg 18:31
So that doesn’t have to be a barrier to having access to those resources.
Paul Kim 18:38
Another example that I came across was the Little Tokyo Hype Squad. I think what they’re doing is amazing. It’s a group that helps a lot of the legacy businesses in Little Tokyo, updating them to the modern standard in terms of providing access to their online shopping or having a social media presence so that more people, especially the younger crowd can actually understand what they are doing and the importance of supporting these legacy businesses. Or let’s say, helping them process like sales data or a number of visits to their website to analyze it. They don’t need to be sort of savvy with technology. They can just go and ask AI to help process that information.
Willa Seidenberg 19:21
I like that idea that it’s one of the tools in our tool chest. It doesn’t do everything, but there’s still a role for us as people.
Paul Kim 19:29
Exactly.
Willa Seidenberg 19:32
What do you think the implications of using GenAI in urban planning or landscape architecture are, especially when you’re talking about communities that have been marginalized by the dominant communities?.
Paul Kim 19:48
I think we’re still trying to sort of figure out how to utilize it without just being a tool to help with the visualization aspect of it and a lot of people are using that as, like, an ideation phase, but I think you can go even deeper in terms of, like, again, using the amount of data that’s available to analyze it.
Willa Seidenberg 20:10
So, let’s talk a little bit about a concept that you write about in your thesis called Asian Futurism. And it’s not just Asian Futurism. There’s a lot of different variations, but in terms of the APIA community, how does that expand the possibilities for combating this?
Paul Kim 20:32
Basically, it’s a movement to reclaim our own narrative, in order to reimagine the future through our own cultural perspective. I wanted to give more than just a handful of examples, because was a few years ago, the Black Panther was a number one box office hit, and everybody was talking about futurism. But they only interpret it in hyper-future in a very culturally African, whatever that means, it’s a continent. But right, like they just, you know, focus on this, like idea African. Everything was looking very much like the visual language of Black Panther. So it almost like that’s what it represented for your typical audience. Afro Futurism is much more than that. It’s also a philosophy. It’s an ideology. It can be interpreted through music, video, art, because it is the easiest way to communicate to other people without having to read a dissertation about what it means.
Willa Seidenberg 21:34
Right.
Paul Kim 21:35
The core idea was that, how can we as African Americans come together and talk about and reclaim their narrative in order to envision the future that they want from their own cultural perspective. And that’s sort of how Asian Futurism sort of started, too, from moving away from this Techno Orientalist stereotype to reclaiming their own narrative, reclaiming their own history, reclaiming their own identity and empowering them to sort of think about how they want their future to be. There’s a rapper named Jason Chu and it was almost a rebuttal to people that were saying, Why are Asian Americans stealing from Black culture? You expect these, a lot of the like refugees from Asian countries to settle right in middle of Bel Air or Beverly Hills? No, they settled in low income areas, and they had to intermingle with Black community members and Black neighborhoods. And when you grow up in that environment, then you also become very much part of that community. Does that mean they’re not part of that community. They have generations of people living in these communities, and I really want all of us to remember that future can’t exist without the present or the past, without reclaiming our own narrative, controlling our own narrative, and being empowered to do so, it will be almost impossible to create our own future.
Willa Seidenberg 23:06
So are there some specific models that you could mention that APIA communities could adopt to better control the narratives of their past and their future?
Paul Kim 23:20
That’s a very difficult question, because if I figure that out, I wouldn’t be sitting here, I’d be starting a business. I’m trying to figure out ways to really think about this notion of data sovereignty. We can’t wait for the legal system or even the government to really come up with an idea. I think it will lag behind if we wait for that. So proactively, what are some ways that we can safeguard a potential right misuse of our own information, our own own narrative? We have to collaborate with other community members and other organizations as well, including, hopefully in the future, local government as well. Something that we need to really actively think about, this notion of data sovereignty. Who owns it? How do we handle misuse and misinformation? How do we prevent digital slop? So digital slop is just information that is just like, not very relevant, not very informative, but they create for the sake of creating. Once it goes out of control, gets out of hand, that’s impossible to reel it back in, and that’s what’s sort of what we’re experiencing.
Willa Seidenberg 24:40
So one of the things that you talked quite a bit about in your thesis was involving communities in this work and with using it. And how would you envision that happening? Because, you know, most of us aren’t as sophisticated with the use of this technology yet.
Paul Kim 25:06
Well, first and foremost, I think tech literacy is important. There are countries that do a better job, especially like in, you know, South Korea, you know, Singapore, they do a really good job. I think the challenge, then, is not just looking from a perspective that we all have to be tech savvy or understand the technology behind it, but you know, different community members can have different levels of understanding, right? So the idea is, how do we provide guardrails for certain types of people who may not really care to learn too much about the usage of it. The goal is then to create some form of a standard operating procedure, a manual of some sort, while providing some guardrails to prevent people from being taken advantage of or stealing the information and misusing them. For me, community members is meeting where they are at in that moment, while also thinking about ways we can sort of scale up or down depending on level of understanding, interest, usage needs. And that’s something that I’m working on right now, and hopefully be able to do a PhD on it, as I’m currently in the process of submitting applications.
Willa Seidenberg 26:26
What would success look like if you had a very community-led endeavor that was using GenAI to further preserve the past and use it for how we’re going forward?
Paul Kim 26:42
For me, success would be different groups coming up with their own use cases and creating their own versions of this AI tool. One great example. It’s the Maori language revitalization example that I use with the Te Hiku Media. I’m not sure if I’m pronouncing it right. New Zealand, Te Hiku Media leverage the AI tool to collect and document the Maori language, because it’s a dying language. So what they’re trying to do is preserve this language using more of this community sourced, open source process. They asked community members to record few phrases and created this sort of language app where people can not only learn but also translate and, you know, do a lot of different things with it. So to me, that was like a really interesting case, a very specific goal, community participation, community ownership and access has also been open to anybody that’s willing to use it with at the same time, in the back end, they were working with other organizations to talk about this sort of notion of data sovereignty. Who owns it? They don’t want another company to create a language app, only to sell it back to us if we want to learn that language, right, right? Another example I wanted to share, because we’re in the L.A. context, and I grew up in this neighborhood. One of my favorite dishes that my grandma cooks is this stewed beef dish, which I later discovered that she learned from her Jewish neighbor, which I thought growing up, was a Korean dish. Obviously, right. she would change it to her own flavor too. When I was invited to my friend and I was like, wait like, latke and this beef dish, I was like, wait what? When I asked my grandma, she was like, oh, yeah, I learned from this lady so and so. Because like, how can you recreate this?
Willa Seidenberg 28:51
And that’s what’s exciting about being in a place where there are so many different people.
Paul Kim 28:56
I’m still discovering things about my own family. But imagine if that gets, you know, multiplied by number of people that we have in the city, those small, micro interactions that may not get documented in a formal history book. And these are all countermeasures that we can, you know, collect and and to showcase that it, you know, we don’t live in a vacuum, like we share information, we share food, music, way we dress, and these are things that are happening, even language, right?
Willa Seidenberg 29:25
How do you think that API, a communities now can ensure that their stories, their histories, are not distorted by the algorithmic bias? What are practical steps that people can do?
Paul Kim 29:43
Participation is important. It’s easy for us to sort of reject emerging technology, because it can be scary if you don’t understand the process that happens behind the closed door. And it’s like so, spitting out these new information that you’ve never seen before. Of course, you’ll be overwhelmed if you’re not, you know, familiar with it. There’s definitely this, like, genuine fear against the technology with, like, potential job loss and uncertainty with various industry that could just, you know, dissolve because of it. So I think it’s important for us to emphasize the fact that you can be scared or you can be critical, and you can be sort of skeptical of the technology, while at the same time being competent and understanding and knowledgeable about the tools about the technology too. So I always want people to at least be familiar with it, at least understand what it’s doing.
Willa Seidenberg 30:46
So, you know, first steps are just use it. Use it on a day-to-day basis,
Paul Kim 30:53
But then also at the same time, I think institutions like whether it’s a nonprofit or business or any forms of entities, I think would really have to come to the fact that when you use it, that information will get swallowed up by this larger language model to train itself. So we have to be very careful and understand this limitations. But there’s also pitfalls that we can avoid, and I can’t say avoid it, because the biases have already been encoded. So even if you don’t use it and stay away from it, you’re still going to be impacted negatively by not participating.
Willa Seidenberg 31:34
So what are you doing now? Now that you’ve finished your master’s degree?
Paul Kim 31:39
I’m in the process of applying for a PhD program in Europe to study continuation of this idea, right, looking at, you know, AI as a potential tool for empowerment, or investigating this sort of social, spatial aspect of metropolitan cities, how that can compare. These are things that I’m sort of wanting to incorporate into my practice, which I think it’s like a logical next step from this thesis, because this is purely looking at it now I want to test it out.
Willa Seidenberg 32:14
And didn’t you win a fellowship recently?
Paul Kim 32:17
I’m working with Asian and Pacific Islander Americans in Historic Preservation (APIAHP). They graciously accepted me to really think about the notion of data sovereignty and potentially creating an a GenAI chatbot that specializes in Asian and Pacific Islander narratives and histories, and hopefully that it could be a tool, potentially to help assist in the nomination forms or other workshops, or social media postings, things like that, where community members can actually engage like a tool that they can use to start the process engage with in the preservation world.
Willa Seidenberg 32:59
Yes. Well, thank you, Paul, this was really interesting. My brain kind of hurts right now. I see a dissertation in your future, and you’ll be able to get much more into it.
Paul Kim 33:13
Because this field is relatively new. It’s really hard to engage with other people that are potentially doing something similar with AI technology. So I want to see any listeners, if they are interested, or if they are working on projects that are similar, or there’s like some kind of potential collaborative opportunity. I’d love to, you know, to hear from them.
Willa Seidenberg 33:34
Okay, Save As listeners, there you go. This is a great opportunity to further the field even more so. Thank you, Paul.
Paul Kim 33:43
Thank you.
Trudi Sandmeier 33:49
Thank you to Willa and Paul for digging into this topic, which is really, truly just the tip of the iceberg and a technology that’s evolving on a minute-to-minute basis. This is going to be something we’re talking about for the foreseeable future.
Cindy Olnick 34:06
Yes, absolutely. And thank goodness we have amazing graduates like Paul who are interested in this and really exploring it in a very direct hands-on way, and helping to translate all of this for use in our field.
Trudi Sandmeier 34:24
Yeah, it’s the future of the past. All right, we have one more episode this season, my friends. We’re really looking forward to this conversation. We’re going to focus on the results of a class that happened this last semester with our awesome alumna and faculty member Sian Winship and the students who focused on really making the history of women in the city of Los Angeles visible in some places that have already been landmarked but the history of women was kind of left out of the story, so we’re trying to correct that oversight.
Cindy Olnick 35:05
We hope you’ll join us. And as always, the best way not to miss an episode is to subscribe. If you haven’t already. We’d love it if you told a friend and gave us a review, perhaps all of the things. So again, thank you so much, and as always, for photos and show notes and links and transcript and all of the good things, please see our episode page at SaveAs.place.
Trudi Sandmeier 35:40
This episode was produced by Willa Seidenberg, with help from Cindy Olnick. Our original theme music is by Steven Conley, save as is a production of the Heritage Conservation program in the School of Architecture at the University of Southern California.
Transcribed by https://otter.ai
