Tech Start-Up Founder Is Treating Alzheimer’s with AI

Will AI be able to reconstruct lost memories? Today on Digital Disruption, we’re joined by Pau Aleikum Garcia, media designer and founder of Domestic Data Streamers. Pau is a media designer and the founder of Domestic Data Streamers, a Barcelona-based studio that has been creating immersive “info-experiences” and generative AI projects since 2013. His work spans over 45 countries, collaborating with institutions like the United Nations, Barcelona City Hall, and Citizen Lab. Pau is the chair of the Master in Data in Design program at ELISAVA University and is a guest lecturer at institutions including the School of Visual Arts (NY), the Hong Kong Design Institute, the Royal College of Art (London), and the Barcelona School of Economics. In 2021, he founded HeyHuman!, an artist residency program that merges music, journalism, and data to drive artistic research and social justice. Pau sits down with Geoff to talk about how his organization is unlocking some of the most human experiences through synthetic memories. This initiative uses generative AI to recreate lost visual memories, particularly for refugees, migrants, and those affected by war or displacement. They explain how AI is not just shaping the future but also helping to reclaim the past, as well as the conversation that led to the creation of this technology. They also discuss ethical considerations of working with vulnerable communities. The project is also being explored in dementia care, where AI-generated memories are being tested as a tool for reminiscence therapy. In this episode: 0:00 Intro 0:52 What are synthetic memories? 2:18 The making of Domestic Data Steamers 3:39 Why images and physical spaces matter for cultural identity 5:00 Dystopian, futuristic idea 8:33 Working with migrant communities 9:47 Where this technology is being deployed 12:11 Low-cost and accessible technology 13:50 The importance of emotional connections 15:08 The images are artificially generated representations of memory vectors 17:30 Working with memories 21:40 Commercial interest 22:43 AI is never neutral 23:55 Skeptic reader 30:09 Understanding the data 34:14 The cost of corruption 36:37 How to stay informed 40:41 The future of AI is bright 44:20 Domestic Data Streamers and the many projects 48:30 Accomplishments and the future Connect with Pau: LinkedIn: https://www.linkedin.com/in/pauerrr/ Twitter: https://x.com/pau_aleikum Website: https://www.domesticstreamers.com/ Visit our website: https://www.infotech.com/Follow us on YouTube: https://www.youtube.com/@InfoTechRG

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:01Geoff Nielson?

  I'm super excited to talk to Pau today. Pau and the work being done with Domestic Data Streamers. They are using design, they're using art, they're using data to unlock some of the most human experiences with us. They've eviscerated the idea that AI is this kind of cold, neutral technology. And I want to hear a lot about what they're doing and how they're helping us humanize all of our experiences with AI. So it should be a great conversation.

• 0:33Geoff Nielson?

  This is Digital Disruption. I'm Jeff Nielsen. And joining me today is Pau Aleikum Garcia of the Domestic Data Streamers. Pau, I'm so excited to talk to you today about everything that the Domestic Data Streamers have been working on in the world of synthetic memories and beyond. So maybe we could just start off by talking a little bit about what are synthetic memories? What is this project you've been working on? And what's the impact of it?

• 0:58Speaker 01

  Synthetic memories is the reconstruction of a visual memory from someone using generative AI models. And it's a project we started two years ago already, almost three years ago. And come from the idea of helping people that because war, political persecution, natural disasters, or any of those reasons, they have had to leave their country, sometimes migrating and leaving a lot of stuff behind. Between all this stuff, sometimes photo albums, diaries, and a big part of this subjective, individual, cultural, visual heritage that they had.

• 1:44Speaker 01

  And the idea evolved into a set of methodologies, tools, and now a whole, almost foundation that is doing the projects all over the world now.

• 1:55Geoff Nielson?

  That's so amazing and it's just, it's honestly one of the coolest applications of AI, you know, of generative AI I've ever come across. How did, you know, you talk about specifically talking about it with, you know, refugees or people who have been, you know, displaced by war. Was that the genesis of the idea? How did this come about initially as an idea?

• 2:18Speaker 01

  So at the studio, Domestic Data Streamers, we have been working for a very long time with data and in a lot of different social contexts. One of those was in 2014, we were in Greece, in Athens, in one of the biggest refugee crises in Europe in the last decades. And almost, I think more than 3 million refugees from Syria came over the border and were located in different places in town, in Athens specifically. So we were helping them allocate themselves in different old schools, abandoned hospitals and spaces that they could facilitate to inhabit for a while.

• 2:59Speaker 01

  And I remember one night, we were having dinner with an elderly woman and she told me, "Well, Ba, I'm not afraid of being a refugee now. What I'm afraid is that my grandkids will be refugees for a very long time." And I said, "How come?" And she said, "Well, the thing is that our home does not exist anymore. Like, our neighborhoods don't exist, our photo albums don't exist. Like, a big part of the things that somehow build up our identity don't exist. And when my grandkids ask, 'Where do I come from?', there will be very few things that can answer that." And that, I think, was kind of the seed, that conversation,

• 3:44Speaker 01

  the trigger of understanding how important are images and physical spaces for a cultural identity and subjective one. So in 2020, we were one of the early studios that could actually do some testing with OpenAI, DALL-E 2, and we were already Like trying to figure out what we could do with this amazing technology. And we said, okay, what if we use it for that specific case? So we started to do like very simple experiments first year in Barcelona, in our hometown. And we invited like several participants, elderly people to come and talk about their memories. And from there, we kind of saw the impact that it had in them.

• 4:28Speaker 01

  Just seeing images that before were only in their heads, just seeing right away. After just orally expressing their memories, they will be able to actually see them in front. And and just that impact, seeing how that was very impactful for them, kind of brought the enough energy to build up the project.

• 4:48Geoff Nielson?

  And what what was that impact? You know, how how did they react to it? And and you know, how how did you feel hearing their reactions?

• 4:57Speaker 01

  Um, so for me, it was very weird, you know, like the the project itself. When you, when we thought about this, it felt a bit like minority report, you know, like this kind of almost dystopian futuristic idea. Um, but because I don't know, we had like this instinct that this could be helpful. We tried nevertheless. And then, um, the first, um, participants were people that we know or people that were friends of friends or the grandparents of people that we knew, right? So kind of close environment. And so we were comfortable there. And and with these participants, um, it for each one, it was a bit different.

• 5:39Speaker 01

  But I remember the the second participant we that that explained her memory, his memory. He was talking about, um, his brother who died almost ten years ago. He was talking about when they were kids and they used to do like.

• 5:57Speaker 01

  Some staff in the field with their mother and so on. And then we generated an image like of that specific situation. And when Juan was the participant, when he saw the image, he started to cry. Like, it was like, oh, I never had an image of my brother like this. And this is a very beloved memory of me. So just being able to actually tangibilize, like physicalize this into an image was impactful. I think for other people was a bit different. Some people actually wanted to create images of sometimes traumatic things.

• 6:36Speaker 01

  Things that happened to them that were important, but were not happy and happy memories. And for them, in this case, it was more about dignifying the past, about saying, that happened. And this is another proof that this happened. It's not only in my head. This is another media that I can use now to explain what really happened to me and how I got here. And so there are several stories and we have been doing that for like these two years with a lot of different people that really wanted to share traumatic events and situations that were like really, really difficult for them.

• 7:16Speaker 01

  But that somehow externalizing them and transforming them into an image that is a universal language, to say so, was also kind of liberating. Can I have seen a lot of people feeling kind of a release when they see the image?

• 7:34Geoff Nielson?

  Yeah, well, I mean, it's so, it's so interesting. And as you said, it's like it's such an emotional, you know, journey for these people, whether it's positive emotion or trauma or both. You know, did you ever feel, Pow, doing this, that, you know, oh, maybe we've gone too far as a studio or maybe there's something happening here that's now kind of beyond our level of expertise? And, you know, do we work directly with people? Do we work with other, you know, organizations? Or, you know, did you feel kind of firmly in control of where this is going?

• 8:07Speaker 01

  I think at the beginning, we didn't. And that's why we only worked with fa- family members and people really close to us. And then when we started to see that we could create a methodology that was more controlled and that there were like certain factors that we could actually um isolate and and guide very well, we like felt very well. But then, when we started to work with, for example, migrant communities, we started to join forces with different organizations. So something that we always do is whenever we generate synthetic memories, and there are always three people involved. There is the interviewee, the interviewer, and the prompter.

• 8:51Speaker 01

  The interviewer also has to be from the same community as the interviewee. That means that they share this common background, they understand each other in their mother tongue and so on. And that's a way that you can really reduce the the potential risks, right, of misunderstanding, misunderstanding, disalignment and so on. And also, i- i- when I say community, I say someone that have experienced something similar, right? Um, even if sometimes it's a bit difficult because the time gap and the different political moments that we're living in different parts of the world, but we always try to work with local organizations that somehow already have like this local community.

• 9:38Geoff Nielson?

  Right. So so what sorts of, you know, you you mentioned, you know, Greece and the Syrian refugee camps. Um, what what sort of places are you actually deploying this right now? And, you know, do you have a broader vision for where it could be? And, you know, where does this go from here?

• 9:54Speaker 01

  Yeah. I think for us, probably one of the biggest moments was when we started to work with

• 10:03Speaker 01

  um care care homes and nursing homes.

• 10:07Geoff Nielson?

  Yeah.

• 10:07Speaker 01

  Um, so I I had this friend who was actually like working in a nursing home. He was taking care of 12 patients with dementia. And he was telling me about this methodology that he was following called reminiscent therapy. And this is a kind of therapy, maybe you have heard about that. It's very common, specifically in the States, but since the sixties has kind of grown. And and it's the use of music, for example, music from the times of that person, um, or food, or smell, or the voice of someone that that person recognizes. And to actually trigger this kind of cognitive enhancement, right? Uh, for a while, it's temporary.

• 10:51Speaker 01

  And and and then we said, what if we kind of cross these methodologies and do uh some pilot experiments? So we did a couple of pilot experiments here in Barcelona. And the results were really, really good. The engagement of of of of the participants were really high. And then right now we are actually in deployment of this more clinical perspective on this. And and we have partnerships with the University of Toronto, uh, the University of of the British Columbia. And we are exploring exactly how, which is the best way to integrate that in a safe, also ethical, protected and controlled way into a clinical environment. Um, but the first results have been like really, really good.

• 11:34Speaker 01

  And I think in that sense, for me, the future goes very much in that direction, understanding that this could be actually used for uh um therapy. And it could be like, I have seen it, but it's subjective and it's not Yeah. um academic in that sense. But I think with uh three, four years of of research with the right partners, we will get into a a really good position and this could be actually deployed and used in a lot of different centers for very like, like really almost no cost. Um, so that's what is interesting, that it's something that it's very accessible. You don't need like really hardcore computing power.

• 12:17Speaker 01

  You don't need like more than an internet connection and just like twenty dollars per month thing. So it's something that could be very accessible for a lot of people and could actually improve the life, not only of the patients with dementia, but also family members, caregivers, and as well as all the communities that have lost all these images, right? Because of the reasons other than health.

• 12:40Geoff Nielson?

  How one of the things that you know, I found so amazing about this and what caught me by surprise is when you actually look at the, you know, the footage of this, it's not, it's not picture perfect, right? It's not like this high quality, everything is perfect fidelity, you know, it looks like Hollywood, um, which is one of the concerns that people have about AI video is, you know, it's got these shortcomings. But, you know, having heard you talk about it, that's not really a shortcoming, right? Like it almost creates an advantage that it has, uh, you know, a dreamlike quality. How did that, you know, how does that work? And how are you, you know, using that to your advantage?

• 13:18Speaker 01

  It was purely accidental. And as we were like really testing the very, very, very early models of, of generative AI with images, um, this, this is how these algorithms were at that point. But then when we started to use like more realistic models, we started to see that the engagement started to be poorer, like lower, like people started to feel less connected to the memories we were generating. And, and we stayed there for a while. We were like, what's going on? Maybe we are doing something wrong. Um, and then we, one day we did the experiment of going back to one of the early models and then

• 14:03Speaker 01

  like participants started to engage again. And then we said, okay, that was the problem. And, and then we kind of, this is, and this is of course an hypothesis, but our hypothesis is that it's not the factual accuracy of an image that actually brings this connection, but the emotional embedding, right? Whenever something is super realistic, you try, without knowing, subconsciously, you try to find the things that were not exactly like you remember. While when you do something that is more blurry and define a bit more abstract, and it's more about the symbolic representation and, and therefore you're not focused on the detail because you understand that this is not a photo of the moment. And, and then

• 14:46Speaker 01

  what we are creating here is instead of an image and or a photography, we are creating an image that works as a vector of memory, right? This, if I say that this ring that I have here, it was a ring that was given to me by my grandmother.

• 15:01Speaker 01

  This ring will not only be a ring, will be my grandmother ring. And this will be a vector to my memory, to my grandmother and so on. So I think these images are like artificial generations of vectors of memory.

• 15:14Geoff Nielson?

  Right, so they're more like, they can be a vessel that people then project their own memories on to versus being, you know, perfectly accurate.

• 15:22Speaker 01

  Exactly, and I think blurriness and non-definition also has a lot to do and match very well with the way our memory works. Because our memory, although if you have photographic memory, it is a very low, a low percent of society. Like, most of us, like, the way we remember things is we remember a couple of things of that situation. And then every time we remember, we reconstruct the rest. And I think it's very similar to how these models work. They do probabilistic imagination at some point. And when they are more blurry, they focus more on the things that you specifically that say that were like this.

• 16:03Speaker 01

  So I think that that's one common thing between early models and memories, human memories, biologic memories.

• 16:12Geoff Nielson?

  Yeah, it's so interesting. And it is not something I would have expected, but it's like, it's it's just really nice that you can do it. And I'm sure it makes it easier to do it with less, you know, power or less, you know, advanced models, which is so cool. Is there, you know, as you talk about this method methodology and how you get this right and walk through the sensitivity of this, were there other lessons or other things that you found? You know, oh, this is something you have to get right that makes a big deal, you know, for the interviewee?

• 16:46Speaker 01

  So I think the most important thing is that

• 16:51Speaker 01

  Everyone knows what are you gonna do because it's a weird interaction to say, hey, we're gonna reconstruct visual memories from your past. And for most of the people, it's kind of a science fiction. So something that we always do is we do a bit of like kind of play. We play with the models. We try to generate, for example, we start with dreams and then we show how the models work. And when they know what they can do and that they've, when they feel that they are under control of the situation, then we can go and start actually asking things. And we always start by what is your earliest memory?

• 17:32Speaker 01

  And from there, we start to kind of find which is the best way to create a symbolic memory of a whole story. Because normally, images try to encapsulate a story that is bigger than the image itself, right? It's not because you've remembered a specific place or a specific person. It's more about what that person meant to you or what that place meant to you. And depending on that, you will build the image in one way or the other. So it's always about the story and having kind of the instinct to find the right image to represent the story.

• 18:12Geoff Nielson?

  Yeah, yeah, and it makes sense. You don't want to start with something traumatic. You want to, you know, ease them into the, into what the technology can do and, and, you know, build up toward it, it sounds like.

• 18:24Speaker 01

  Yeah. And then what is very important, of course, is to explain that you are not creating an image of that moment, a specific photographic image, because otherwise people can feel disappointed at some point, like, hey.

• 18:36Geoff Nielson?

  Setting the expectations.

• 18:37Speaker 01

  Yes, where is my mother? I was expecting to see. And, and then, yes. And so I think setting up the expectations, showing examples of other things that have been done before are important. And, and then you have to, if you are part, an interviewer or an interviewer or a prompter, you have to just. Um, be ready for whatever because sometimes people connect like right away. And sometimes it takes time for people to just go through this and then say, ah, no, this was not exactly like this. It was during the night, this was, and the table was green, not brown. And then you need like to fine tune and find the right way to to get to the image.

• 19:21Speaker 01

  And sometimes you don't get to the image, you need to find another memory, right? Um, I remember this um, kind of interview that we had with a ninety-four-year-old man, an engineer. And he remembered very well when she wa- when he was four years old. And the first time that he was in a car. And and he remembered the model of the car, he remembered the color of the seat. He remembered everything in so much detail that for us was really difficult to reconstruct that image. And we spent almost an hour and a half building up that memory.

• 19:58Geoff Nielson?

  The right prompt, yeah.

• 20:00Speaker 01

  When he saw it at the end, he said, yes. And it was, it's a joy for everyone, like, we had been one hour and a half just figuring out that it's, but it it was worth it. So sometimes you need patience on that.

• 20:17Geoff Nielson?

  It it it's such an amazing technology and, um, you know, Pal, he- hearing how you're using it. You know, it's very clearly centered on using it as a force for good, how to, you know, how to help people, you know, in need, whether it's, you know, refugees or people who are trying to reconstruct their memories, you know, be because of, you know, Alzheimer's or, you know, advanced stage. Have have you been approached or are you working with any more commercial organizations that are interested in this for, you know, corporate use or marketing or anything like that? And, you know, if so, what does that look like? And and does that application concern you at all? Or how do you feel about it being used more broadly?

• 20:56Speaker 01

  Um, I mean, we have collaborated with Google, but it has been with Google Arts, which at the end is a foundation and it's trying to push for this project to to expand through other spaces. So I think the commercial use of this, I think I saw something, I I saw like this company that were actually doing something similar and for people who were almost dying. And they wanted to reconstruct like the book of their life. Um, and this was kind of a commercial project, not very successful. I think the marketing was quite creepy, but but I I think, I guess, at some point, there will be commercial interest.

• 21:39Speaker 01

  In our case, this is part of a bigger research that we are doing with artificial intelligence, specifically generative. Um, that is trying to understand that, of course, there is this dichotomy, no? There is these people saying this is the end of the world, Harari, um, this will destroy everything that we know and so on. And we have these other people with this very naive perspective that artificial intelligence will solve all our problems. We will not need to work anymore or, um, just it will find all the cures to our diseases. Um, and I I think we are a bit in between in like this critical perspective on how we can use that, having into account the the responsibility that it have to hold something like this.

• 22:24Speaker 01

  And I I think really for our generation, we are, it's too late to be pessimistic. So we we really need to build up stuff with this amazing, but we have to say it, it's really, really amazing technology that can actually change a lot of things, but it it's um, it's not neutral. Like this kind of technologies, artificial intelligence is never neutral. And I I think it was Nicolas Carr who said that the the value of a well-made tool um, lies not only in what it produces for us, but what it produces in us, within us, right? So I I think when we are creating systems, technologies, methodologies, tools, I always try to think more in that dimension.

• 23:09Speaker 01

  And maybe that's why we are not so interested in like this more commercial aspect of of the technology. Although, of course, it drives the market, it drives a lot of things, but. But sometimes it doesn't drive a lot of impact in people, right? Um, so yes, I think Synthetic Memories is a good example of this. Of an experiment that started as a kind of artistic experiment and experience, and then started to grow from there. And now it's funded mainly with research funds. So it's people that is really like hopeful that this could be actually used for dementia patients,

• 23:47Speaker 01

  which is a gigantic problem in the world right now with a, with a society that is getting older and older. So, yeah, I think that's that one way. But we have like so many other projects like this one. Um, Skeptic Reader, which is a plugin that we released a month ago. It's a plugin that we like published online for free. And, and it's a kind of a bullshit detector for the internet. So you can use it in YouTube, in any new media channel. And it, it uses like standard L L Ms to actually find logical fallacies and, and ways that articles could be manipulating or just biased towards one side and not the other. And just not giving all the information or contextual information.

• 24:32Speaker 01

  So you can actually have a critical opinion. And, and this again is another example of how we can use this technology to actually engage with information in a deeper way. I think Synthetic Memory is more about how we can connect with each other and with our past in a deeper way using this tech. And Skeptic Reader, it's a technology to actually be more critical. Instead of just delegating if something is truth or not, just try to be more critical about what you read.

• 25:03Geoff Nielson?

  And sorry, how I missed it? What's that, what's that technology called?

• 25:06Speaker 01

  Skeptic Reader.

• 25:07Geoff Nielson?

  Okay, Skeptic Reader, yes.

• 25:09Speaker 01

  Yeah, it's a Chrome and Firefox plugin and it's very.

• 25:15Geoff Nielson?

  Yeah, as you said, it's a bullshit detector, right?

• 25:19Speaker 01

  Exactly. If someone is listening to this podcast, they can actually try it with this podcast.

• 25:24Geoff Nielson?

  Also, it uses audio too, not just text.

• 25:27Speaker 01

  Yes, well, it uses the transcript of YouTube.

• 25:31Geoff Nielson?

  I'm so curious to hear to try it later and see if we're full of shit on this podcast. That's hilarious.

• 25:38Speaker 01

  It always finds something and I'm sorry. Always find a way.

• 25:44Geoff Nielson?

  Yeah. So do you typically point it at journalism websites or, I don't know, government websites? What is it typically pointed at?

• 25:53Speaker 01

  Yes, mainly like media outlets. So it's more about like from Fox News to The Guardian and everything in between. But of course, now a lot of people is using YouTube as an information channel and any kind of like channel that is more like publishing news. Any kind of updates on wars, on situations, political situations, natural disasters and so on.

• 26:20Geoff Nielson?

  And how long does it? You've just got me curious about it. How long does it take to actually analyze an article or a piece of media?

• 26:27Speaker 01

  It's, we're talking about five seconds, something like this. Yeah.

• 26:30Geoff Nielson?

  Yes, I mean, that's such a powerful tool for consumers to be able to, you know, before they even consume something, is this worth my time? You know, what kind of slant does it have?

• 26:39Speaker 01

  I'm starting to doubt that there is something that we are skeptic about our own creation, about skeptic reader. That is because people is lazy in general. I don't want to say that, but a lot of people is very like likes comfortability. And this plugin actually makes you uncomfortable sometimes because, for example, I, there are journalists that I love and I love how they write and how they think and so on. And I started to use this plugin with them because I will not use it with journalists that I already can smell the bullshit. I will use it with the journalists that I like. And then I started to see certain biases and certain situations that.

• 27:20Speaker 01

  I said, oh, this is making me think more and more and more, right? And sometimes you need, yeah, I mean, it's a, I think it's more for the people that is using it right now that I know mainly come from the academy, right? So it's people that is already in that space of research and really want to go to the to the the end of the stuff.

• 27:45Geoff Nielson?

  Well, and and there's a, there's a separate issue. I mean, maybe there's two separate issues. The first one is, um, are you actually willing to stop reading journalists that you like? You know, if if there's bias? And then the second one is, are there any, is there any journalism that is truly free of bias or close to free of bias? Because if there isn't anywhere to turn to where there isn't that bias, you're you're kind of stuck, right?

• 28:08Speaker 01

  Well, I I I don't think so. I don't think that because something is biased, you should stop reading it.

• 28:13Geoff Nielson?

  Okay, it's awareness.

• 28:15Speaker 01

  Yes, it's it's just understanding that this is not neutral, that this is not the truth. It's just perspective, right? And I think that is missed most of the time. And and right away in this moment of polarization in most of the Western world, you can feel that, like, people say things as if they were the only truth possible. And and this is very dangerous, I think. And Skeptic Reader actually was designed to kind of depolarize, to say, even if you read people that you like, you can find problems in the in these arguments. So, yeah, it's also about showing that there is a lot of gray areas.

• 28:55Speaker 01

  And I think now simplistic messages are very at fashion. And I think we need more, sometimes more gray, more spaces of, yeah, I don't know.

• 29:07Geoff Nielson?

  More skepticism and nuance.

• 29:10Speaker 01

  Yes, and not only skepticism, it's also about creating common spaces that can join forces of people from different perspectives. And I think if you can be critical with the people and journalists that you like, then you are opening a door for also people that think totally different to to share this conversation and think, be more aware that you can be similar to them in in so many ways.

• 29:37Geoff Nielson?

  Right.

• 29:37Speaker 01

  And so it's more about like building tools for critical thinking. Yeah.

• 29:41Geoff Nielson?

  No, I love that. And, you know, to me, it comes back to what you were talking about with this, you know, sort of mission of how do you use these tools to, you know, unlock something within us? Or, you know, humanize us more in some way. Is that, am I articulating that properly in terms of your, you know, sort of mission for the studio? Would you frame that a little bit differently?

• 30:03Speaker 01

  Well, so the studio domestic data streamers, actually, the focus is to fight indifference towards data, which is how we can actually transform statistics, data, and like this very raw kind of information landscape into something that can actually make people move. It was not long ago that there was like this huge new hit in Europe, but it was, I think, something even bigger in the States, that it was the cost of corruption in the country, in the whole continent. And, and it was like really hardcore. It was over 90 billion, something like this. So a figure that when I saw it, I was angry, but I was not angry about the figure.

• 30:44Speaker 01

  I was angry about the journalists that put the figure there because no one can understand the difference between 1 billion and 90 billion. Like, only billionaires, that is a very small fraction of society. So I said, look, like, we truly need systems that can actually make us understand the situation because the bigger realities we're living to, they are big numbers. Like, when we talk about massive migration, climate change, economy, like, all these things has to do with big numbers and statistics. And we, our brains are not, we don't have the hardware to actually understand what is hidden behind these numbers. And if we don't understand that, how are we gonna do anything to change it? If we don't get angry because there is corruption,

• 31:26Speaker 01

  because we cannot give a dimension to this corruption, how are we gonna do anything to change it? And so for me, actually, transforming this raw data into something that is comprehensible, that people can interact and understand in a deeper level, is a way of making a change, of making people want to do things differently. And, and this is like the small parts of like this whole ecosystem that we try to tackle with the studio, that is transforming data and information and information technology into systems that can help us understand better what is hidden behind.

• 32:02Geoff Nielson?

  Right. So for people, for people out there who, you know, have data but are struggling to tell stories with it, or are struggling to, you know, change behaviors or outcomes with it, do you, do you have any principles or any, you know, beliefs around what are the best tactics for getting people's attention beyond just, you know, throwing a big number at them?

• 32:25Speaker 01

  I think the most important one is to talk about something that is important to the people that you're talking at. I think familiarity, that's why we use the name domestic, because we are always trying to figure out the most domestic, daily way to, to express information. And the biggest, I think, the most common example is whenever there is a fire, that they give the size of the fire in swimming pools, in Olympic swimming pools. And, and this kind of brings kind of an idea of how big is that area, right? So bring

• 33:05Speaker 01

  always metaphors. And we can only understand something when we can connect it to something that we already understand, right?

• 33:14Geoff Nielson?

  Yeah.

• 33:15Speaker 01

  We need metaphors. And that's why it's so important to use art, poetry, and, and design to actually bring this data to life. That's one, I think, very important idea behind communicating information. And the second one is to truly talk about something, why this is important. Because sometimes we just grab information and we put it there, but we don't talk about the consequences of this information being there. So I think a lot of times we forget the why. Why I should care about this big number. And this is more about contextualization. Give people context. So whenever you show the information, they

• 33:56Speaker 01

  say, wow, okay, we need to change something, right? Otherwise, it's just numbers. And numbers live there in this abstract, symbolic, I don't know, universe.

• 34:11Geoff Nielson?

  So applying that back to the, you know, the case of the the cost of corruption, did you, did you come to a way to represent that in a way to get people's attention? Was there, was there an answer in that case? Or is that one you're still figuring out?

• 34:25Speaker 01

  Of course, we do a lot of projects around this. And the first thing that we do is pointing out whenever there is data that is not being explained correctly. So for example, banks are very good at explaining data, but they explain it in a way that is very comfortable for them. So for example, a bank will always say that they have recovered 1,400 properties that were in default. They will never say that they have left 1,400 homes uninhabited and 1,400 families out of their homes.

• 34:56Geoff Nielson?

  Right, yeah.

• 34:58Speaker 01

  So it's also about understanding the vocabulary that you use around information and the euphemisms that are used sometimes to talk about things. So something, sometimes it's just about giving clarity. What are you actually saying when you are saying this? What are you hiding? Because data, of course, can hide a lot of information. Language, I think, not only describes the world, but also constructs it, right? And if we define, and if we create a language for information to be there, correct, a very clear language, people will understand the world and will construct a world that is closer to the reality.

• 35:41Speaker 01

  And I think right now we live in a, in a situation or in a moment that a big part of the society is kind of disentangled with the other. Like, we are like so separated because we don't use even the same language of data, the same language of words to describe the same, the specific same world.

• 36:02Geoff Nielson?

  So I'm just, I'm just thinking about that, Pow, because you know, you mentioned earlier that AI is not a neutral tool, right? It can be used, it's a tool, it can be used for anything. And it sounds like you believe the same thing about data, right? That despite our best intentions, data is not inherently neutral. It can be, you know, manipulated. So how do you, how do you use that to depolarize? Like, is there a way to use this to bring people together? Or what do we have to do to make sure that we're more informed about that? And can, you know, I think you mentioned earlier, like, find common spaces?

• 36:40Speaker 01

  Yeah, going to this idea of not neutrality, I think that there is a story that I love that was about the, during, like, one of the first rebellions of the Jews in Rome was over the data, was over the census, because they did not want to be counted. Counting something in the first, is the first form of power, because it is understanding everything in its parts. So it's like giving kind of the size of something so you can control it. And this problem is still happening in some parts of South America, because being in the census means that you have to be part of the military, right?

• 37:22Speaker 01

  And a lot of people is trying to hide their kids, so they are not on the census, so the militars cannot come to this house and pick up the kids when they are in age to go and join the militars, right? So data has always had something to do with power and control, and that's why I mean data, and specifically artificial intelligence, is never neutral, because it's something disconnected to a worldview, and, and someone holds it, like, it's at the end a, a, a piece of information that you can use in, in your advance. And at the same time, data is kind of reductionist, because it's...

• 38:00Speaker 01

  It's beautiful and it's very tempting to distribute the entire world into a single code, right? And universal law that can govern any kind of phenomena, like two hemispheres, five continents, male and female, animal and vegetable, singular and plural, right, left, four seasons. Like, we like to classify things into categories, right? It's very comfortable because it gives you like this false feeling that you control it, that you understand it, right? And we try to do that to, of course, have a feeling that everything is a bit under control and we can comprehend it. And actually, it has helped us a lot, but unfortunately, the world does not work like this. It has never worked like this.

• 38:45Speaker 01

  And, and I think a big part of the polarization we are living right now is because this classification bias that we have, that we say, oh, this person is a liberal, or this person is, I don't, you already use all the prejudices around that group to classify that person. And, and this is becoming more and more because I think algorithms and artificial intelligence kind of give more power to this idea, to the idea that everything can be classified and predicted. So, how LLMs work are based on this, on prejudices, on biases, very strong

• 39:24Speaker 01

  biases that are very functional, of course, but that kind of power or empower the intrinsic bias that we have as human beings. And I think the way to de-escalate that is to kind of break these borders. So, for example, Skeptic Riddler is a way of doing that as we are showing the gray areas, even if they're in the people that you like. And synthetic memories, for me, is a way of doing that because we are doing the same with different generations of people. Now, you can see what other people went through, right? And you can see your eyes to see that you don't need language, like, it's just a visual reconstruction on that. But I think it has also to do a lot with the fields of, and the professions

• 40:08Speaker 01

  that have always been leading certain tools. So, for example, artificial intelligence is a, is a field that is prominently

• 40:19Speaker 01

  Developed by white men like you and me from from western countries. And and and most of the times they come from engineering and like informatic engineering schools. So that's a very small portion of the world and that's a very like.

• 40:37Geoff Nielson?

  It's biased, right? They are they brings its own bias to the, yeah.

• 40:41Speaker 01

  So I I think the future for artificial intelligence are a brighter future. It will be, of course, a part of a more diverse set of people thinking about it, like from lawyers to journalists, to poets, to all kinds of people from different walks, different countries, trying to figure out which is the right data set, which is the right bias and building not one general AI, but several very diverse, very biased to towards different things, eh, models that can be used for some things, one for some others. And I think that will be very helpful for the polarization, the fact that we don't have one way, but several of them.

• 41:24Geoff Nielson?

  So so this to me is one of the big, you know, trends and and risks, I think, of our time, right? Is can we democratize AI? Can it go down a path of being, you know, open sourced and for everybody? Versus how much of it is controlled by the same four, you know, mega corporations? And they have all the data about everybody. Are you are you optimistic about that power? Are you worried about it? How do you see that unfolding? And and, you know, what role, if any, do you see, you know, your organization is playing there?

• 41:56Speaker 01

  You know, I I I was very pessimistic when Deep Sea was released. And I think the open source kind of had a kick. I I don't know anymore because the the the stage is changing so fast that it's really difficult to see if private corporations or open source will kind of go wild and and and win this kind of battle.

• 42:23Speaker 01

  I I certainly see artificial intelligence as a kind of a centralization of power. Because at the end, it's it's.

• 42:33Speaker 01

  For a big corporation, it's easier to integrate in the already existing tools that we are already using.

• 42:38Geoff Nielson?

  Yeah.

• 42:39Speaker 01

  So it will concentrate more power. And of course, the other way, the other day I was talking with someone at Token AI and they were telling me, well, you know, the thing is that the data that you could get from Google is very, you can get a lot of information of someone, of course. But the data that you can get from the conversations of ChatGPT from one individual, we are going to another level of intimacy, right? And I think it was Harari who said, we are going from the battle of attention, from all these social media assets and so on, like I want to stay you the maximum amount of time in my platform,

• 43:20Speaker 01

  to a battle of intimacy that is more, not so much about the amount of time, but about the amount of trust that you give to a certain private company. And for me, this is like probably the most dangerous part, that we are going like beyond a certain amount of, I know, control and trust to certain corporate institutions that of course,

• 43:44Speaker 01

  even though they try their best sometimes to do like the right thing.

• 43:48Geoff Nielson?

  Maybe, maybe.

• 43:50Speaker 01

  They are corporate institutions and they have stakeholders and they have to make a revenue out of it. So they will prioritize the revenue over certain social well-being as we have seen with Meta and other organizations. So I think that's very problematic. But of course, as long as there are other options, it's okay.

• 44:14Geoff Nielson?

  Yeah.

• 44:14Speaker 01

  I think.

• 44:15Geoff Nielson?

  Yeah. So I want to come back to, that was super, super enlightening. It gave me a lot to think about and I could, I could follow that train of thought for a very, very long time. But, but just, you know, in the interest of time as well, I do want to come back to domestic data streamers and some of the projects, you know, you're working on. I mean, where do you see this technology going in the next few years? And what, you know, upcoming projects, you know, that you're willing to talk about, do you have in the pipeline?

• 44:46Speaker 01

  So right now what we are exploring is how we can help certain organizations with information entanglement, that we say. So for example, here in Spain, one of the, well, we were talking the other day with, it's kind of a public service institution for, for children that have no parents, right? So they are hosted in a public institution and, and they have a lot of kids there and sadly they don't have enough time to publish all the, to report all the things that are happening there because the ministry asked them to report on every

• 45:28Speaker 01

  little thing that happened. So a big part of the time of the caregivers is focused on reporting, on bureaucracy. Yeah, not care. No care. So actually, what was scary to me is that one of the social workers told me that from the five days of work, there are three that are spent in bureaucracy. Oh my gosh. And two only in taking care of the kids. And, and then, so what we are trying to do is like workflows that for example, they can use voice notes from like Telegram or Signal and they will be transformed directly into a report with a write, a language, everything correct. And then they will need to draft it in just like five minutes instead of two hours.

• 46:11Speaker 01

  And, and only with that, we were like kind of getting that we were actually improving two days less of work in bureaucracy. So that's of course, two days more of caregiving and then taking care of the kids. And this for me like really enlightening. And other than this, we are doing projects mainly with cultural institutions and museums. We, we will be releasing a new project at the Barbican in London, I think in April, March, I'm not sure about. And, and other than this, we are doing several projects more connected to the way museums can use artificial intelligence in their environment. So for example, how you can explain history

• 46:55Speaker 01

  from a more contemporary perspective, how you can personalize the narratives to the different audiences that you have, and this kind of. Examples and exercises that we like really enjoy.

• 47:11Geoff Nielson?

  Oh, that's so, that's so interesting and you know, comes back, I guess to the, you know, to the viewpoint of how do you make it for individual people versus just one, you know, kind of blanket engagement.

• 47:22Speaker 01

  Well, you know, one of the last project we did, it was really fun. And it was around a facial that we were using these facial recognition models to detect whenever a politician was falling asleep into the Congress or the Parliament. So we created these bots that will connect to streaming cameras from Congress of over thirty-six different countries and create head maps of the sleepiest areas of the Congress and Parliament. Because of course, no one is watching these things.

• 47:54Geoff Nielson?

  No, no, yeah.

• 47:55Speaker 01

  But we can now with these bots. So I think that that's an interesting project because it turns a technology of control in a technology of accountability.

• 48:05Geoff Nielson?

  Right.

• 48:06Speaker 01

  So the idea is to release this algorithm and make it open source for anyone that want to use it in their own Parliament and probably in a couple of months.

• 48:15Geoff Nielson?

  I love that. You know, Pow, just, just taking a moment to reflect on, you know, the, the amazing and interesting stuff you're doing around, as you just said, accountability, but also, you know, recovering memories, you know, democratizing AI, humanizing data, you know, you, you mentioned that, you know, domestic, you know, the, the domestic data streamers project started more than 10 years ago. Did you have any idea that this is where this was going? And, you know, how does it compare in your mind to that, you know, original vision of what you guys would be up to?

• 48:48Speaker 01

  Um, no, no, there was no vision at all at the beginning of this studio. It was kind of, yes, I think a group of friends with Alexandra, Danny, Juan and Paul and Axel, trying to figure out like which was the, the right way to explore data and art, which was the thing that we were curious about. I think there was a, a mythical quote from Bruce Moe, the designer, and, and he said something like, If you're a design studio, the main job you have, as the name suggests, is to study, right? And I think this is something that we have always tried to do.

• 49:27Speaker 01

  And to understand that our work is not only about creating things, but to understand them. And I think that that that kind of have brought us here. And in not just solving problems on trying to find the right solution for a specific need of a client or a partner, but more trying to understand what is the problem that is behind and the technologies that exist in the environment. And it's like having the right environment to later on find solutions or find ideas that can help with that. And sometimes are very lateral, not what you expected at all. So in that sense, it was never expected to get here and do this kind of projects.

• 50:13Geoff Nielson?

  I love that. That's, you know, it's so exciting and the stuff you're doing is so cool. So I can't wait to see, you know, what's coming out of the studio next. I want to say a big thank you, Pao, for joining us today. You've given me lots and lots to think about, so really appreciate it. Thanks for being here.

• 50:30Speaker 01

  Thanks for having me.