Queering AI – online panel on art and AI

Show video

Meine Damen und Herrren, Mesdames et Messieurs, Ladies, Gentlemen, and everyone between and beyond the binary. Welcome to the panel “Queering AI”, which is part of the series of talks "Taming AI" a project by the intelligent.museum by ZKM | Zentrum für Kunst und Medien Karlsruhe and the Deutsches Museum. My name is Margherita Pevere, I'm an artist scholar, and today I have the great pleasure to be moderating a panel with Janne Kummer, artist, Sara Morais dos Santos Bruss, scholar, and Jake Elwes, artist. I started this panel by evoking a movie cabaret in its interpretation by Zizi, the deep fake great drag act, which is an artwork by Jake. We will hear more about this and the occasion we are here today is to discuss imagine potential, hopefully open up and expand what Queering AI may be.

We have been living in times where queering has become a force of unsettling binaries, unsettling normatives. It can be disturbing, it can be disrupting, but it can also be future shaping and opening new ways of being together in the world. So there's an ethical aspect in queering. We will get more into it in a bit. And then we've also been living in times where AI has become a little bit of a key word, a little bit of a trend, a little bit of great advancement in technology, which has also been monopolized by a big corporations.

There is an nice German term for this "Spannungsfeld" which is the tension field between two poles. So in this panel, we can expand this tension field between what queer can be and what AI can mean when we take AI away from the monopoly of big corporation. And for this reason, we have artists and scholars here who have been employing and questioning AI from a variety of perspectives. I will start with asking some question to each one of them so that they have the chance to introduce their work.

And then we will continue the conversation drawing from their practice. My first question is for Jake, because in the ZiZi project using your own works, you started with the idea of queering data sets, data sets are a group of data, that are used for training facial recognition. Mostly, not only, but in the case of your artwork with images of gender fluidity, transness, queerness, and drag kings, queens, monsters and things.

It is then developed into a deep fake drag cabaret. So there's a fascinating tension between the construction of drag identity, how much fantasy, how much sincerity, how much cost, how much artists cost in each act, and the extraction of data in deep fake algorithms. Can you please tell us a bit more about the project, how it started, how you work with the London drag scene to develop it? Because I'm particularly interested in how you bridge the care and preparation of drag acts and the footage you had to train with the algorithm with them. Absolutely. So yeah, this is a project I've been working on for the last few years with 21 of London and the UK's top drag artists. Like you said, drag kings and drag queens and drag things, make sure we had a real diversity which to me represents a lot more of the drag and queer community that we get in London than maybe RuPaul's Drag Race would.

So yeah, like you said, the project started back in 2019 with a piece called "queering the date set". So here is a little video clip. Can you see it okay? Of queering the dataset. So effectively what's going on here is I've taken, like you said, a sort of standardized dataset of 70,000 faces, but these were very normative faces which was being used to train facial recognition systems and injected just a thousand images of queer bodies and makeup and faces. And effectively it broke down these systems and plays with this place of otherness and queerness.

And I think this is something that we should be constantly asking ourselves in a way, as a queer community, is, who are building these systems and who are they building them for and to serve? And do we want to be recognizing these systems or do we want to be breaking them down and messing with them and confusing them? So this is where the project started if you like. And from here we started think about, about the ethics of deep fakes, of how deep fakes are often used in a non-consensual and in an exploitative way for really insidious purposes like deep fake pornography. So in a way, how can we, in quite a tongue-in-cheek way, and I think this is a sort of queerness as well, it's like you said, there's that constant tension or contradiction between making something really fun and playful here, but also working with these technologies which can be misused and exploited. So for us, we were like thinking about how could we use deep fake technology to uplift our community rather than to further oppress the queer community as a performance tool and to celebrate. So here, effectively what we did was we wanted to make sure i that everybody involved in this project knew exactly how their data was going to be used and got paid when their data was used. We created our own data sets, which I think again is a really important point because right now all these conversations we're having around large models, around GPT, around Dall-E are talking about data sets, which have been created by companies.

And I don't think so many artists are really challenging what these data sets are doing. And common crawl, one of the most common ones used underlying a lot of these models, has actually a list of words which automatically filters out. has actually a list of words which automatically filters out.

And a lot of those words are words of queerness, words that queer activists will use a lot. And all of those websites, that section of the internet automatically gets removed from these systems. So for us, we were like, how can we just train an AI on queerness? What does that mean? And can we use these tools in a consensual way? So going back to this project, this evolved.

So moving from a kind of exploration of a latent space after we've queered it with drag performers, I wanted to think how can we actually collaborate with the drag performers and bring them into the project, but make sure that they knew how the data was being used and also that with the deep fakes, I wanted to make sure that the only people who were given permission to reanimate the other queer bodies in the project were people from the community and people from within our cast. So here, effectively, we have one of the drag performers called Caramel, who performs a Beyonce track. And here you can see we've turned their body into a skeleton. They've created the movement.

And then on the right we have a deep fake body of Sister Sister, who we filmed actually in a cabaret venue that's shot during lockdown and turned their body into a deep fake that you could then feed any new movement in and recapture their body. This piece is actually showing at the moment, very exciting, it just opened yesterday at the Victorian Albert Museum in London. And we've got 21 performers performing to you in the space all simultaneously doing Beyonce acts and David Bowie acts and Shirley Bassie acts. So you can see a little bit more, here are some images of the kind of amalgamations, and again, this is like for me, a real queer move is to take these technologies and do something with them that they were never intended to do.

So the engineers and scientists that wrote the open source code, which I then took and hacked with, it was never intended to be used on more than one body at once. But I wondered what would happen if I fed in 21 different bodies into the system. And you get parts of all the different drag performers and this body, I dunno, disintegrates into the floor. This is for me, a wonderful moment because this is like a queer failure. Talking of Jack Halberstam, “The Queer Art of Failure”, which was quite influential for me.

When these tools break down and finding the poetry in when they really fail and fail on queer bodies. And actually for me, this is an example of AI bias because effectively the data sets that we created of each of our drag performers never dropped into the splits, partly because a lot of them weren't capable of doing the splits in the first place. But what's brilliant is when Kara's movement drops into the splits. The AI does not know what to do. So this is Lilly Snatch Dragon and Sister Sister and Oedipussi Rex. And they literally disintegrate into the floor in a kind of puddle of textiles.

And for me, this is talking of AI bias because it was never in our original data set. So the latent space, the boundary of everything that had learned from our drag performer never saw one of those performers doing that particular movement. So when we fed in a new movement, it didn't know what to do and it glitched in a similar way to a facial recognition system that doesn't understand how to identify trans people or people of color, might glitch and fail on certain bodies and identities.

So anyway, this is like a really fun way of maybe exemplifying some of those things. I'll just quickly mention our last project. So this is quite fun. We've been doing deep fake drag double act as well.

So working with my collaborator in London, called Me the Drag Queen, and getting her next to her deep fake doppelganger clone and they performed musical theater songs together. So they did one which was, anything you can do, I can do better, which for me is a satire on the idea that AI is going to take over from human artists. And maybe proving once and for all that AI is never going to take over from a real human drag queen.

But I think all of this is thinking about utopia. It's thinking about how can we empower ourselves as the queer community to think about how, I mean, I not had a wild idea recently, which is what if only queers and people from marginalized communities were allowed to build these systems and train these systems in the first place? Wouldn't that maybe serve society a lot better? Because, it would serve the rest of society, but it would also serve us first and foremost. And I think maybe that's quite needed to shift that from the sort of very normative identities, building these systems into something very queer. So I think, yeah, for me a lot of this project is about a deep fake drag utopia.

I think I'll finish. Thank you very much, Jake, for presenting the "Deep Fake Queer Act". Which show mato your artistry as well as the performers you work with and the community you work with. Involving communities are always a powerful tool or disruption on activities because you reappropriate a space that otherwise was either negated or neglected. You mentioned two things that bring me then to the question I have too for Sara, one is certainty and the cows that emerges when a data set is fed with something it does not recognize. So there's a glitch in the certainty, uncertainty.

And then there is reappropriation we will talk about, it may be more later, also in relationship to the legacy of media art. But Sarah, actually in our exchange before the panel, you shared some reflections about the tension between, in your words, AI puts forward the notion of certainty, bit certainty of place, identity, desire, which you take to be colonial at heart and antiqueering these effects as AI produces positivistic knowledge on the body and the self. So subversion and co-optation become then strategies of queering beyond acceptance and refusals.

Can you please tell us more about how you worked towards your ideas, especially drawing from your publication "Queere KI" it's in German from Transcript Verlag and which is open access, which is now reworked in an upcoming publication, "Queer Reflections on AI. Uncertain Intelligence" published by Routledge. Yeah. Thank you so much, Margherita, and to all of you of course. Thank you for guiding us through this exciting conversation.

I'm really happy to be here. The two edited volumes that you mentioned on queer AI in general came out of a conference that I organized together with two collaborators at TU Dresden with Michael Klipphahn-Karge and Ann-Katrin Koster, and we were really frustrated with, I think many of the things that Jake has already mentioned. This idea that technology was this deterministic framework that was apart from culture, that wasn't really, that was impacting us without the possibility of completely negating the status quo, which is actually the lived reality that people adapt, technologies, subvert them, use them in ways that they're not intended, and so on and so forth. And yeah. So I was personally really inspired by the work of Kara Keeling, who has this notion of Queer OS, so queer operating system. And she basically says, if technology is something that is working with the body, impacting on the body and changing the body, or actually for shaping our ideas of identity of gender and sexuality and so on and so forth.

Then there must be a way of developing something, as you said in your introduction developing an operating system that is queer, right? And that not only caters to queer identities, but actually is in itself queer. So produces queerness in a way. And I guess one of the things that I really found fascinating within the AI discourse and what made it so adaptable to that is that AI functions more than anything as an imaginary.

And what I found was that there were a lot of people that were actually claiming, that to make AI better, we have to have better data, we have to have better, we have to have more certain ideas of what AI is. We have to know what exactly the technology does and we have to be very clear on what we are describing when we talk about AI and I was actually of quite the contrary opinion because I thought that precisely the openness and the spectacular or the kind of this imaginary that opens itself up within AI, AI consciousness, this ulterior being, this other that is present and impacting our every day. I thought exactly, this is something that is actually really generative in a sense because it really opens itself up to many reworkings and subversions and also dis-identifications.

If I were to speak with José Esteban Muñoz, who's also been very influential to precisely that notion of queer AI that I have, because I think like this notion of dis-identification, which I also see by the way in Jakes work because it's so camp and over the top and there's this kind of this recognition of, queer bodies are not supposed to exist in that way. And at the same time, drag has always been this over the top representation of, the binary gender norm. And then at the same time, and I guess we'll get to that later, so I was really looking at AI as this imaginary and as this almost, yeah, almost religious belief in...

If we look at the hegemonic view in this long-termism and super intelligence and AI consciousness and so on and so forth. So I think to me the AI in itself, the way it works technologically and the way we understand it as an entire system of labor, of knowledges that are produced, of data that is annotated and data that is selected to be annotated and curated and so on and so forth. I think this multiplicity and this I don't know, pluriverse of bodies that are already coming together to produce this supposed vision of unity or this idea of something that is pristine and that is factual and that is just this one thing, I think that is in itself already very something that I just wanted to break open and in a sense point towards how many, conflicting and multiplicitous and generative different bodies are already participating within these technological systems. And of course that doesn't mean to gloss over the fact that there's a lot of hierarchies and there's a lot of inequalities within that multiplicity.

But I think it was to me it was already like thinking of an AI that has just this one thing. I think, yeah, it just didn't make sense to me. To me it was very pluriversal to maybe use that word.

It was evoking a lot of different fantasies at the same time. And I think in particular the way that we look at other cultural iterations, so especially the arts are very good in working with that. So that was why I was really interested in that as well. And maybe I'll stop here for now. Thank you so much, Sara. I thank you.

You just provided us with so wonderfully inspiring material and food for thought, that we can come back to it when we have the conversation around. But there is something that you mentioned regarding co-opting that also made me think about the word by Janne because the legacy of media art, which I mentioned earlier, media art is a broader field, is not only a critical approach through theory. So not only criticizing idea, but actually making, opening up the black box of technique of technology, opening up the devices, hacking them, hacking computer, hacking motherboards, making them do things differently and other things, make them do art things. Understanding how things are done by whom, what they do, and to whom. So this approach has been adopted by artists to borrow your own words, cooptation of that technology and the power dynamics and hierarchies that technology embodies, especially when it comes from big corporations.

So in your project: "The House of Monstress Intelligenzia", Janne, you bring together practitioner with diverse profiles because this resonates with my own practice as a transdisciplinary artist. And I'm interested to know more in detail how you worked on coding partly with devices, but not only for the narrative, but also a critical tool to subvert paradigms and normativities and to coopt a rare appropriate technological space. Thank you so much Margherita, and also thank you, Sara and Jake, for your presentations.

I will draw on everything you said and just add on. I think it's very it was a very good start for me, a good platform. So maybe I will share a little presentation.

I try to skip through it a little bit. Yeah, so I wanna talk about this project we've just mentioned, "The House of Monstress Intelligenzia", it's a project that I took together with media and performing artists allapopp. And we started it in 2021 actually while we were doing the project, we collaborated with two other artists. One is called Luna Nane, Luna is a creative coder and visual artist, and Portrait XO, which is an AI musician and also visual artist. See them again.

Maybe how it started, and it was similar, I think, to what Jack was already telling about. This is a project, because I don't know, if some of you remember, but in 2021 there was this website called "This person does not exist" and it's this StyleGAN website and every time you refreshed your browser, there were like new faces, like super photo realistic faces created. And I was super amazed by this and spent like hours on it sometimes.

And then at one point I just noticed that at the periphery of these realistic faces, there were like these glitch buddies, like the kind of queer trend twins or like different kind of bodies. And I found it super interesting visually, but also as a form of okay what..., you know like of course knowing that an AI maybe doesn't want to tell you something, but taking it as maybe this is kind of a futuristic thought or the AI want to tell us about the bodies of the future.

So how about we take this AI glitch buddies as the starting point for our project so as Jake already said, so it's the glitch, like the so called malfunction in the AI. And I'm drawing on the term glitch also from Legacy Russell: "Glitch Feminism: A Manifesto", where she also sees the glitch as an entry point into a normative system right there. It's like this malfunction, there's like this break where you can enter and then maybe rethink or like shape things from new.

And we were questioning, okay, could these bodies help us to envision a queer feminist future? To speculate about it. And yeah, so that was a starting point. And then of course we had like different questions like what.

Is it actually possible to create a queer feminist AI? I'm coming from performing arts, but I've been working as a media artist since a few years. I am super used into this DIY myself into something. And when we started, we knew theoretically a lot because we've been working what's like the subject of machine intelligence and AI, but we never worked with it actually. We thought of the project, like to divide it into three aspects.

One for us was really like, okay, what is the actual theoretical, like the status quo, theoretical knowledge discourse also from queer perspective, from queer angle, but then also how do we get to hands-on work on this? How do we dive into AI and actually be able to manipulate it from, not just from a user's perspective but from insight. How I empower myself towards this platform capitalism is actually by learning things, is actually by going into the technology and trying to understand it. That doesn't mean that this has to be my artwork and this artwork still can be like speculative or as you were saying, like framing it, Sara, but for me, this understanding actually gives me the possibility to think about it differently also. I feel like I only have the possibility to do this because there is so much knowledge out there and like the form of sharing to empower each other that it's actually not so difficult. I think there is a big problem..., a certain kind of threshold

that people think, oh, AI, this is this black box. I will never be able to understand. But I think if you get like the basics, you understand how it works.

There are different machine learning models. There are different ways of how to generate images. It's not super hard. So this like to understand the cross set up. Maybe I can skip through this quickly, when we were thinking about how to actually create a queer body, and also knowing that certain kinds of machine learning technologies or techniques use classifications. Of course everything is somehow classified if it comes to AI and there is human labor behind it.

But then we were thinking, give the AI as much space as possible to go crazy on whatever and to create our own data set with it. So we decided on using unsupervised machine learning and we created our own data sets with selfies of us because this is like the data we had the most. We also worked with labeled data sets. So we combined our unlabeled data with labeled data sets and did experiments. For example, here I trained myself on a data set of ponies. There came all these kinds of bodies and these bodies were the base for us to think about bodies differently.

And to question like a normative kind of body that is the default everywhere. So the outcome was like a performance which was fully AI generated. We said we are only gonna use AI generated material. But we also, while we were on stage, we were like doing this.

We're working with Colabs a lot, so we were trying to generate material on stage. So actually, the audience could follow while in an aesthetic environment, could follow how these things are actually done without being like autodidactic. Like we didn't comment on this, for example, and then reusing the pictures that are actually, that were, for example, generated or the text as a material to use on stage. So this was our artistic approach. And yeah, as I mentioned already, for us it was super important to also have like people to talk to while we were developing the project, people that have been working with the subject of queer feminism.

I like Emily Martinez or Josephine Ho, who was moderating another panel. It's a great AI researcher and I think this is also super important within this whole topic of how to queer AI to share knowledge and to empower one another. With your methods, with your tools with just knowing, yeah, you can do this. It's easy.

I can help you explaining it because I feel like there is also, for a queer person, if you have all these tutorials just made by this white male, Silicon Valley dudes, it's very much it's different for you to watch this tutorial and be excited about it than another person that you feel connected to will explain it to you or share. Yeah. And maybe that's the last point. So what we were trying to do is also, and yeah, we can talk about discord also as a platform, which is problematic in a sense.

We opened up a discord server where we actually shared all the tools we used. All the our kind of tried to break down our own approach. Also give a channel to communicate if somebody created something and... And then we had also like an Instagram takeover, where we thought, where we tried to explain what we were doing. So it was we tried to put a lot of effort in this communicating or practice. Did that work the Instagram takeover? Yeah.

More, more or less. But as everybody who did did like a social media post already which is should be short but insightful. It takes a day, so it wasn't... Oh yeah.

It takes forever. So much work going into it and this is not sustainable. Definitely. Thank you Janne, for the insights in your project. It's really interesting to go into details how the creative ideas develop from one intuition then shapes up and becomes a full-fledged artwork. There has been certain recurrent topics or recurrent points, in your contributions that I would like to draw upon for us little bit of conversation together.

One was the tension between certainty and failure and potential that Sara mentioned, and these in different ways came up in your presentations. But then you all made me think of something I read by the scientist Jonathan Jeschke, he's someone working on ecology he's a researcher who I've been in conversation with and as a scientist – so a completely different field – looking at the tools that machine learning and artificial intelligence make available, he speaks of dark knowledge. So like in the ecological observation or in other fields, you collect a huge amount of data sets and then there is some knowledge there that is not easily accessible, it's dark, and with certain tools you can access it. And we found it in an elegant formulation that a little bit steps away from these mythology of the artificial intelligence and rather looks at pattern recognition in machine learning algorithms that is a little bit closer to the truth and less, let's say, ideologic because the same word, artificial intelligence is very much ideology, ideological because you attribute to a machine a feature that was used to define human behavior. So how, to what extent can a machine be intelligent, but most of all it doesn't make sense to say that a machine is intelligent and maybe can do other things.

And this in my perspective, says much about how the discourse about machine learning, feature extraction from dataset is colonized by a certain world view. And your artworks bring back this generative potential, creative potential in the latent space. It was also one of the points I had with Janne in the preparation preparative conversations.

So maybe you can say something about your views on what Artificial intelligence may be, and rather what can we do as practitioners when I say we, refer to the diverse field of practitioners to reimagining this kind of technology beyond this normative ideology. Maybe if you all still thinking I can start us off. So actually I don't have an answer, but I think like it really, like your question really troubles me in a sense that it really makes me think and it, or I have been thinking about it as well because I think on the one hand, of course works like Jakes and Jannes is fantastic and it's fantastic and really allowing us to rethink what technology does. There are very real consequences to technologies not working or excluding queer people and non-binary people, right? If we think about the trans drivers who were locked out of their Uber accounts and couldn't go to work, they couldn't own a living, just because facial recognition software doesn't recognize them. That's harsh and that shouldn't be the case. And at the same time, I think Jake already pointed to this, that we are always working within a problematic framework that actually, yeah, it requires fundamental change.

And to me, I think the fundamental change lies outside of the technology. But that doesn't mean that we throw our hands up in the air and and let Elon Musk do his thing, right? That's also not great. So yeah, I think that also really in, in a very low key sort of common quotidian way I think that every single glitch that we can actually produce, and that's also and I think that's what I meant with the uncertainty, I think that a lot of times I just enter like wrong, of course there's a luxury in doing that, but entering wrong information or just playing with the sense of uncertainty or with the sense of, being this fixed identity, not just for AI but in front of like tech companies in general, I think is something that if we have the luxury to do it is really something that, that is worth doing.

Yeah, and at the same time I think that really critically thinking about how these technologies can serve us and what Janne was saying about, demystifying that it's actually not that hard. It's not that difficult to play with AI or to do something with AI. I think is really good for some practitioners because there are actually people who can develop things that in a way that is useful to, or that can serve our communities, right. Yeah.

Janne, I'll pass on to you. The one thing that you said, which which was, that actually lies outside of AI, like the main problem. I think this is, I would also say this is the basic thought we all should have. What I still was thinking with using this technology, I really feel like sharing, like your own experience is also something that I think is super important. For example, if we talk about playing with these tools, I don't know if we take like the tools like Midjourney or, like all these image generative tools that are around now, Dall-E, Playground AI or Leonardo, and I still find it as super problematic, like if you're given a prompt, for example, like what will come up, that would be a normative result in a sense, like a normal person, for example, if you try it, will be on Midjourney, it's like a white male, middle-aged cis person or a young white female pretty person. And then it's just like I feel how to subvert this technologies.

It's like if you want to queer, you have to learn the queer art of prompting in a sense, because you have to prompt yourself around the norm. Yeah, it is shit that it's like this way, but I feel like, okay, maybe this also needs to be something that is made, that it, that could be an entry point in in how to turn this technology around. I don't know if it makes sense, but yeah, I think that's really interesting. And I think building on what you're both saying, actually, I've been thinking about disparate. There's maybe an inherent queerness to how a lot of these AI tools work.

And I think from what you're both describing, it feels like, this idea of a latent space, I think a really beautiful and useful concept in how we think about machine learning systems and actually not necessarily anthropomorphizing and thinking of it in terms of consciousness and intelligence, which I think can us into slightly problematic areas, which can be a bit of a distraction. But instead thinking about a spatial metaphor where everything that the machine has learned is in within this space and everything it hasn't learned is outside of that space for that particular model. And I think that's a really interesting idea, but These classify models are built to say with a certain percent of certainty this is one thing or another thing. And in a way it's yeah, how can we de-binarize that, what could that mean to de-binarize AI? And is that something that's possible to do and also programming uncertainty into it. So instead of it's 99% listing, it's can it be uncertain like what the thing is? And actually a lot of machine learning safety people were speaking about programming uncertainty as a way of fixing some of the issues around how these tools are working.

But yeah I do feel that there's something like intrinsically queer about moving through this latent space and actually reducing things to mathematical coordinates. So if you are not giving it the labels, if you're doing unsupervised learning, and it's just building up this mathematical space of potentiality effectively, of all the in-betweens, in this like multi-dimensional space where everything can be continuous and things aren't necessarily fixed and discreet, to me that feels like a really beautiful queer metaphor. And if we are not reading labels outta that either, we're not giving it the human classifications, then it can just drift in between this space.

And actually, I think I'm really interested in seeing when artists are playing with those concepts. There was one piece, a machine learning porn piece, that I did really early on, which was for removing pornography from Yahoo search engine, then reverse engineering, re-engineering it's made create porn But it never got told that these were penises and vaginas. So this is like a wonderful example in a way. When it moved through that latent space of what it had learned to look for to remove pornography from the internet, it creates images of intersex bodies, which I really love.

And that's an example to me of like, when you do reduce these things to mathematics and then you can move between them. The humans never said, these are penises, these are vaginas. So everything is between the yonic and the phallic and it's creating these in between images. But yeah, there's that kind of potentiality to go back to Muñoz, it feels like it's always there underlying how these technologies are working and being built. It's just then the humans that come along and put the labels back in. But not to confuse it with something that isn't human.

I think that's also been a mistake. I think this idea of saying that it's going to surpass human abilities, is stupid because it's built with human abilities and it's a reflection of us and our own data. I think Jaron Lanier said something really nice there.

It's like saying, it's like comparing ourselves with AI is like comparing ourselves with a car. It's like saying that a car can run faster than the human runner, but... What's the point of it? Yeah. They're just two different things. It's a tool. Yeah. Dear Panelists, since we are very slowly approaching towards the end of our conversation within the Queering AI panel I wanted to remind you of the question we planned that each one of you would ask to our fellow panelist. So if that's fine for you, we can indulge in this moment and namely Sara asks a question to Jake and then we move on from this Jake asks Janne and Janne asks Sara.

Yeah, sure. I guess we already touched upon it a little bit because Jake I think I, I really share this interest in the uncertainty and multiplicity things and And the actually utopian potentiality that is multiplicity and deep fakes, because I think it's in particular this deep fake is something that is really shaking up a very conservative understanding of truth in a sense, especially in an AI generated truth and really shaking up and redistributing what were the power knowledge divide. What do you think about, because you do touch upon this like non-consensual deep fake and the violence inherent to these types of appropriations and also, I guess sexual harassment discourse is more generally. Because from a non carceral feminist sort of position, I would say, the reality of the crime, whatever that might be, is not necessarily that which is the most, the best way to assess how a victim has been harmed and so on and so forth. So I guess, do you think that your work can really, can also be a commentary or in a way, do you think that it addresses these issues in a sense? Or what do you think about the relationship between those instances and your work? Maybe? You mean specifically with deep fakes? I guess the deep fake and the question of realness, the question of artificiality and authenticity and so on and so forth.

Do you think that they are in any way useful for addressing these or for building upon, I don't know, non-carceral ideas of, being together, working on consent, working on what it means to be sexual together or to be desirable to one another and let sexuality flow really, so to speak. Oh, that's a big question. Just random thoughts. I'm sorry. No, it's really interesting. Thank you. I really appreciate your take. I think absolutely though it's at the core of the project.

I dunno how visible it always is, but in a way I want it to be playful and accessible as well. And, actually not doing this in a super dry serious, there's quite a lot of, quite dystopic work made about this, which actually can be quite inaccessible. So for me, the hope is that we can throw this on at a drag venue.

We can throw this on at a queer rave, we can throw this on at a museum and, it can be accessible to people who are interested in drag and AI, but actually AI isn't the easiest thing to enter into. And and fine arts as well. So I think, crosses quite a few of those bridge and throughout the whole process, this idea of consent is, pretty key for us.

And it's still evolving this project. So all of my collaborators, it's a constant discussion with them and making sure that they're aware of, how we're going to use the data going forwards and whether they can withdraw it to any point. We wanted to think about this, it's almost like research ethics.

How can we create an environment that's safe for everybody? And, we didn't want anyone coming into a space and waving their arms and controlling the body a, a queer body in a space because to us that felt really problematic. And actually we keeping it within just the community that can reanimate each other's bodies, I think, yeah, that was a key aspect for us. And yeah, no, I think I'm also a radical faeries I'm trying to work out, whether that kind of radical queer playfulness can still always exist.

And know, I think of some of my earlier works, it was harder to bring that. But actually then when I got into drags, so my boyfriend does his PhD in drag performance and used to run queer venues and drag venues. And I think for us, actually, that was beautiful coming together moments where drag in a way is a perfect example or exaggeration, yeah, gender nonconformity and really pushing that against the biases which are being talked about in these systems to break them down and subvert them.

So yeah, we had a lot of fun with that. I dunno if that really answers any of your questions, but No, that's great. Thank you so much. Yeah, I think this question, Sara, brings up a fundamental element in the use of data that is consent.

And that has also had the backlash of artists whose artwork, whose visual artwork had been used for fitting it into the data set. Useful training, most common algorithms behind the so called AI art wave, although it must be reminded that we are mostly here talking about visual arts using AI. And whereas we would need probably a different panel and a few more hours to expand the conversation to machine learning does go beyond the use of images, the use of facial recognition and art goes beyond the using portraiture as an expression, as a tool of expression.

But perhaps Jake, maybe now it's your time to ask Janne a question about her work or her ideas. So I guess for me, I'm gonna ask a super basic and simple question. I just want to know more about your journey in getting into artificial intelligence really, and how you came to it from your backgrounds as an artist. I'm quite intrigued to hear. Yeah. Happy to answer.

Yeah, maybe also it draws on the question that you had before, Margherita, because I'm a good example for how to make technology in general maybe, but also particularly AI more accessible for oneself and empower oneself in actually learning it. Actually I come from performing arts, but I've always think media or technology always was a topic for me. I grew up with computer games and stuff, so I never was like scared of it.

It was always like an integral part of my life and my identity, I would say. I was a bit sick of not knowing, like talking about this and not knowing how it actually works. And then I started, to play a bit of with Arduinos and stuff and like to code like a little. And and then getting like really step by step trying to teach myself how things work. I mean with AI, it was always something that was also on my bucket list.

Oh, how does it actually really work? What are new.... and then there was like this open call and it was like, particularly for performing arts in AI, like this intersection. And then I usually take a project to learn myself a new skill or teach myself a new skill or to dive more into something. My research or as a researcher, I think like the main topic is the representation of bodies in the digital realm and how it reflects back on how analog bodies are perceived. And then, like that you have this feedback and so I, I think like working with this image generated bodies came also from my field of research and interest. As I said we had collaborators like Luna Nane, and Luna is a creative coder, so she gave us the platform to start with.

She explained how Google CoLab works. And then it's if you don't know how to code Python, like even we don't have to, but still to know how to read it and then just how to go through it. And then, and then you know how it works, right? You watch tutorials get better, you get new ideas, you do another thing. And yeah, and actually I keep on working with AI tools since then, so it's really, I'm super grateful that I made this start and encourage everybody to do it if they want.

That's odd. Yes. Thank you, Janne. But that's exactly why you obeyed your work in particular made me think of the legacy of media art because there is interesting in understanding how things work, how these machines, these devices work inside.

Unscrew them, open them up, and then okay, start dealing with the uncertainty you just created. But then this is a way of understanding how things work. Also I just want to add, I really believe in learning by doing also. Like even if you don't know yet, but then just start and then you understand so much better than if you would read like ten papers or watch ten tutorials.

Of course you can start with this, but I think like really it's good to go into, you can't break anything. I think this is always the big fear or no, I broke it and now it's, but yeah. And even if sometimes things get broken, you can also learn how to repair them, don't you? Yeah. And then there is also, for me, interesting in your work also the journey from being a performer or someone working in the performing arts, then having this interest about the body. And so this link is also something I can relate it. And just another stone I wanted to throw into our pond, and then maybe we'll have the chance in the future to talk about it, is how bodies are not only how they look like.

I also think that the current mainstream attention on facial recognition, bodily features flattens the bodily experience to appearance. Whereas especially within the queer communities, it is known that bodies are not necessarily how they look like. And this is another potential that hopefully can be explored in the future. But then maybe now Janne is your turn to ask Sara a question.

I already thought, Sara, I have so many questions because I feel of also the publication is so rich and there are so many things to ask you, but I maybe stay with this topic of representation and actually the images that are represented I've been thinking a lot because it already happened in the past, if for example, companies have like their image identification, and then you have this really diverse group of people, which now you wouldn't know if they are actually, if they are real people or like AI generated. And if the, that the how you, how a company, for example, gets perceived, maybe not necessarily is how it is structured. And so I wanted to hear if what's your take on, especially with like image, but also avatar generation, right when we go to deep fakes because there are like the glitchy deep fakes, but when the development goes into the direction that it's more and more flawless like that the images look more and more real, that everything looks more and more like non glitchy actually.

And I found this yeah, this really concerns me actually also for it because it's like the take actually taking queer bodies, creating queer bodies or creating marginalized bodies and using it for example, advertisement without actually paying any body. And yeah, I just want to hear your thoughts on that and where you think it could potentially go. Thanks, Janne, I think that's also a huge question because of course this is not only happening to queer bodies, I've done some work on click and crowd work and how, the representation of kind of happy click workers, happy data annotators, and there's a book by Philip Jones called "Work Without the Worker", where Phil Jones actually talks about how, there is a linkage between humanitarian aid data annotator.

So data annotators are basically being sourced from refugee camps and they're predominantly women and so on and so forth. This exploitation issue is something that is happening all across the border and not only to queer bodies. Of course queer bodies are... So maybe I would just again say these solutions are not within the technology and I don't know if they can be, and at the same time, of course, many of the things we've already touched upon are really valuable to this question because I think the question of who is allowed to create data or select data, curate data and so on and so forth, is really important. The consent that you have about your data, what you can do with your own data, but also the consent when people extracting your data is important. And then for the representation, I think I'm conflicted as well because on the one hand, of course big companies are never gonna be in the service.

Like they can put rainbows everywhere, all they want, and they can celebrate pride in whatever. And we all know that they won't change certain policies within or they won't stop businesses with authoritarian regimes that, that are antiqueer and so on and so forth. So I think this is just something that we need to know and acknowledge and be aware of. And for like individual representations, I would almost say, how nice, welcome.

Come, and why don't you experience a little bit of your own queerness, open yourself up to that. And I think I'm personally of the opinion that everybody is... that nobody's served with a gender binary. Even if masculinity is not this one thing and it's a multiplicity and femininity can be a multiplicity and take on a multiplicity of bodies and shapes and forms and acts and performances and practices.

On that side I would say it would be also great to have a little bit more generosity and... Yeah, let everybody be queer. I know that comes like with its own problematics within certain communities and certain communities have to be protected. But staying with Muñoz and the utopian I would say let's all hang out and be queer and groovy together and share a Coke or whatever it is that we think is a good queer practice.

Thank you, Sara. I think that closing with the idea of Utopia is a powerful, is a powerful projection. Of course. Utopia is a possibly more inclusive and supportive space where everyone can find space in tolerance and welcoming not as a disciplined, normative space of whatever kind. I wanted to ask if you have any further comments, because I really enjoy the conversation with you as a way of bringing the topic of machine learning as a possible tool to imaging identities.

And then of course, this brings up other questions that unfortunately we won't have the time to explore today together, but maybe there are other things or other ideas that came up during this conversation. And you have very urgently the need to express now. Janne.. I would be very happy also if people see this and want to reach out and people work on like similar topics and like really want to continue the discussion with you all. Like within this panel, or also anybody who's interested.

I'm super happy to be reached and to continue this conversation. Yeah. Yes, that's a good point.

Of course, on the website of this project, you will find the names and address of each one of the presenters in the panels. So I'm sure reaching out is always a much needed part of the game because artists and research need input and feedback from audience, peers, strangers, monsters, and also sometimes critique. So there will be platform for this, but Jake you also look like something was burning in your heart and wanted to say something. No, not particularly, but I think moving between an AI queer utopia, that's what we should all be aiming for.

Then, thank you so very much to each one of you. I truly hope we'll have occasion to talk again in the future in person or on other semi virtual platforms like today. I thank again, the intelligent.museum, ZKM and Deutsches Museum for having us here today. I thank you the listeners, random encounters that have been generated by this panel and goodbye.

Thank you so much. Thank you.

2023-06-30

Show video