Talk: Georgia Ward Dyer, Artist and Researcher

Talk: Georgia Ward Dyer, Artist and Researcher

Show Video

Hi. Everyone thank you so much for coming along today yes. My name is Georgia, wood Dyer I. Am. Only, very. Nominally. A violinist. I'm. An artist and researcher. And. This. Is what I'm going to be talking to you about today why artists, should work with AI but I thought I'd give a little bit of an introduction about myself before. That. So. Although I suppose I would call myself a philosopher, but maybe serve lapsed philosopher, because although I have a degree in philosophy. I found. That. Exploring. Them through philosophy, as, an academic discipline. Was. I found it somewhat limiting, and so this is how I came to you art practice. And. But. I'm still interested in the same ideas that motivated, me as a philosopher, so, some, ideas about the relationship, between word, and image for, instance I'm. Interested, in, meaning. What, meaning is, how. Its constructed. Where. It's located which. In, itself is a bit of a tricky word about locating meaning. How. Its, dissolves. The sort of boundary between meaning, and non meaning, and. Because I'm interested in those things. In. My art practice and then also interested, in things like, taxonomy. And, categorization. And. Archives, but, also, folklore. Magic, and myths so all these disparate things that's, what are patches as a like aren't they but. I've often worked with scientists and technologists. And. In, my, creative practice I make things but also write things I, am. Going to talk a tiny bit about some, of my own work at the end. But. I just wanted to give a bit of a background about what my sort of interests and motivations are. But. For the past couple of years I've also been working at. Nesta. The innovation foundation I work, in the sort of think-tank, bitch as a researcher, and. I work there on emerging. Technologies in the arts as, well as some futures work. And. And. I do that through some, sort of more speculative, research projects, and workshops, but, so, research. And. Policy, work for, government. Public. Bodies like the Arts Council and so on. So. Just before I start, a, quick. Note about diversity. Firstly. When I talk about artists, so why should artists, make work with AI I mean, this in the broadest sense of. One. Who works through. Creative, practice, I don't want to get caught up in, labels. About specific. Practitioners, so all welcome, and. Secondly. Acknowledging. My own privileges, as, white. You know highly educated from you know privileged. Socioeconomic. Background I don't, think it's possible to overstate how. Important. It is to have a diversity, of who these artists, are when I say artists should be working with a I have, a narrow. View of what those artists should look like and. I don't just mean diversity. In. The, sort of race gender. Identity. Sexual orientation way, but in the widest sense of, diversity. Of viewpoints and, disciplines, lived, experience, and so on, and. So with this in mind and. Knowing that I, hope, that the, audience that, you're making up today is a mix of different, backgrounds. And some. Of you hope I mean, sure lots of you from you al but also perhaps some members of the general public to you I've, worked hard to make this talk as. Accessible, and, in plain, English, and. And. So I'll try to stick to that and. Of, course I'll, be happy to come back to anything that isn't, clear we've, got the Q&A later here we've got a break later as well when you're very welcome to come and chat me so. Today, I'm, going to talk about why. Artists, should make work with AI. The. Short answer to this is because it's. A valuable thing to do. But. What is the value in here is it valuable for. So. I'm going to unpack this and, I'm, going, to roughly, divide my, time talking, a, bit. Confusing that about the time I start a bit late didn't we do I have until. -. We. Started yeah okay, trying stitch that. Yeah. So who is it who is it valuable for where's the value, so. Dividing. It into you I'll be, looking at it on, one hand as.

Valuable. To everyone - to, the, general public - AI. Researchers. To. The. Sort of wider picture and then, in the second half I'd be talking about and the, value to artists, themselves of. Working with AI so another way to think about this it's, sort of my. First question, why artists should work with AI and the first part I'll look at why artists. Should work with AI why it's why, the fact that it's artists, doing this work is important, and then, I'll be looking at why. Artists. Should work with AI so. Why for artists that's important to work for that hope. That's clear between the two there I was going to pretend this as a gif and then, I thought you'd all see, through that and know, that I was the sort of Mechanical, Turk behind. Her. So. First. We're looking at this one aren't we white artists should work with AI. Why. Is that important. Um so. There are a few different, reasons and. This reads when I wrote this down I thought it reads a bit like a manifesto, actually, I don't have a problem with that manifestos. Are very important, in art, and, then across the history of art so. Reading. This out so the world needs artists. To challenge AI I should. Put the emphasis right the world needs artists. To challenge AI the. World needs artists. To make sense of AI, artists. A critical. Artists. A good, at complexity. Artists. Are ahead of the curve so. We're. Going to work through these so. Obviously artists. Have, always been early, adopters of, or. Experimenters. With new technologies. But. They're not sort. Of obedient, users. To. Use the parlance of tech companies of, users, of technologies, and artists. Working with tools, which. I'm going, to slightly used interchangeably. With the word technology, is. A bit like. Anyone. Who's tried, doing user. Testing with children, for. Instance. Or. Used testing with not just your peers from your course for instance idea, is that analogy mildly, every, think about tropes around children and creativity as well. You. Know about the first of every child is an artist Picasso. Stuff an, artist, will use the tool but they won't use it how they're supposed to and they'll break it and they'll, break it a hundred different ways and they'll, repurpose, it and they'll adapt it for some strange, reason that, they want, and. Artists, have obviously always done this for technologies. And. So again as I was saying thinking about technologies, as tools I was thinking about some. Example. Of the. Way that artists use tools or use technologies, to give and I thought a really simple one would. Be about paint. Which. Has been around for, ages and, it's, quite a simple, technology. To, get our head around. Been. Around yes since I looked this up in it it's I think twelve I think it's twenty seven thousand years ago so it's nice figurative, cave paintings, from twenty seven thousand years ago. But. Really it's about how those how at all as simple, as paint gets changed and transmitted, over time and. So here's another this.

Is Not a sort of linear progression through, time as you may have guessed his, Jackson Pollock you. Know not, the first person to use paint in this way but certainly the most famous, you. Know so it's. Challenging the. Sort. Of mechanism. Of the tool itself that sounds very overwrought, to say about paint that as a mechanism of use but how you use paint, can be challenged. Here's. Another example just, to work against, just, having such a famous, man Jackson Pollock his Daphne, Orem who use paint to, to. Paint on film. Strips, which were then played through a special machine that she created to, create. Electronic, sound so sort of electronic, music machine, using. Paint I mean. These are slightly some idiosyncratic. Examples. I've picked but the idea is to show that you, know even something as simple as paint has. A long, long history and how it's been used and, and. Of course how you know how this that, Daphne Orem wonders from 1957. Of course there are even more exciting, examples. Of how painter sees now and, it doesn't follow any kind of linear trajectory it's, just led through all these different revolutions. And artists. Are doing this, with a I already, but it's. Important, that more artists, do this for they I to and, of. Course we'll have some examples of that today but you. Know where are we on this timeline with, a I as, a tool rather than paint sorry. If. This is the timeline for paint. Where. Are we for those artists using a i i'd, i wouldn't want to say we're at the beginning because that would be rude, to the artists who are here today who've, got amazing work that they've done with AI but. But, no doubt i mean one thing for sure that, we can know is that however artists, see using AI today it'll be very different in twenty seven thousand years time so i just think the idea of being at the beginning, of this moment, of exciting. Experimentation. And and, especially with, a tool, like AI where the pace of change is. So much faster than the pace of change with paint that took you know thousands, of years think someone, thought to put it in a tube. So, you, know I'm not the only I I know I'm biased but not the only one who thinks that artists, have an important contribution to make in terms of the developments, of technologies.

Douglas. Eric who's who's the project lead for a Google project, called magenta and whose. Aim is all about you know. Experimenting. With how machine learning can contribute to creating, art, music, has. Talked about this and has talked about drawn, an analogy, with the camera, and has, said has. Talked about how. Artists. Who are working with AI now you, know that's really, the future of what, it will look like to. Making the tool as the first stack. So. Yeah so there's going to be a trajectory, and it's gonna when we're really only the beginning and. And. It's and all, the way through I think it's artists, who are going to, mechanistically. Challenge, those technologies, so challenge, how they work what, they can be used for how. They use. But, there's another way in which artists. Challenge technologies, not, just how they're used or what they do but also the, impact that they have and this is just, as important. May. And maybe has sort of reached the, public consciousness. A bit more and so, I'm just going to give a couple of examples of this. So. Artists. Mentioned. Here's there's Caroline cinders, who has a project called feminist data set, and. Joy balamani who, has created. This poem. Performance. AI entire, women I've put these up with the names everything because to encourage you should look them up and get, to know them I couldn't do them justice in the time I've got today, but. These are both both. These projects, sort. Of. Challenge. This idea, of algorithmic, bias that probably a lot of you have come across the idea of this that. Because. Of the data that's used to train some algorithmic, algorithms. And the, advent of AI as a sort of integral, part of how societies, run. Curd. Or does, already, amplify, existing inequalities. And. Here's another example just because. It's very topical it's just from a few weeks ago and Kate, Crawford from, the AI now Institute, and artist Trevor Paglen did, this image net roulette where. You could upload a selfie of yourself and, then be labeled by an AI image recognition. Tool. Trained, on image net which is a very, significant. And, sort of seminal data set images. And. That. The results were sort of you know people were like oh this is really fun we know what does the AI think I am but.

But, It's, quickly, showed, up as, sort of problematic size about I'm. Creating, from enough from this article now a, dark-skinned. Man was labeled as wrongdoer, offender, an asian woman as a jihadist. So. I think again this, is about how artists. As I was saying before like with the paint yes they they sort, of challenged the mechanism, of how it all works but, also they critically, challenged, what impact, it has to, and so, they, work to make sense of of what, AI is for, the public this is quite an important communication role. Bringing. Something, even if it was in specific communities quite well-known about at the idea of algorithmic biases, not didn't. Just appear at, two weeks ago but, certainly, bring it to to a wider public is really important. But. So I just wanted to take a moment of pause and think, hang on. What. I have is just with my training sets what is AI, I, was just at a workshop the other day and it started, with this and let, me tell you we spent a large percentage, of the day, in. Agony's, they were trying trying. To work out how to define AI and, this was a roomful of you know engineers. Computer, scientists. And policy, experts people who knew the fields. You. Know that that the data set that I was just talking about that was used in Kate Crawford and trigger patterns piece this one, I'm Imogen, er that was not assembled, by an AI that was assembled by, people. So. Even. If we were tried going, to try and say okay well forget all, these confusing, bits of training sets that's just try and give, a purely technical definition, of what. A is and this is what happened the other day I miss workshop and it was very funny to watch um you to have people, disagree completely on what that would be and it goes there oh it's, just statistics, crowd. There's. The, sort of oh it's just about it's. An engineering problem. You, know even if you said okay we'll look can we agree it's a computer program now what does that mean again I said an algorithm, which. One, I mean what do we mean by algorithmic, lines of computer code is, it a neural, network sort, of catchphrase that people have got to know now or is that the weights of, a network I mean it's it's a complete nightmare to. Try and just try and unravel that and I'm not suggesting we do but I'm trying to make a point here about it being complicated, so.

Even. If we were to try and give a technical definition, of what a is we, couldn't. But. We wouldn't want to do that anyway because it would be a complete, fallacy, as. We've seen AI, is is, culturally, embedded and, at least the, following ways that I've already gone through know. If. If for instance we think about AI as. Just. I think how maybe most, people might think of it that the, thing which powers, I don't, know, recommendation. Algorithms. Like Netflix, or, transactions. On Amazon, or assistance. Like Syrian Alexa. You know already that's a hugely culturally, embedded. Especially. Is the success of those depends, on, the. Sort of relationship success having a successful relationship with, the human interactions, in order to improve, as services. So. It does not a clean divide between, the technology and the culture, anyway. On. Top of this this. Wasn't complicated enough. We, also have, like. The there's. A project at the Leverhulme, center. For the future of intelligence, called AI. Narratives, which is a project about, I. Suppose. Shedding light on what the public perceptions, around AI are and it's what you'd expect as Terminator, figures and sci-fi, but. That that question is not just important, sort of anthropologically. To think about but. Also is a practical, one when it comes to policy. AI. A sort of well known as a field so field of research. This. This danger, of overreaching. On height. Has. A has, a policy, impact, if. If it doesn't deliver and and if you look at the sort of you can find graphs, of AI research progress, and sort of peaks, and troughs of AI winters, and so. So, as a total, mass basically. It's, what I'm getting at and now obviously I'd, want to acknowledge that all, technologies. And I'm sure you're, all that you al you all. Know this already all technologies, as shaped and defined by culture and society, and by their use over time that's. Not a new thought, or. One specific to AI but I think it's particularly. Significant, here as I, said because of the particular nature of how AI and, particularly. Machine, learning, which. Is what, mostly. People mean, when they say AI these days anyway. How. That works and secondly. How potentially pervasive. And diffuse. Applications. Will, be you. Know AI, is this a paragon, example, of complexity. Around. Water. Technology, is what. Its impact. Is how, it will be shaped by us. But. Art as, a creative, practice. As. A medium, as the sphere of influence can handle that kind of complexity. If, we think back to our paint example, you, know there there, are different analogies. I could use but you. Know art practice. And our, practitioners, Rafael you know will flood that will flood the space will fill the vacuum with. All possible kinds, of challenges and experiments, and both, in at the individual, level of you, know an, individual's, art practice. As. I was saying earlier about why I came to art rather than the philosophy, it's an approach that allows for a contradiction, and allows for multi valence allows. For ambiguity, and complexity. But. Also as, a discipline, and at the institutional, level you. Know where, we are today, it, allows for that kind of complexity, to. So. So that's you know that's one of the things in fact particularly which digital technologies, a good enabling, but. Yeah so hopefully, I was showing that Oh a, is, complex but, the art is very good at handling complexity.

Artists. Suck so. The last one on that list was about artists being ahead of the curve and, so. I don't stop the garden because I'm we're gonna hear examples, of artists working with AI but. If we remember about how that pace of technological change is, is not that some nice straight line I. Just. Wanted to give an example so, here's a slide showing the performance. Of Ganz between. 2014, and 2018 so, in. The, plain English version so an. AI tool which generates, images of faces, on. The left in, the black. And white is how well it did at generating realistic, looking faces in 2014, and on the right how elected in 2018, so that's, an incredibly short time for huge, improvement. And. There's an example from a. Project. That Anna and I did together actually about this about, why, artists. Are well placed to, both. Keep up with the pace of change but also to document it, Anna. And I worked on a project a couple, of years ago in. Which we used voice. Track as, as one of many steps we used voice recognition software. To as. A step in creating a contemporary. Retelling, of very, classic, fairy tales this one beating the Beast and even. Involved. Us reading, as, sort of the usual. The classic version into, a voice recognition algorithm. For it to so transcribe, what we were saying even, a few months on from when we first did that project. The results were completely different because the API that had been released that we used, had. Improved you know that its performance and improved dramatically and we. Were intentionally, exploiting. The, ways in which it got stuff wrong and misheard us one. Because of the aesthetic, sort of result. That we wanted but also because. It gave us an insight into you. Know for example what, this tool was trained on it turned out this one had lots of references to sort of anime, characters, and they so we thought okay well it's clearly trained on closed captioning. And. Just a few months later none of that insight, or aesthetic, could have been available to us because it, was quite, seamless. So. Artists, the ones who are grabbing those moments documenting. Them maybe. The only public documentation, of those things so, that's enough about how great artists, are and I've got my eye on the time as well hopefully I can squeeze a bit more time in and, we've, done enough talking about how brilliant artists now, let's talk about now, I'm addressing, the artists, and saying why you should work with AI. Something. To talk, about two aspects, of this firstly a. Sort. Of race, through, with an eye on the time again and the, role of a are in the creative process and, then, secondly, this. An idea of AI as, mirror. So. The. Role of AI in the creative process I'm just going to really quickly. Move. A step away from the, the idea of AI. As, creator, AI. As original, creator I know that that dominates the headlines and, if you are interested in that there's plenty out there to sink your teeth into but, I think it's a distraction although, there's really interesting research on this about. Computational, creativity.

There's The Goldsmith's, computational, creativity, work, there's. That's work before that as Margaret Bowden on there, just an that I think. It's a distraction from what's really exciting at the moment which is AI, not. A sense of completing creator, that's going to replace artists. But instead as a mediator, that, makes the creative process more seamless and intuitive, and. I'm going to give some examples. So. Here's, an example. From. Rebecca, fabric who I just learned. I think he's joining CCI. Yes. Great. Lucky. You and with, all I can say and so, she. Created this tool called wek inator which. Again. I've put the details out so if you're interested to him or go, off and look it up it's. Brilliant but I'm essentially allows you to set. What, inputs, you want to give and. Then what outputs you want to get using, your laptop. With. Barely any with no coding, acknowledge needed, and it's incredibly quick and intuitive to learn and very fun, she's. Made some other great tools as well but. It's it's it's fun it's quick its accessible, it's flexible, and. The way that the way that you use, it in your, creative process that's, it it's, not like oh no now I've got to sit down and spend hours. Trying to install something, on my laptop. So that I can code with it, another. Example of a tool which I thought I should mention because there's been some noise about it in. The sort of creative AI community. Is runway, ml. So. This is a, project. Which is a similar, intention, to try and let. People use, experiment. With AI. Tools, in quite a straightforward way without coding knowledge needed and I should. Mention though I think, it's important to talk about this that runway. Ml is not free I mean, it's free to download but then you, have used sort of spend as a credit system with. Computation. Time that you use and, this is really important to note about who creates and owns these. Tools if they going to be so important in the creation process, I. Think. It, could be this watershed, moment, like, with Adobe Creative Suite where suddenly everyone can use Photoshop, and everyone can be a graphic designer but, there is a tension with, things that are proprietary. And what and what artists. Want to do with it which is you know experiment. And. I think it's, not just even about a software question, because. With, machine, learning particularly. It's also becomes a hardware question even, if you, know types. Of flow is free to use. Then, you still need GPUs, to run it so um it. Gets quite naughty and I think we should be focusing, on the importance of creating conditions in which in which, using these is accessible, to everyone and, obviously when I say accessible, it's not just about the, financial, access and. We'll the court you know of the resource access which what the hardware or whatever it's, also about skills. And. But, I don't really need to say much about that because here we are at CCRI, so I think we're good on that front I'll, say. But. Yeah, it's not just about there being toriel's online with skills it's also about having community, spaces to, to. That. Support people to learn so. They're two important things to mention. So. There is this possible characterization, of the role of AI as as. One which supports, and helps the creative process but there's also another one ai, not as creator, not even as mediator but as disruptor. Annoying. Word to use but. This. Is the idea that actually AI could help take creative practice, in unexpected, directions, so. Obviously. The history of using rules and, algorithms, in, the creative process is a long one there's a long history of computing, and computer, Arts, and, even before even without computers, just you, know people artists using rules, in, that process like the realists, or Aleppo, all use this it's not a new idea I'm, such, a beeper AI on this little continuum, of a fad or do we think of it as doing. Something, a bit more maybe there's a bit different or even more exciting.

And. I'm, running out of time so I'll get this very quick example, of. Generative, design, where, which. Is sort of the idea of turning the design process. On its head and not, saying oh we need to sign a chair okay, well I will have to have four legs and okay, well here's what has used to look like how can you change this a bit and improve. It instead, it's okay. I need to design a chair I'm going to describe what the what the design problem is it's something, that can support the weight of a person and, that's, got a sort of flattish, bitch bottom shaped bit on the top and then and then you hand that to the AI and, then it explores all the possible solutions and comes back with something really wacky and weird that a person, probably wouldn't have thought of by. Themselves. So. I think, yeah. The sort of idea of AI not not only as. Potentially, having a big impact in, improving. Or making, more intuitive and existing creative process but also, interacting. With the, creative process, more, like a collaborator, would rather, than a sort of obedient, slave. Now. Very finely. Then and AI. As a mirror I'll probably have to skip a few bits here hopefully, I've done enough to. Convey, to you the idea that, AI. Is very tightly, intertwined, with, ourselves us. As individuals, but. Also us and specifically. As a society. And. I'm, going to talk a little bit more in practice to you so one. Example from my practice about, how intertwined, AIS, with ourselves and, then how we can how we can exploit. This to make work that is, a reflection, of ourselves as, their, poetry collection that I created and, asking. People to generate text, using, the predictive software on their phones. So. You know you just sort of tapped the letters, the words that come up and then you can generate reams of text, and. And obviously what came back yes of course that gave insights, about the technology, itself. But. But, it really gave very intimate, insights about and, I hope yeah, you couldn't risk big enough to read isn't that gave really intimate, insights about the people who contributed, text what, their pet names were for that loved ones how, much they hated their boss and, where they were going around holiday, all of this really intimate, information so. You. Know already if you're if you're a part of the modern world there's. A better sort of algorithmic versions, of you fractured. Across all, the services that you use and there are other artists, who've looked at this of course I would I would name check Erika's score to use watching this and hoo-hoos adapt its at bat saying everyone's, producing, an image of themselves for an algorithmic. Is intentionally, or not so. Even, if you're not cheaply interested in making work about AI as a tool, to start. Thinking about our, identity. And how it's made up of our communication. Or you know our personalities. Is is really interesting. And. I'd. Really like to make that quite clear making, work with AI doesn't. Always have, to have the technology itself, as a subject, matter. And. I'm sure Anna will touch on that later. So. Lastly. In. My, own practice there's. Another way in which AI, is, a great tool in this way not, to talk about AI itself but. As a sort of reflection, back on a human way of doing things I wanted, to make another point if. I haven't ran the point home enough already about AI, being, a reflection of ourselves individually, in ourselves in society and culturally. Also. You know in terms of how some. AI. Tools were developed there's, a there's a direct, link you know between a. Scientist. Called Jana Kane who created a. A very. Important, kind of AI, tall. Convolutional. Well created. I'm giving him a lot of credit here but convolutional. Neural networks, has, talked about how. He's. Influenced. By the, structure. Of the human brain and in, terms of designing, the structure of these. Kinds of neural networks so. There's there's there's a literal. Link between. How. We are and how a are, some kinds of AI are and. I've been, interested in this in my practice to. Explore. Philosophical, ideas about, as I said earlier I'm interested in word and image. Through. Its really using AI so, I'll just have to do a super, quick race, on. The left hand side of these pairs are drawings, that I made that are sort of intentionally, abstract, but look a little bit like maybe letters, maybe. Glyphs and then, I sort of asked an AI what it sees in the image and then, and. Then I asked. The AI which. Specific. Bits of the image made it give that answer and so. On the right hand side what you see are the parts, of my drawing that, sort of yielded that specific, answer, isolated. And this one scorpion for instance. And. That's that was, obviously, interesting, to, sort of reveal, these things about AI but, also it, was about my. Interest in how, we construct. Meaning and how we look at images and when we see an image with a word, underneath, we, immediately.

Make Connections, and how we read meaning into images, in that way so, so essentially, if you want to make work about. Individuals. Or about. Culture. Or about anthropology, or about society, then, work, with AR and it will give you that, so. Finally, a very, quick. Well. That's my call to action but very quick and diversity, throwback, I should, say that there are many levels at which you, can work with AI and I think I've made that clear with some of my examples you don't even need to know how to code and, that, said and the, more you learn then. The, more knowledge you have and the more flexibility. And power you have and if you're going to break the tool you probably want to know as much as possible how, it works. So. Again we're in the right place for me to be saying that today. But. With. So many directions to go in and thinking back to my example with paint paint. Such a simple, substance. And yet, so many different ways to use it and and test, it and and create, new things with it and there, are so many different directions to going with with how you're, all going to get off and work with AI now, that. I think, the. Most important, thing to remember is criticality, and that, should be your guide in which way you develop or where you want to go to because. There's no good learning how to it's no good learning how to do any of these things as we, know the idea of the sort of purely technical perspective is, a fallacy anyway, that. That the critical perspective, has to be your guide, and. I will end there thank, you very much. You.

2020-08-26 17:10

Show Video

Other news