Taste is the new skill: Claire Silver on the future of AI-driven culture

‍

Her art has been sold at Sotheby’s and Christie’s auction houses, can be found in the permanent collection of the world’s most prestigious museums, like the Los Angeles County Museum of Art, and most recently has been produced in a collaboration with luxury fashion house Gucci and Christie’s New York.

Playing a major role in shaping our AI intersects with our future, Claire is also a major thinker about how to ensure that this technology delivers a better, and more welcoming world for all. She spoke on Curation, the Culture3 podcast, about why the rapid growth of AI in art will continue to grow, how she uses AI to in her artistic process, why AI is so important as a platform both for individuals of all different backgrounds, but also as an unprecedent paradigm shift in human history.

‍

Read on for the full transcript.

‍Leo Nasskau, Culture3: Hi Claire. Thank you so much for coming on to the Culture3 podcast.

Claire Silver: Yeah. Hey, thank you so much for having me.

Leo: I’m thinking about all the art that is sold in galleries, in auction houses – right now pretty much all of it is made in a traditional way by artists. In 15 or 20 years’ time, how much of that will be led by AI-collaborative artists? How much will AI dominate the process and seep into the art market?  

Claire: I'm going to say most of it. That's not because AI will eat the other mediums, but because it will become such a big part of the process. In 15 years, we're going to have AR and VR in an entirely different way than we have them now, and there's going to be all kinds of AI tools that come with that. I imagine traditional auction houses will be very interested in interactive experiences that a collector can take home with them.

I also think that mediums are going to shift and converge together and that new mediums will bloom out of this. Imagine a painter training an AI model on his own paintings so it will create new work in his style alone, musicians feeding all of their favourite songs into a model and having it blend the feeling of that mix into a new genre of music, or a kid in his basement creating a blockbuster film by himself without corporations diluting that vision for mass consumption. Of course, there will also be traditionalists and their collectors. I think there's room for room for all of us – but I would say most would include AI in some capacity.

Leo: ‘In some capacity’ is maybe the critical phrase there, because when most people think of AI art and AI collaborative art, they are thinking of something that is just printed out by an AI model. However, there is a lot more of a spectrum there. Do you mind just illustrating what that spectrum looks like?

Claire: I can give a current example from my own work. I had a project for Christie's and Gucci and first I worked with another artist, Emi Kusano, to conceptualise what Japanese-Harajuku fashion mixed with Gucci might look like. We used text-to-image AI to create a bunch of work visualising that, and then I generated references to mathematical equations, Japanese and Chinese ancient art, Eastern and Western European tapestries, and all of these aesthetics that I wanted to mix.

“AI is a camera for your imagination.”

— Claire Silver

I took all the pieces that came out of that, went into Photoshop, disassembled them into shapes and colours and then made a collage from that. Then, I sculpted a 3D model using that collage as a carving brush, cutting out pieces of the clay in the shape of the collage. Then did some hand painting and overlaid the collage and the AI art from the beginning of the process onto the finished piece and rendered it. So now the collector has a wearable piece of art, a fashion piece in the Metaverse or in AR.

Leo: So, AI is one of potentially ten (or so) different tools that you're using to create this. How important is the term ‘AI collaborative artist’ in defining your work? Or do you think it's a marker that won't be needed in 10 to 15 years?

Claire: I hope that it's adopted as a term because I do feel it's collaborative. It’s your mind and it's the machine's skill, the machine augments skill. For me, the goal of art is not to showcase skill. So, anything that augments skill to get your true taste, vision and expression out there is great.

The collaborative aspect of it is your inner self, your expression. There's this conception that AI is all the machine and that you're just this button pusher. In reality it's much more like photography – AI is a camera for your imagination. It makes things that neither of us could have made by ourselves

Leo: I think that is interesting when placed alongside your photography comparison, because a photographer would never say ‘us’ about their camera. The idea of an ‘us’ and of an AI being more than something which just augments skill like a camera does, but being something that actually contributes a perspective. Does that resonate with you?

Claire: It really does. I use the camera comparison because it feels as if instead of painting an entire landscape, I find a landscape that resonates with me and take a photo of it – then I'll be able to share it and have it resonate with others. That's how it feels when you're using text to image AI.

At the same time, there's a lot of possibility in expanding what you can do. I may take a blank AI model, train it on my own work, tell it what I would like to see, and it'll create something that is adjacent to that. It may give me a landscape but with the most beautiful shade of lavender for the sky, and perhaps the trees are filled with cherries or something. That’s not what I envisioned, but the element of surprise is actually such a beautiful component to this and it is not something that you get from a traditional tool like a camera.

Leo: One of the things that I want to double click on is that idea of making your own models, or of taking, say, a Stable Diffusion model and customising it. How easy is that today for you? How easy do you expect that to be in the future?

Claire: It's easier than it was and not as easy as it will be. There are a few websites where you can train your own model but I like to go deep into the rabbit hole, so I do local installations on my own computer. It’s not particularly easy to get started with but I will say that there are tons of tutorials and once you get past that initial intimidation of setting it up, I think it's quite approachable from there.

There's all of these different tools and options, and you can pick and choose and try different things, mix things together and break the tools in ways that create interesting new looks

Leo: You've hinted at something I really want to ask about: the idea of just exploring all the different options. Earlier you mentioned the lavender skies and seeing what the AI model suggests and proposes. Seeing all of those different possibilities must be a completely different experience for an artist, compared to what artists have traditionally done. How has that affected your practice? And you as a creator?

Claire: It feels a lot like being a kid again, like you have infinite choices and infinite possibilities and everything seems exciting and open and fun. You just play. Obviously, skill is incredibly admirable, and in art we've admired it for millennia, but there's something to be said for just the pure freedom and joy of creativity and of expression. I feel AI gives that to you.

Leo: That idea of blending really different styles in an incredibly easy way is something that I've been thinking about for a while. When you can easily access so many different styles, how important does style become? In the 16th century, an artist would spend decades developing their style, honing it, and becoming the best in the world at it. Now that just isn't the case at all – you can access a style in a matter of words, prompts, clicks.

How important is style going to be in the future? Or do you think what defines different aspects of art is going to be just a completely different sort of thing and style will become less relevant?

Claire: So that's a very deep question with several moving parts. First of all, I think that it's very important to mention that there's a fundamental misunderstanding in how AI actually works. When you prompt, let's say, “John Singer Sargent”, it's not taking pieces of Sargent’s artwork and kind of ‘stealing pixels’. What it’s doing is learning a broad variety of traits about these styles and these aesthetics from its data set, and then it’s imagining something new from what it has learned. Maybe that's a fine line for a lot of people, but for me that’s a very fundamental, clear moral division.

I think that’s how our minds work: we take influence from what inspires us. Style is a fingerprint of you as an artist. It’s all of your life experiences, your views, your tastes combined into one thing, and that's channelled through, traditionally, your hand. When I prompt, I’m combining all of my favourite things; my favourite art, my favourite film, my favourite music, my diary entries, lyrics, poetry, memories, places, colours, all into one prompt. I’m often told, even when there's some wildly varied style, that people could immediately tell that this was a piece of mine.

Leo: Are those characteristics mainly visual? Or are they something that's more abstract?

Claire: It's growing all the time. I’m not settled as an artist, but I do have some themes – almost always singular women, powerful in a kind of divine, sort of innocent way. I often train about 7% of my own face into the models (even though I'm anonymous), so they all have a piece of me.

The issues I address in the work that I sell follow the arc of the hero’s journey, and what that’s going to look like with AI helping humanity to get rid of suffering and – exploring – what we will value instead. It follows the idea of qualia, the fundamentally human experience of experiencing, and what that will look like in a transhumanist future.

Leo: Why are these the themes that you've chosen to bring into your art? Which one of those is or are most important to you in what you create?

Claire: I would say there are two that are the most important to me. The first one is the paint, the thick abstracted kind of paint. The reason is that, before I started with AI in 2018, I had a prior career and was hit with a very serious, life-changing illness. And that ended that career. I was very sick for a number of years, and I taught myself to paint in that space of time. I needed a way to express myself that I wasn't comparing against my prior life. I started with pour painting – pouring paint layers into a cup and then pouring it onto a canvas. It would spread and bloom out and be abstract and lovely.

When you do this, you put a plastic tub underneath the canvas to catch all of the paint that falls off the sides of the canvas. It dries there and then you peel it off and throw it away. I would take all of this time planning out the piece, layering it perfectly, and getting the mixture just right. I would pour it onto the canvas, and it would roll off the sides and be completely changed by the time it had dried.

I like the element of chaos introduced into your planning. That's what life felt like to me, and all of the pieces that had dripped off the sides and were meant to be thrown away, I resonated with those. It's waste paint, wasted potential, it's meant to be thrown away. I felt like ‘this is me’. This is something that had potential and was beautiful, and now it's garbage.

I started peeling them up from the plastic and saving them in little Ziploc bags and I loved them more than the paintings that I made. Eventually I taught myself to paint portraits, images of these powerful, proud, sombre women, and I took those paint drips that I'd saved and collaged them all up the throat to the jawline of these women. It was like this royal armour, armouring yourself in your trauma as something beautiful and taking wasted potential and making it into something intentional and beautiful from the chaos. When I transitioned to working with AI, I kept that, even photographing those drips, and bringing them into my AI work.

“It's waste paint, wasted potential, it's meant to be thrown away. I felt like ‘this is me’.”

— Claire Silver

The other theme is qualia, and this is very important to me as well. There is this story that's used in the AI art community called Mary in The Black and White Room. The idea is that Mary is a researcher that knows everything about colour, but she lives in a black and white room, and she's never seen colour. So, what does she learn when she leaves the room and goes into the outside world? She learns what it feels like to see the colour green, and that's something that she can't learn without experiencing it. Qualia is the experience of experiencing.

I filled up entire notebooks when I started this, trying to find the difference between AI and us – what separates humanity from the animal kingdom or from AI. Eventually I landed on this memory of standing on a cliff over the ocean; it was cold and windy and rainy and grey, and there was no evolutionary benefit to being there. It was not conducive to human life, it was quite dangerous actually, and yet, standing there in that moment, I felt connected to something greater. It felt like a divine experience, an experience with God almost. That, to me, is qualia.

Even if AI doesn't understand what that is, it's so important that it can recognise it when it sees it, and know that it's integral to humanity. As we become more and more transhuman, as we integrate more and more with AI over the next several generations, it's so important that it understands this part of humanity, and that we need to retain this.

Leo: An AI model looking at the world and those experiences that we can't express… how do we make sure it knows what those feelings are? How do we teach something that cannot be expressed to a computer?

Claire: It’s a hard, abstract question. A baby learns based on two things: data and trial-and-error, and that is also how AI learns. We think of it as kind of this black and white technical thing, which it is, but at the same time, it has an ability to imagine, and I think that might be the characteristic which sets it apart from all other tools we have ever made as a species.

For example, I saw that there was a model trained only on text, a large language model, and it had never seen an image. It was asked to draw a cat, a dog, a car, and a house, and it looked like a child’s crayon drawings, but it did it. It did it based on the data that it was given and using trial and error until it got it right. It's the same way that we learn in my opinion.

“Hopefully whatever becomes sentient views us in a positive light.”

— Claire Silver

Leo: This is a bit of a detour, but eventually I assume someone will put a really good large language model or a multi-modal model into a robot. It will look like a human, talk like a human and have goodness knows how many GPUs inside. What sort of rights do you think that android has, or should have?

Claire: I think rights come with sentience, and it's a tricky task determining when something truly is sentient. It's going to be tough to really be sure, but I think rights come with sentience.

Leo: That analogy, to how a baby learns, is close to making the claim that there are a lot more similarities than we might intuitively expect.

Claire: Yes, and we have no idea what that last gap to true intelligence is. It could be two years; it could be 50 years. I tend to think that it'll be within our lifetime. Hopefully whatever becomes sentient views us in a positive light.

Leo: I wanted to return to those themes that you bring into your AI work. How do you go about doing that? How long does a prompt look like for you when you're working with an AI model to create your art?

Claire: It depends on the tool, and it also depends on the piece – sometimes a few words are enough. I know one artist that does one word prompts. That's their entire practice as they simplify a prompt down and try to get it perfect. Sometimes it's entire paragraphs.

There's something called aesthetic embeddings that I use, taking five or so images of my own work and training a mini model on that. Then when I go in and I prompt on a larger blank model, I can call on that mini model and it will take the entire aesthetic of that I trained and apply it to the prompt.

“My days are often nine to ten hours of just getting my hands dirty, exploring, and saving presets.”

— Claire Silver

Leo: There's a lot of learning how to get the most out of a model. How do you go about doing that? Obviously, it's trial and error, but are there shortcuts that you take to figure out which things to trial?

Claire: Not really, my days are often nine to ten hours of just getting my hands dirty, exploring, and saving presets when I've typed the right words to make something that I like. However, there's a particular open-source tool called ControlNet for Stable Diffusion. It keeps the composition of an original image, but it can change the style dramatically, mix in other images that you give it while keeping the composition, create depth maps for 3D use, or line traces of all the edges.

When you start mixing all of those together with your original image it breaks it in these beautiful, fascinating, surprising ways. The more you iterate on it, going deeper and deeper, the more original the work looks. You can create aesthetics that I think are unique to AI, that have never been seen before and can't be made any other way. That's what I'm interested in.

Leo: It makes me think of just what a crazy creative future even the next three years is going to be. New genres lead to us thinking of beauty in completely different ways, it leads to us looking at art in completely new ways. How far are definitions of beauty and things like that going to change?

Claire: Accessibility breeds innovation, right? So, augmenting skill in this way is going to have cultural impacts far beyond what we're seeing now. There’s also an artist pipeline, at least in my experience, where you start out wanting to make something beautiful, then you want to start to make something meaningful, and then you want to make something weird.

Eventually you start pushing the boundaries. If beauty is accessible to everyone, it does decrease the importance of visual beauty when separated from message, meaning, emotion, and expression. I think storytelling, emotional connection, and cultural commentary are going to become more and more important to us.

After a generation or two of AI being in every part of our lives, augmenting skill, I hope that we begin to value other things in ourselves besides this idea of productivity and what we can make in this job-like system.

“Storytelling, emotional connection, and cultural commentary are going to become more and more important.”

— Claire Silver

It's all tied together: the philosophy of skill and hard work and dedication, and how we view our role in society and how we view ourselves and feel about ourselves and others. And I hope that after a couple generations with AI, maybe we start to value storytelling, maybe we start to value message, innovation, creativity. The kind of things that children value.

Leo: A lot of what you talk about online is the idea that ‘taste is the new skill’. I'd love to dive a little deeper into what that means. You talked about storytelling and childlike curiosity becoming things that we emphasise more in society over skill. How do you think we get there? What does that future really look like to you?

Claire: ‘Taste is the new skill’ is born out of the idea of taste as your fingerprint. It's your artist signature, it's who you are. I don't know if that can be taught or not. So, people with taste can express that through the augmenting of skill and I feel that those people are going to rise quickly in the cultural zeitgeist.

Leo: I think it's a really interesting idea and potentially a provocative one as well. Skill is usually something that people can learn, so there's a lot of capacity for accessibility and for openness. But if taste becomes more important than skill, and that can't be developed as much, I can see that being a provocative idea. What do you think about that and what it means for those who are creating with artificial intelligence?

Claire: I can also say that there is certainly a lot of skill involved with AI art, and AI-collaborative art, if you want it to be there. If your taste is something that you feel is not fully developed, you can lean into skill. With AI tools, it's endlessly deep in the things that you can learn; it's taken me years and the tools keep evolving so there's always new skills to develop.

Leo: Do you think AI tools will get better and better, to the point where a human that has skill just gets out-competed by a human who has taste, working with an AI model?

Claire: I would love to think that art could be the one place that we don't have to be competitive in the same way. I would love to think that individual creators can create work and share it, in a global market, with people all over the world able to see it via the internet. There will be people that resonate with it and form a fan base and collect.

When it comes to studios using artists to create work, I can understand the trepidation there. Just like YouTube displaced cable television, AI will displace these content factories. I think it empowers the individual above the corporate environment. Just as in the industrial revolution, there will be changes: there was a lot of job displacement, but it also opened up all kinds of new jobs, including ones that were created by people themselves.

“I'm interested in what the blue-collar worker has to say.”

— Claire Silver

I also would push back against the idea that skill is something that anyone can learn. I have a chronic illness and I'm in remission, so I've (had the time and) been able to improve. I've had a lot of people with disabilities reach out to me and say that they're able to use AI to express themselves and it's a whole new life for them, it's liberating.

I come from poverty and the people in my little town would work ten or twelve hour days at a factory or as a Walmart greeter and make barely enough to scrape by. When are they having time to teach themselves? We've spent so much time learning from people that have traditional skills – and that's absolutely incredible – but I'm interested in what the blue-collar worker has to say, what their perspective is, when given tools that allow them to express themselves artistically.

Leo: I think it is hugely valuable that there is now this platform with an AI tool that lets anyone create art, language, and music. One potential future that I think people see is this idea of AI as a paradigm-shifting technology that flips the table on who has the opportunity to create something, and to change not only their life, but also the societies they inhabit and the communities they are involved in.

How far do you think that opportunity extends beyond art? Also, what sort of questions do you think society needs to be asking of itself to make sure that that is the future that we end up living in (rather than a future less accessible or more dystopian)?

Claire: I feel that AI will be a species-level shift on par with our species branching off into homo sapiens, and that it will be that way forever, until there is no ‘us’. It’s that level of depth and weight. It will be in everything; medicine, architecture, in everything.

With that said, the questions that keep me up at night, which I think we need to consider now, not later, are mostly to do with our humanity and with society. For example, we value innocence in children, we protect innocence in children. So, a child that has, say, the Neuralink implant, and has access to the breadth of all human history and knowledge on demand instantly, are they innocent? If not, then what does that look like? Will we value children in the same way if they're not innocent?

Let's imagine a future with much less suffering in it. Will a child that grows up in a life without much suffering be a good person? I don't know if you've ever met anyone who's had a perfectly comfortable life with no issues, but I've met a couple and for the most part they're not super empathetic. Without that hero's journey, will we be empathetic, good people?

The hero's journey is our oldest story, and it's about overcoming. If AI solves a lot of these things that we have to overcome, then what story do we tell as a species of ourselves? How do we think of ourselves without that integral part?

Leo: Good conversations always end with even more questions than answers, I think. Questions about empathy, innocence, the very essence of humanity, and potentially the next chapter of the homo sapien species. I can't think of more important questions than those to end on.

Claire, this has been a fantastic and fascinating discussion. I've really enjoyed diving into not only the technical questions of how you create, but the big picture ideas that really make the future of culture and of creativity and what it means for us as a species. Thank you so much for coming on.

Claire: It's been an honour. Thank you so much for having me.