Artificial Intelligence: The Good, The Bad & The Useful
An interview with Benedict Carpenter van Barthold on AI, human creativity, and the dangers of apathy in the face of change.
Benedict Carpenter van Barthold is an accomplished visual artist with work in collections across Europe, including The British Museum and a number of permanently-sited sculptural commissions. He is the Principal Lecturer in Fine Art at Nottingham Trent University and earned his doctorate examining the relationship between technology, working processes, and the evolution of meaning in the visual arts.
This means two things: he is very busy, and he is uniquely well-equipped to comment on the developing intersection of fine art and artificial intelligence.
Pertaining to the former, the original intent of this article was to interview Benedict for an Imprimo ‘Artist in Residence’ profile, but he had to reschedule. Pertaining to the latter, this rescheduling was due to a prior speaking engagement, undertaken as a member of the board of directors of Vieunite.
Learning more about his role within Vieunite, and his explorations of novel technologies within his own artistic practice proved catalytic, inspiring this altogether separate conversation about artificial intelligence, the sanctity of human creativity, and the dangers of apathy in the face of change.
Imprimo: You initially rescheduled because you were giving a talk on AI the other day. How was that?
Benedict: I was at a company that I’m advising. They’ve got a couple of AI/art projects that I’m helping them with. And they’re both really centered around the art market and the consumption of art. The core product is an innovative glare-free digital screen that is optimized for displaying textured surfaces, and it looks — at first glance — like a painting or a drawing rather than a screen. And the screen is connected to a marketplace. I’m helping them by talking to artists to engage them to put their work into this marketplace.
The AI becomes relevant when we start working with museum collections, and also CC0 (aka public domain) images. There’s an algorithm that’s been developed that can accurately determine when something was made and what style it’s in. It uses that to create a library, mainly for historical work. But then there’s another AI in development, which will assess the work that’s uploaded for its biophilic traits, which are those traits and features most likely to promote a sense of wellness in the viewers. This way it can curate content that makes viewers feel like they have a connection with nature. Further down the line, they hope to connect this tech with biometric proxies; the data collected by smart devices that can suggest how well we’re sleeping, our heart rate, blood pressure, etc.
Imprimo: So theoretically, you would then have a database of art that’s been gauged on its ability to trigger a certain biological response, linked to a device that’s getting live biometric feedback, that would curate your space, based on your mood. That’s science fiction.
Benedict: It gets better than that, though, because the other thing that they’re working on is an AI image generator called Vieutopia that’s trained to produce works in particular historical styles. Potentially, we can then use that data loop that you identified to generate images in an evolutionary way, which progressively grows more and more therapeutic. One of the things that I’m advising them on is how they can do that without upsetting artists or treading on living artists’ toes and impacting their revenue. They want to mark themselves out as being more responsible than and different from a lot of the other AI image generators, which — illegal or legal — are operating in ethically dubious territory. We’re training the data set on CC0 images. And they’ve been really diligent about removing images from their training sets that have watermarks or are claimed as someone’s intellectual property, or are liable to infringe on copyright.
Imprimo: Even with the due diligence you’ve done to ensure the use of responsible training sets, I could still see artists taking offense. The majority reaction amongst artists to AI tools, in general, seems to be a knee-jerk offense. You, however, seem unfazed; even enthusiastic. Do you see AI as a positive disruption that will ultimately bring net good to artists and the art world? Or do you see this as a negative step that cheapens human effort?
Benedict: My position is that the genie’s out of the bottle, and it’s not going back in. And I’ve got a certain context I bring to thinking about this because I’m not just an artist, I’m also an educator. A student who’s making the decision of whether they’re going to study creative arts will be graduating three years from now, and there are already lots of examples of generative technologies, which are extraordinarily powerful, and are capable of producing digital images that have a sense of human creativity and human feeling about them. At the moment, I think the cultural value of those images is novelty, but that will change when people get used to the technology and use it for commercial purposes. As an educator, I think if we don’t talk to our students about that, then we’re doing them a disservice when they graduate.
Imprimo: So what does that mean for human creatives?
Benedict: I think as long as we remain in the business of curating human value, and curating and creating human interest, it’s not the end of human creativity, but it will necessitate the acquisition of new skills.
Artists need to understand how it works to take steps to ensure they don’t get ripped off; that much is non-negotiable. Then on one end of the spectrum, you’ll have artists who develop real expertise in using these technologies as a fundamental part of their workflows, and then at the other end, there’ll be those who position themselves in reaction against it. I don’t think that last one is very good, because that will pretty much condemn the arts to become some kind of weird heritage craft.
So it doesn’t really matter if it’s positively disruptive, it’s disruptive. And the only way in which it becomes positive is if artists engage with it, and help shape it. We need to ask ourselves, who do we want to manage the tool? Who do we want to be using it? Personally, I want to understand it as well as I possibly can, then I want to be able to use it wherever I can as part of my workflow.
Imprimo: You don’t think this sudden increase in accessibility — that anyone with an idea, irrespective of their artistic skillset can now generate something that at least approximates art — inevitably devalues artistic craft?
Benedict: Where we’ve got an edge is that humans are interested in other human subjectivity and human understanding. And the more time you spend with generative AI, the more you realize it’s algorithmic. You get out what you put in and that’s it. With the material and embodied experience of painting or sculpture, something irrational happens. It’s not empirical. It’s not scientific. But it creates a very strong, emotionally enriching feeling of connection. Now, I have not seen any AI art that’s been able to go anywhere near that.
Imprimo: Do you think AI art will ever achieve the same intangible emotional resonance as a human creation, or will it likely never be more than the sum of its parts?
Benedict: While an AI image generator can use the learning it has undertaken in the course of digesting trillions of images to produce a succession of discrete objects in response to human prompts, those images only share the common property of being part of an algorithmic process. They’re not part of any other story, or any other cultural economy until a human curates them into that economy or into that story, and I think that another human would struggle to take one of those images from that series and see it as being part of a broader project.
In contrast, if you imagine an artist who’s made a series of paintings you can see the development of that story, and that’s got a completely different cultural value. Even if one was to isolate just one of those paintings, I think if the art’s any good, a human user would be able to see that isolated painting in some kind of hypothetical connection with another series of works and see the narrative around it in a way that you don’t with an image which has been made algorithmically. Once the artist has made that series of paintings, maybe one gets curated into a collection, and it carries on developing meaning. It’s got a kind of latent potential, which is enlarged through interpretation, through criticism, through exhibition, through provenance and ownership. It goes on growing in a way that I don’t see happening with AI-generated images. I don’t think that philosophically speaking, there’s anything to stop an AI image from entering that larger economy. But it will only do so if artists use it, and as soon as that happens, it shifts from being algorithmic to being something else.
Imprimo: So until genuine artists re-contextualize the output of image generators as genuine art, they’re not really a threat to anything.
Benedict: I think we’re living through a moment that’s very much like the invention of photography. You’ve got this disruptive introduction of a new technology, which suddenly makes the craft skills of embodied expression look, well, a little bit expensive. And that’s freaking everyone out, particularly since it also looks as though it’s stealing artists’ vision in order to create those things. Now, that latter part I think, is actually a little bit hysterical. I think that it probably isn’t really doing that. Because well, if it is doing that, then the work is probably not very good.
Imprimo: You echo something that’s been bouncing around in my own mind. My initial reaction to ChatGPT as a writer was fear. But I’ve played with it a lot since, and for fun had it write an academic paper. It spat out the requested information to a tee, but there was no depth, no analysis. It did exactly what was asked of it, and nothing more. I find the same in a lot of the generative art I see, even the “good” stuff. So I feel if anything, AI has really only raised the bar of mediocrity.
Benedict: Yeah, really good way of thinking about it. As an educator, ChatGPT is hugely challenging if the way in which you assess students is to get them to write an essay where they’ve got a high degree of discretion regarding the content. I think — as you said — that we look at that and it looks scary. But if you switch it around a little bit, you could look at that and think actually, that’s useful. Because if my students are producing work that is at the standard of simple kinds of recall and mimicry, then they’re not producing analytical writing. They don’t have any real insights.
I think if the hard lifting of creating is removed, then the utopian view of this technology is that it will make us better. Now, for example, when I’m doing a piece of writing, it’s not about the grind of putting the words on paper. It’s more about what I want my voice to contribute. And if you’ve only got more time to give to that, you’re exerting more energy drawing out something which is latent to yourself and your own particularity that you didn’t have time to do before. That’s got to be positive.
Imprimo: I guess if you’re forced to strip away the “physical labour” of creation, all you have left is a concept and a pretty stark view of whether that concept is contributing something novel or if it’s just more noise.
Benedict: You’d asked earlier if I think this tech is a positive disruption? I think it probably on balance is because I think it threatens mediocrity. And I don’t mind if this comment upsets people because a lot of artists will need to be made really angry before they start engaging with this. And those are the people that need to be reached. This technology is like photography, but even more disruptive than that. And if you’re not with it, you’re going to get left behind. Use it, figure it out. Get mad at it. Try and smash it. But pretending it’s not there is not a very intelligent option, because even artists who don’t want to use generative AI will be affected by it. AI is already curating content. And as an artist, you might want to understand how that happens.
Imprimo: A lot of the conversations around AI art, while perhaps rooted in some competitive insecurity, also pertain to the unauthorized use of work in data set training. Whether or not the AI presents a serious commercial competitor, it’s not really fair game, is it?
Benedict: As an artist, I think there’s a lot of work that needs to be done here. We need some really robust protocols so that artists can share digital images of their work and know that those images are not going to be used for training purposes. I would like to see some kind of industry-standard being produced here. If you make an image that is your IP, I think that it’s reasonable to say you get to control how that image is used. An artist should be able to say “No, I do not want that to be used for training purposes.”
And that’s one of the things I think, by the way, which I think is really great about Imprimo. In order to have these recognized standards, we need authoritative metadata. The use of blockchain technology to reinforce the claims of ownership of intellectual property is going to be exceptionally valuable as the legalities of AI develop.
Stay tuned for more from Benedict…
This conversation has been condensed and edited for clarity.