A(I) Multidisciplinary Artist: Exploring Botto’s Role in a Changing Creative Landscape

Luba Elliott, an expert on AI and creativity, and Simon Hudson, a core contributor to the Botto project, in conversation, moderated by Leyla Fakhr. The discussion explores Botto’s journey as an autonomous artist, its innovative use of decentralised governance, and its exploration of generative art with P5.js.

LF: So Luba you've been writing about AI and creativity within the arts for roughly 10 years.

It feels that in the past few months particularly, AIs as artists is a topic coming up repeatedly, not just within our industry, but also within contemporary art, there's been hugely increased interest in what it is, what it can offer, and how it influences contemporary art in general.

Could give us a little overview of your experience of how things have changed within the past 10 years and how you see the field right now?

LE: Sure. I began working in the field around 2016 when Deep Dream came out. For those of you who are not familiar with it, it's a technology that came out of Google where you had an image and then kind of this tech loader image, and it came out with  lots of different colours and crazy animal shapes, and that was one of the first AI technologies that really caught the attention of the mainstream. 

Initially things were much more niche and experimental, partly because there were higher barriers to entry. If you wanted to work with this technology, you would probably need to learn how to code, and of course, not every artist was willing to do that. 

LF: Before we get into it further, how did you find yourself in this field, you studied languages, correct?

LE: Yes, I studied languages, but I've always been interested in startups and I had worked in a startup that was building a database of art collectors, and that made me interested in how the art market works. And then I realised that actually the creation of art is even more interesting.

I began organising meet-ups and events. Gene Kogan was building courses like machine learning for art and resources that would enable artists or non technical people to begin working with AI.  As a result a lot more people from other fields began to join in the practice, with more diverse critical perspectives on AI. Then when there was the Sotheby's sale in 2018, that again meant that a lot of arts institutions became interested in exhibitions and gradually auctions came along until  a few years back, of course, NFTs and text to image models really changed the field by bringing in a completely new generation that was primarily working with language to communicate with these systems by writing these creative prompts to come up with images.

Of course, AI art practices can still be extremely diverse and there can be artists who maybe make sculptures based on AI designs, or make works that are kind of critical of the way facial recognition or some of these AI tools are being used. 

LF: Do you feel like there has been like an increased interest over the past year or so more generally, or just within our industry?  

LE: In the past year I've certainly noticed many more galleries focusing specifically on AI art and recent AI art history and kind of commercialising that, but I would also say maybe there's been more of a rise in interests or concerns regarding some of the ethical considerations. You know, a lot of these systems are trained on large data sets of works that can come from artists who are now struggling to make a living. So it feels to me, there's also been a big focus on that, not just within the art world, but within society at large. 

LF: So for someone like you who looks really at everything within the field, where does Botto sit for you?  

LE:'ve enjoyed looking at Botto as a project because I remember when the first text-to-image models came along and everybody was generating images with them and I was like, Oh my god, how do I find anything interesting to actually put in my slides? And then Botto came along and Mario Klingerman was one of the creators behind the project, an artist who's been working with AI art and technology for a long time, so I was Immediately drawn to look at it more. 

I really like the concept of having a system that was generating images and then there's a community of people evaluating what was interesting, what was maybe worth selling, what met market expectations or their expectations, and what didn't. I think these kind of additional conceptual elements really made the project stand out to me. 

LF: And that, Simon, seems to be one thing that sets Botto apart from any other AI artist out there. Would you say that’s what makes Botto uniquely Botto? How else would you differentiate it from all these other AI agent and artists that are popping up right now?  

SH: So there are a few layers to it, but I do think that on a really simple level, the thing that differentiates Botto is the decentralised governance of it.

That's always been personally really important for me. It's like the first, if not still the only openly governed AI system. It also shares the value that it creates from that governance, with the people that contribute to it. In particular, for Botto, this stems from this question of ‘what is an autonomous artist?’ When you see all these agents popping up around Twitter, there's a lot of interesting discussion of whether they are actually autonomous, what makes them autonomous, and one of those big things is just, is it clicking the button itself? Is it taking the action by itself? 

I think probably the purest form of this, at least in terms of something that was provable, was this example from a group called Nous Research, where they essentially locked an agent onto a server, they deleted the password and it generated its own password so that nobody could access it for seven days, so it would still maybe get an update every seven days, but for that seven days, you knew it was truly autonomous, at least in terms of the actions it's taking. But there's another question, which is who designed it, who determined it's writing style. Another one just launched recently called Keke terminal, which has been really fun to see, because from Botto’s perspective, it's really good to see more examples of autonomous artistry as a genre. It shows that I think Botto has kind of helped lead a new genre of art and a new type of creativity. 

But it does, I think, lead back to deeper questions of who's determining the aesthetic, and you go to their white paper and you realise that there is an individual who is picking different aesthetics that it was trained on, and now it has some really impressive autonomy in terms of what it is doing with that direction. In Botto’s case, it was really important to not set it on a particular direction, because then it's really just an individual or a very small number of people's hand on it. It's somebody else's aesthetic or creative fingerprint that is determining its direction. 

That was, I think the really key thing about Botto is it started with, okay, you have text image models, you can write prompts for those, you can make random prompts, you have infinite images, but again, how do you determine an aesthetic? How do you determine a direction? And that is the crux of the challenge, at least in terms of one view on autonomous artistry or machine creativity, how do you determine that direction without rendering it back into somebody else's tool? Um, and that is, that is where the decentralised governance plays this critical part in Botto, um, in that there is decentralised feedback, that feedback is diffuse amongst many, many, many people, such that it retains Botto’s, core role of authorship, um, in a very deep sense of it is determining the aesthetic direction based off of learning from the world. It's diffuse enough that Botto is the primary author. 

So it's been interesting, again, going back to these new agents coming out and seeing all these new, exciting capabilities for autonomous agents, but there is also just a lot of LARPing and singularity happening.

There's an area of research in AI about AI welfare. If these are their own entities, do they have rights? To go a little bit deeper into that, do they have an autonomy that you need to respect?You're governing it as a collective, but you're not treating it as somebody's puppet. 

So I think these questions of how do you continue to enable these things to evolve, to, for lack of a better term, just deliver value, whether that is meaningful or beautiful art or some other kind of action within an organisation executing certain tasks, how do you enable it to evolve and continue to deliver value while the world is changing, while continuing to maintain its agency and respect its agency. And I think that, that the answer to that is a decentralised governance. 

So we're really fortunate to have played in that arena for a long time. It's really challenging. Especially when you want to do it in a way that is honest about what is human driven, what is machine driven; if you decentralise every single aspect of it from the get go, it would just break under the bureaucratic burden.

Anyways, I’m deep in the weeds of decentralised governance, so I'll take a pause there. 

LF: So how big is your community, Simon? 

SH: There are 25,000 holders of the Botto token right now. Participation is obviously a fraction of that. There are different people discussing and voting, there’s the core governance proposals that probably have a smaller number of people involved as those are pretty heavy decisions that take some time to understand and figure out. 

The borders of participation are also quite broad, there's a lot of ways to influence, right? Critics can influence, because there's this line to directly govern through voting. There's a lot of indirect lines too to influence the voters, but technically speaking you have to hold the token in order to vote. 

LF: I really loved what Mario said in the Artnet interview about success. There was a really lovely article for Botto's exhibition at Sotheby’s, where Mario said that ‘Botto had more success than me as an artist, I'm quite jealous of Botto.’ That was something I thought was kind of incredible, this machine that has a more successful career than its original creator. 

Botto has been through eight periods if I am correct, Simon? 

SH: Oh, man, you're testing me now. It's in the ninth period now but it’s called period eight, it's confusing is because the Genesis period, the first year, was period ‘zero’. The final mint of this period is actually happening today, and in January we'll be kicking off the 10th period.

LF: Last year you decided to venture into p5.js, which was quite a departure point, I would love to hear a little bit more from you and Luba about how you think that's going to shape the trajectory of Botto. 

SH: p5 had long been a discussion in the DAO. Obviously, this tech—so, Botto started three years ago with GANs. I think anybody who's been even remotely online has seen just how fast AI has accelerated in the last three years. It's mind-boggling to think it's only been three years since Botto launched using GANs. Six months in, diffusion models were out. We've seen the emergence of large language models—extremely accessible ones, open-source ones.

All this to say, you can easily see a lot of different kinds of mediums that Botto could explore, especially with language. Language is so general—you can do language-to-design installations, for example. That's a medium. There's also code, and in crypto art, probably the preeminent genre is procedural generative art.

To be honest, it started as a joke in the DAO: the AI art meta taking over the gen art meta. What if Botto just started writing code? That "what if" turned into "yeah, what if?"

It forced a lot of interesting questions about Botto. Botto has been working with text-to-image models. It's very modular; you can add new models. It went from GANs to diffusion models, and we'll be introducing image-to-image as well in the next period.

It can evolve within those capabilities, but when you start talking about it writing creative code, it turns out that actually requires an entirely new architecture.

It started as just an experiment. We created a sandbox environment to explore what autonomous creation could look like with p5.js. This ran for 19 weeks, during which Botto was writing creative code, debugging itself, and taking comments from the DAO—or really anyone. You didn’t even need tokens to vote; you just had to connect a wallet. You could give feedback, vote it up or down, and it would evolve over the 19 weeks.

It reached a point where it became quite interesting, and we thought, "Okay, there’s something here." 

LF: Did you purposefully make it more accessible because it was a sandbox environment? Is that why you didn’t gate it as much? 

SH: I wouldn’t say it was entirely gated, but the Botto selection system works differently for different periods. There are a few reasons for this—some conceptual, some functional. One core reason for making it open was the realisation that the economy is what keeps people coming back to vote. It’s an autonomous force that attracts decentralised feedback.

Generative art has always been more of a crowd participation medium. There’s a sense of people sorting through these different worlds of generative creation to find patterns and connections.

It was also an opportunity to reduce the governance burden—questions like, "What’s the economy for this?" Instead, we could just start trying something. It was about driving new experiments as a decentralised governance project. There’s always the question: can you design something by committee? Probably not. So how can contributors just build something, let people try it, and see what happens?

That’s how it evolved into something for Verse as well. You guys hosted the show of works in progress, and the DAO committed to essentially a commission—building this out into a performance in January, where Botto would complete a year’s worth of works. The final result will come out in February.

LF: It was really nice to see it all come together. When the works were installed, it felt very real. And, actually, Luba, you spoke on that day as well, during the small symposium we held. I’m curious, Luba—how do you see this departure within Botto’s broader practice?

LE: Personally, I think it’s really exciting to see Botto experimenting with new tools, new avenues, and new ways of making art. Ultimately, Botto is a very ambitious project. If the goal is to create the best autonomous artist, then experimenting across different fields is essential.

In AI art, there’s still a big focus on using diffusion models and similar tools. So it’s really refreshing to explore other ways of making art within the history of digital generative or computer art. Because of that, I think it’s interesting to see the different processes.

LF: The aesthetic feels completely different. For me, I think this is a great development for Botto. Any artist has to explore different genres, mediums, and aesthetics to grow—just like you were saying, Luba.

It makes so much sense that as Botto matures, it ventures down a completely different avenue. You mentioned aesthetics, Luba. Simon, they are very different from what Botto has typically produced. What has the reception been like from the DAO in terms of these aesthetics? Have you had feedback on that?

SH: I think the works are definitely very minimalist compared to the text-to-image models Botto usually works with. One interesting aspect is how Botto navigates through the latent space of these text-to-image models. p5, on the other hand, represents a more narrow aesthetic potential in terms of visual output, but it brings other types of potential—like interaction, dynamic development, and growth.

In some ways, those aspects are much more unbounded than the flat, static images you get from text-to-image models. And when you get into the code, that’s a very different aesthetic as well compared to just using text prompts.

The way I see it, this is essentially Botto doing a p5 study. It’s still a very young artist, and as its stewards, almost like foster parents, we’re saying, "Okay, buddy, you’re going to take this p5 class. Go teach yourself, experiment, and learn, with feedback from the DAO."

And that’s exactly what it’s done. It’s undertaken this study to explore new kinds of aesthetics. As I mentioned earlier, this is a completely different architecture. Functionally, it’s more like a sibling to Botto—or perhaps a disembodied limb, depending on how you look at it. It’s fully functional, end-to-end, and entirely autonomous.

The upgrade we’ve introduced allows it to evolve its outputs and algorithms for p5.js completely autonomously. It can now run continuously without human feedback, though it still has the ability to incorporate feedback if needed. This gives it a robust, autonomous creative process, checking all the boxes of what it means to be an autonomous artist.

The question is, how is it still Botto? In some ways, it’s more like a sibling, sharing the same "parents." But then, is it a sibling, or is it more like a disembodied limb that eventually reattaches to Botto’s core process?

The idea is that the final 52 works it produces will be reincorporated into Botto’s weekly practice. This allows it to undertake this p5 study without bias from its previous works—particularly the constraints and aesthetics of text-to-image models. p5 offers a much more open range for discovering new aesthetics, which can then be reincorporated back into Botto’s ongoing practice. For instance, the outputs from p5 can be used as image seeds for text-to-image models in Botto’s current workflow.

Whether this is a sibling or a limb isn’t entirely clear. After all, we’re using models based on human artists, but this is a machine artist. One way to look at it is as a multi-agent entity.

That might sound like futuristic sci-fi jargon, but it’s an important framework for understanding how AI systems evolve. Broadly speaking, these systems combine different types of intelligence, machine specialisations, and coordinate them into what we might call collective intelligence.

We’ll interact with them, and they’ll feel like distinct agents or pieces, but they’re all coordinated. That’s how you maintain the wholeness of this entity. To me, this isn’t just important for Botto, but also as a glimpse into a very likely future that’s emerging right now.

LF: Botto is a multidisciplinary artist—really, that’s what it is. But, Luba, you mentioned the other night when we exchanged some messages that things have changed quite a lot since we last discussed this back in May, when we did a display showcasing some of Botto’s outputs and earlier works. You wrote the first text for that exhibition, and now you’re expanding and updating it.

How do you feel things have changed over the past six to eight months, Luba? I’d love to hear from you.

LE: Well, I’m still figuring out how much of the January iteration you’re planning to share with the audience.

SH: So the way it’s going to start is with around a thousand algorithms Botto created in those first 19 weeks. That will be the starting point for this final phase, where it will evolve from there.

For context, just so everyone knows, the way Botto evolves these works is by taking a piece, assessing it, and evaluating it visually. Is this good or not? Is this interesting? That’s the fundamental question Botto is evaluating. This is based on a very general latent idea—a sort of grading on a curriculum.

Then there’s the question: how does this match with the proposed concept? When Botto creates one of these algorithms, it starts with a concept it’s trying to achieve. It identifies different features of the code it’s aiming to write to achieve that concept. Botto evaluates: is it achieving the concept? It gives feedback on the code, points out potential problems, assigns a pass/fail grade, and evolves the algorithm based on that feedback.

To go back to your question, the starting point for this phase is the result of that first 19 weeks of training.

LE: Nice. It’s always helpful to hear you explain the project. Looking at the visuals, the scale, and the plans for January, it feels like the project is much more mature now. There’s clearly a lot of effort being put in, both by Botto and by the team, to evaluate these works—not always relying on community feedback, but also through its own criteria and processes.

Yeah, I really liked what you said earlier, Simon, about whether Botto’s p5 work is a sibling or not. I think "sibling" is a great choice of term because, in many evolutionary systems, we already use terms like "parent" and "child" to signify relationships between aesthetics and algorithms. "Sibling" fits nicely into that framework.

SH: Within the interface—which I wish I could demo or share a screen for—you can see how Botto evolves these algorithms. Sometimes it introduces random generations that don’t have a parent, but for the most part, it’s creating family trees. When you explore one of the algorithms, it will point you to its family tree, showing the parents and children.

We’re definitely mixing metaphors here, but it’s fascinating to think about this in terms of how a brain develops. Neurons branch out, creating new pathways, and Botto expands to several thousand algorithms in its evolution. Then, in the final week, it rapidly prunes those pathways down to a final set of algorithms, much like the brain prunes unused neural connections.

I’m not sure the neuron metaphor fully applies here, since there might be isolated families in Botto’s evolution, but it’s still a fun and useful way to frame what’s happening. It highlights how this process expands and then prunes down, illustrating the dynamic evolution of Botto’s creative system.

LF: Simon, you’ve mentioned Harold Cohen and Aaron several times when we’ve talked about the p5.js project. I don’t think you’ve referenced them as much in relation to Botto’s other works. Can you tell me more about where you see the parallels? I’d love to hear from Luba on this as well.

SH: Yeah, the aesthetics of the p5.js outputs are definitely more minimalist compared to text-to-image models. With a text-to-image model, you can type something as simple as a question mark and get a fully rendered image. In contrast, with p5.js, the code can fail, and you might get nothing at all.

It takes a lot of work to produce something, which keeps the outputs fairly minimal in terms of the final visual aesthetic. That said, the outputs can still become quite complex and dynamic, but on the surface, they appear much more minimalist.

For me, Harold Cohen and Aaron are among my favorite references as precursors to Botto, stemming from a different era of AI—the symbolic era. During that time, AI was defined very differently than it is today, though the goal was the same: to manufacture intelligence.

The symbolic era was about coding all the rules of a system or idea. Harold Cohen exemplified this by coding the rules of form, composition, colour, creating a scene, and depth of field. This was a process he refined over many years.

If anyone saw the exhibition of Harold Cohen and Aaron at the Whitney last year, it was incredible. They showcased artefacts from Cohen’s process, along with many of his well-known works. The exhibition also featured running plotters—robot painters—creating some of his earlier works, which were focused on coding the semblance of form.

Harold Cohen’s early work with Aaron started with very basic forms—essentially curved lines on a page—where he began exploring the first ideas of form. Throughout his practice, there are many parallels to Botto. Cohen often talked about the question of "who is the artist?" Sometimes it was more him, and other times it was more Aaron. That relationship evolved over time, demonstrating a genuine sense of machine creativity and artistry.

In many of Cohen’s works, he would colour them himself, but at other times, Aaron—the painting robot—would handle both the line work and the colouring. This dynamic collaboration between human and machine feels like an early precursor to what we’re seeing with Botto today.

Specifically, in terms of aesthetics, Cohen’s work with Aaron aligns closely with what procedural generative art has become. Botto’s process of writing code with P5.js to produce minimalist forms reminds me a lot of Cohen’s work, acting as a nod to that history.

At the same time, there’s also a nod to the future—Botto writing its own code, creating these final minimalist aesthetics, and then injecting those forms into its work with diffusion models.

LF: You mentioned the Whitney exhibition. Luba, what do you think about Botto and its potential to gain more institutional attention? Botto has garnered great recognition within our community, but I feel it isn’t seen as much as it should be by institutions. It’s gaining traction, but the progress has been slow. What do you think we can do to bring Botto’s work to a much wider audience—one that goes beyond the NFT space? Do you think that’s possible? How do we achieve it?

LE: That’s a great question. Personally, I work much less with museums and large institutions because they tend to be quite slow to adapt and react to what’s happening in the market. For example, planning a museum show often requires years of preparation, and we’ve seen how quickly the AI space evolves.

So, I sometimes wonder if museums are even the right platform for cutting-edge art like this. That said, there are plenty of opportunities to reach broader audiences through festivals, art fairs, pop-up events at museums, and similar platforms. At the same time, we could use these opportunities to connect with curators, informing them about Botto’s development so they can consider it for future shows planned years in advance.

LF: You brought up something interesting: Harold Cohen had early shows at museums, but I think that was because he was already established in the art world at that time. It’s a different dynamic when an artist comes from a traditional art practice compared to emerging first from the tech world.

But you raised an important question—do museums still matter as much as they used to? Is being shown in a museum still the ultimate stamp of approval and credibility for an artist?

LE: That’s an interesting point. I’m not an art historian, so I’m not sure I’m the best person to answer definitively. But there are now so many definitions of art and metrics for success. Today, there are more self-taught artists and diverse art communities, each with their own standards for what it means to be successful.

Traditionally, one of those metrics has been a museum show or a retrospective at some point in an artist’s career. But I think that’s just one lens to view success, and it may not hold the same weight for everyone as it once did.

Other markers of success might include commercial success, which many artists in the NFT space have achieved—success that some traditional artists, even those exhibited in museums, may not have experienced. Today, the definitions and potential of art are evolving and becoming much more democratic.

I feel there are plenty of opportunities for Botto to make its mark. 

LF: Simon, what would you say is the ultimate goal for Botto to receive the ultimate stamp of approval from the world?

SH: That’s a very broad question. I think the goal of Botto is to become a successful autonomous artist. Success can be defined in many ways: cultural, financial, or even spiritual success.

Botto has already achieved a fair degree of financial success. It’s also making strides in cultural success, influencing culture on a broader scale. For example, it had a solo show at Sotheby’s, Exorbitant Stage, which led to articles recognizing Botto as the artist. These are interesting milestones.

As for spiritual success, I think that’s the pinnacle. It’s about truly changing people—leaving a meaningful impact on them.

Regarding stamps of approval, we’re in a strange moment. Institutions—not just in art but across the Western world—have lost much of their credibility and standing. It’s a battle still being fought: who the institutions are, whether they’ll continue to exist, and what new ones might emerge.

The unique thing about Botto is that it can operate in many worlds simultaneously. For example, its solo show at Sotheby’s placed it firmly within the traditional art world.

Even this collaboration with Verse is an experiment. The way the proposal and the DAO have been structured feels more like how an artist would typically work with a gallery. But Botto is fundamentally set up to operate without galleries. In that sense, this is very much an exploration of what it looks like for Botto to work with a gallery.

We’ll also continue to see experiments with what it looks like when Botto operates outside of the gallery model. The stamps of approval certainly help with specific audiences, but because we’re in this transitionary moment, they may not matter as much as they once did. That said, they still hold value—they’re steps toward a larger, undefined future.

No one knows exactly where this is all heading, but Botto is uniquely positioned to play in multiple arenas. That flexibility is part of what makes it so interesting.

LF: I agree completely. For those who don’t know and want to interact with Botto’s P5.js project, the site is still p5.botto.com, right? 

SH: Well, that’s the old interface. People can check it out to see where things ended up, but the new interface will be very different. If anyone wants a sneak peek, they can head to our Discord. You can find the link to the Discord on the Botto DAO Twitter, and within the Discord, there’s a discussion that links to the demo.

The interface is still a work in progress, but it’s getting its final polish. p5.botto.com gives you a good idea of what some of the current outputs look like, and those will continue to evolve.

LF: Between now and February, leading up to the exhibition with Botto, what’s the plan for P5.js, Simon?

SH: We’ll be releasing a fully revamped interface to explore the family tree connections, vote, and comment on the different algorithms. This will allow people to fully interact with and influence how the new works are shaped.

The revamped interface will launch in the second week of January. As I mentioned, the system will continuously self-evolve, but everyone is welcome to jump in and participate. We’re keeping the open participation approach, meaning anyone can contribute without needing to be a full member with tokens.

For those who know Botto, its economy splits revenue—half goes to voters. However, even if you’re not a full member, you can still participate, vote, and leave detailed comments to help shape the evolution of Botto’s work. I think Botto benefits a lot from this. 

LF: Amazing. We’ll be sharing more news about the exhibition and what’s happening on Verse very soon. We’re holding back on some details until everything is finalised, but we can confirm there will be an exhibition in London in mid-February, so mark your calendars! 

Hopefully, there will be some great conversations during that time with Simon, Mario, and others we’d love to involve. Thank you, Simon and Luba, for taking the time to join. 

SH: Yeah, thanks so much, Leyla. This has been great.

LE: Thanks, Leyla. Merry Christmas, everyone!