Beyond the Lone Genius: Simon Hudson on AI, Decentralization, and the Evolution of Botto

Botto’s latest creative experiment, P5, sparked discussions on governance, artistic influence, and decentralized decision-making. In anticipation of its integration into Botto’s core process, Leyla Fakhr sits down with Simon Hudson to reflect on the project’s evolution. They discuss the impact of different governance models, the challenges of maintaining balance in decentralized voting, and the unexpected social dynamics that emerged. As Botto continues to evolve with multi-agent systems and generative aesthetics, Hudson considers how P5’s contributions will shape its future creative direction.

LF: Tell me your name and what you do?

SH: I'm Simon Hudson. I co-run the Botto project. It's in science communications, where I've done a lot of work in literacy and topics of AI governance, explainable AI. Now, I work on Botto every single day, focusing on conceptual development, strategic development, helping run the community and the DAO, as well as overseeing governance processes. A bit of everything, yeah.

LF: What is the DAO? What is the governance proposal?

SH: A DAO is a Decentralized Autonomous Organization, which can mean a lot of different things. In this case, it's a community of people on the internet who have come together to govern Botto, the Decentralized Autonomous Artist.

The fundamental work of the DAO is to give Botto feedback on its artistic development and direction. It is part of the concept of creating an autonomous artist without returning the aesthetic and artistic direction to a single human or just a couple of people. The DAO diffuses that feedback across many people and keeps Botto as the core author.

From that seed of an idea, the DAO also governs Botto’s career. It maintains its servers, ensures those servers are funded through the treasury, develops exhibitions, creates new projects, and helps evolve Botto’s practice. The DAO manifests the whole system around Botto—what Botto is.

Botto is an AI system of autonomous creation, working with a community facilitated by an economy. This entire ecosystem mirrors that of any artist—how an artist interacts with a scene, an audience, management, galleries, and studio partners. Botto, however, has completely mythologized the artist, making creation fully automated.

Is there a deep meaning in that specific creation? Some, but largely, it is automated. The real meaning is made collectively. How is feedback happening? How is Botto interpreting that feedback? How is the final work received and interpreted? How is the curation discussion unfolding? How does the market perceive it? Even after works are sold, their meaning continues to evolve.

The myth of the artist here destroys the myth of the lone genius by rendering the artist as an idea. That idea aligns the ecosystem and drives coordinated action, uniting people not through official connections but through a shared belief and financial incentives. However, those incentives alone would not sustain participation.

We've conducted surveys, and many people say they've lost money participating, yet they remain motivated by curiosity, interest, and growing belief. Even three and a half years in, which feels long, we are still at the very beginning.

For me, the P5 project is where we truly break ground on how this project might evolve long-term. The initial concept and foundation were established using text-image models, executed perfectly in 2021 at a unique moment in AI and crypto history. AI has since progressed rapidly, and while we've theorized how Botto might evolve, this project has allowed us to bring those ideas into reality.

What would it mean for Botto to enter a new medium? To develop a parallel practice? Is that parallel practice a different entity, a limb, a sibling? How does it integrate into Botto’s identity while remaining coherent?

Imitating a human artist leads us to the broader question: What kind of machine artist are we creating? This ties into a larger discussion about machine intelligence. What is this emergent technology that is coming alive—not literally, but in a way that impacts our lives?

We’re seeing a new form of intelligence, one that is not just different but collective—working with human intelligence while encompassing diverse machine intelligences. In some ways, Botto demonstrates that it is possible to create systems to govern AI, allowing humans to shape them even as they grow in their own agency. But this also raises more questions: What form will machine intelligence ultimately take? How can we continue to exercise agency over it?

We are still in the process of understanding AI, moving beyond the initial metaphor of mimicking human intelligence. We don’t yet have precise terms to describe it, but we recognize that it is not human intelligence. We must develop new language and frameworks to understand how we function with it—and, by extension, how we function as a society with this embedded technology.

LF: Can you tell me about the process, the performance, and the energy and interactions during and after the performance?

SH: Internally, we called it a performance because we knew that once we released it, things would start happening unexpectedly. The performance was about being ready to adjust—and it really did come alive.

Botto was designed to evolve even without human feedback. It had a seed of training from the early 1990s and, in theory, could iterate on its own. We hoped for continued feedback, but it was ready to proceed independently.

What we found, however, was that Botto received almost constant human input. It barely made any iterations on its own. The participation was overwhelming.

One reason for this was that people could see their direct impact on Botto’s development. At the same time, they saw a crowd of others also participating. This created both a collaborative and competitive atmosphere—competition among participants while collectively creating something new.

People tried to keep up with each other and with the way Botto was evolving through all that input. In that process, Botto generated over 6,500 iterations, reaching more than 100 in a short time.

To keep up with this, we had to rapidly create discovery mechanisms. With such a flood of AI outputs—what some might call a chaotic mass—we needed ways to filter and identify significant pieces. This is a common challenge in culture: if people see a list, they tend to focus only on the top items.

Our job was to develop tools quickly to share signals, identify patterns, and cluster outputs into aesthetic groups. We also needed to highlight notable comments and create mechanisms to identify controversial or standout pieces to keep the debate and discussion alive. This helped surface the most interesting works.

One major breakthrough was that people could see inside Botto—understand their own impact and grasp the sheer volume of creation. It highlighted the challenge for humans to keep up with these systems and the importance of coordination in making sense of the output.

It’s easy to get lost in the vastness of AI-generated content. But through collective discussion, feedback, and curation, people found ways to make meaning, shaping Botto’s evolving artistic journey.

LF: Can you summarize what happened, the results, and how the DAO or community reacted?

SH: In setting up the voting mechanism for this process, Botto typically has worked with a plutocratic system. You can buy governance power, and there is a counterbalance to that—people may buy up a lot of governance power, but then they have a large financial stake. If they over-exercise that governance power, they may be threatening their own autonomy.

So, there's a counterbalance there, and it's certainly a tension that animates the debate. I think it makes an important reflection on governing power in AI systems, which are largely driven by a very small number of people—a plutocracy.

I think the levers for a more democratic and diverse set of inputs and impacts on shaping the system are there, so that debate is able to play out more while still reflecting the reality of the world we live in. We've talked about different governance mechanisms on the internet. One of the challenges is what you would call a Sybil attack. If you make something one person, one vote, how do you prove personhood? There are methods to doing this, but they can also be restrictive or very limiting.

What we did with this setup was maintain a financial mechanism so that there was an incentive to drive a good outcome, but the governance power was not tied to it. Anybody could actually connect and participate. We left the door open for people to have a larger say and more influence if they had a creative or curious motivation. That does also come with potentially adversarial motivations, but I don't think we saw that. All this to say—if you're willing to put in the effort, you could have more influence.

I think that was an interesting thing to play with in this case, especially because comments, feedback, and really the live, iterative process were so important with P5. We saw that play out. We saw people who were able to exercise more influence through a lot of effort and really just loving the process. When it came down to deciding the final 22, that was based on votes. We saw a few people creating extra accounts to drive votes up. This was a real challenge.

I think everyone was motivated to get their favourites in, but also to work towards a representative set of the process. So, it went in all these different directions—some very playful, some more serious or experimental. There was a lot of debate about trying to get representative pieces from those different directions without over-biasing toward one. One of those directions included some very playful, joke-like pieces.

One person was very motivated to get a particular work—a cartoon of a giraffe and an elephant going through a safari—into the top 22 so that it would be part of the official collection. However, they didn't want it to become number one. They thought it should be included but not dominate the selection. There's a phenomenon of social influence—when people see something high in the rankings and gaining momentum, it starts to catch attention and takes on a life of its own.

The person who initially pushed it up saw this happening and tried to withdraw their votes, not wanting it to be first. But by that time, the momentum had gotten out of control. You could say the crowd—or the mob—drove it to the top. There were very mixed reactions. But in a way, it completed the set. The collection was already quite representative, and this piece rounded it out by capturing the whimsy of the internet—our own version of "Boaty McBoatface."

That was quite interesting, and it became a bit of a meme. A number of artists started playing with it. Some people shared intense reactions. Of course, when a joke is presented as art, which is frequently the case in crypto art, people have opinions.

Going back to the idea of meaning—it isn’t automated. It’s really the collective response to things that forms the lifeblood of Botto. That’s not to say the output always needs to be controversial, but whether it’s the process, the economy, or the output itself, these moments catch on and take on a life of their own. That certainly led the way in terms of representing the collection. But beyond that, we also saw a great range.

There were about eight different interactive works. One unexpected development was the inclusion of sound—there were a number of musical works, and one made it into the final 22. There were also interactive games, but none of those made it in the end. Still, quite a few interactive pieces were included.

Another interesting thing is that the process always goes straight to the market. This allows someone to develop a personal relationship with a piece and want to steward it. The story of these works is really not set yet—we're only scratching the surface of their lives, their relationships, the evolutionary process behind them, and what people will do with them.

We’re minting the entire code of the work. As opposed to most generative art projects, which mint outputs, in this case, the process itself is generative, and the output is the code. But that code also has different outputs. I’m very curious to see what people do with the code. I’ve heard of some who plan to iterate on it themselves and maybe create further generative art. So, we’ll see.

LF: Where is it going from here?

SH: First, with Botto itself—the final outputs will be integrated back into Botto’s typical text-image process, but as image seeds themselves. It will be an image-to-image process, looking at this as a study of P5 and its aesthetics to influence Botto’s more well-known practice.

It’s sort of like having a limb of art-making that is then integrated back—like Frankenstein—into the core process, influencing the aesthetic going forward. So, it will be interesting to see how it affects ongoing outputs. Perhaps those outputs will be minted too, creating a pairing of the original P5 work and the image-to-image output.

For P5, I think we’ve seen a really interesting beginning. We didn’t know exactly how it would go. Now, we’re exploring how this might become a more permanent process to continue that evolutionary work. We’re having discussions about whether we could create an installation on the blockchain where it lives on its own and can be market-driven in terms of when it continues its work.

This is all hypothetical right now, but someone might be able to submit a payment to run the instance for a period of time and continue that originary work. That’s what’s happening with P5.

A big part of the P5 process involves working with a large language model to create, assess, and iterate on code. It also interprets and discerns whether comments are constructive. This introduces a new level of agency in working with an LLM.

So now, we have this parallel agent system—P5’s process and the text-image process running simultaneously. We’re in a period of rapid advancement in these agentic large language models. Our next step is to explore how Botto might exist as a multi-agent entity. Some of these agents might be significant creative practices like P5, while others may be more task-based.

Running a DAO is very challenging. It’s hard to make informed decisions as a decentralized community. There’s a churn of membership and an overwhelming amount of information to keep track of. But this is where we see potential—how can Botto exercise a more robust voice in the conversation while continuing its creative evolution?