An Intercontinental Conversation between @falco, @harshitrnnh, and @pindar

Pedro Falco

As human beings, our main production is data. As a visual artist, I do generative art. My work seeks to reinterpret data and use it as raw material. Through programming I write images that reflects the trail we left.

Harshit Agrawal

Harshit is an artificial intelligence (A.I) and computational artist. He uses machines and algorithms and often creates them as an essential part of his art process, embracing becoming the cyborg artist. He often juxtaposes traditional art styles and visuals with machines and computation, creating a space to both direct, and be guided by the machine.

Pindar Van Arman

Pindar is an AI Artist exploring the intersection of human and artificial creativity. Winner of the Robot Art Prize in 2018, his robots use a broad array of deep learning, generative algorithms, and feedback loops to bring his AI creations into the material world one brush stroke at a time.

Pedro Falco of Argentina (@falco), Harshit Agrawal of India (@harshitrnnh), and Pindar Van Arman of the United States (@pindar) are SuperRare artists specializing in generative and AI art. None of them had met or spoken before this show, but hearing that they were going to be exhibiting art together the three got together for a conversation. This is a record of their conversation where they do a deep dive into each others art, process, and cultures.


Let me begin by saying I am really excited to see each of you here on SuperRare. Big fan of both generative and AI art. After falling in love with several of your new works visually, I was hooked, and had so many questions about how they were made, and who they were made by. 

So how do I begin getting to the bottom of that. Maybe I can direct my first question to @falco.

In your description of your work Gray Matter you talk about how it is like a nervous system, with millions of dots, passages, connection…  In Ether Crib, you talk about a voyage through the ether.

Gray Matter
Edition 1 of 1
Gray Matter is my first generative artwork. It’s a continuous flow of data that drifts like a nervous system, leaving a trace, drawing a path. More than a million gray dots, invite us to see secret passages that thought takes to find new connections.

These are wonderful descriptions but it did make me wonder, did your generative work begin as an idea that you created the artwork from, or did the artwork come after the ideas you describe?


I was iterating on generative systems and making it more and more complex. So the generative system comes first. I like this idea of creating something and then giving it a name. It’s like something was born and I look at it and I see the meaning of that. In a way it is an attempt to consider myself as a creator, so I design a system, the system draws and finally I am the one who makes the futile attempt to communicate things. I try to escape to the idea of art or design as a moment of inspiration. In the specific case of “Gray Matter”, I was deeply obsessed with reading how the brain works. I was interested in this idea of little things participating in something bigger. Small agents move and participate in something important and each point, no matter how small, has a crucial impact on the entire process. So like I said, I’m always on the edge of what is abstract and figurative, I always try to connect the dots and try to choose a concept that deals with technique, what I see and what I think.

I had a similar thought when looking at @harshitrnnh’s work. 

I was looking at your Machinic Situatedness series, @harshitrnnh, and saw what clearly appears to be budhas in the compositions.  Knowing that you are an AI artist I realized that you are probably working with a dataset that includes imagery of Budha. This made me wonder what you thought about collecting and managing the datasets? Does your artwork start at that point, before, or after?

Machinic Situatedness 5
Edition 1 of 1
Machinic Situatedness is a series of artworks that uses artificial intelligence (A.I) to explore the subject ‘cultural situatedness’ and influences in the genre of AI art. These works are created by an A.I drawing inspiration from Budhhist painting styles to create an abstract, dream like output using GANs. This series asks the questions- what is an AI machine’s cultural underpinning and how can we broaden its scope? These draw a lot of reference from Nam June Paik’s TV Buddha work, here alluding to the cycles of transcendence which we undergo as a species, continuous cycles of trying to become something more than ourselves, which we are now channeling through the evolving role of AI in our lives and AI itself is by learning more from us.


Datasets play a very crucial role in my AI art practice. I take a lot of time to curate the datasets as I want them, curating each image of it. In some of my works, I create datasets too, by some underlying AI process like image to image translation and then use that to further train my art generating AI system. The artwork, however, begins prior to that in the conceptualization phase, where I think of what I want to communicate with the piece or series, what I want it to look and feel like, how I plan to produce it and such. Once that plan is set, yeah, dataset creation or curation is one of the first steps in the process. In recent projects, I have explored dataset collection by services like mechanical turk too. Another aspect that I think about often is what kinds of art are digital archives or datasets available readily? It usually is of European or American art, and as one can see in the AI art space, most of the art is an outcome of those available datasets, be it Wikiart, Google Art archives or large museum collections like that of the Met or some more European ones. This tends to perpetuate the aesthetics of art of these regions even further with AI art. As an Indian working with AI art, I try to often work with art closer to home, as you can see in the Machinic Situatedness series.    

Looking at both of your art, I am curious about your datasets.  So let me begin by asking you, @pindar, what dataset do you begin with? You paint portraits of specific people. Is their face the only data you use, or is there more? Furthermore,  after defining the data, are you surprised by the machine’s output, or you know roughly how it’s going to turn out based on your setup?


I have no idea how paintings will come out when they begin, and this is because I do not know the dataset before it begins.  There are some specific inputs that I begin with, like images of the person to be painted and images and data from the robot’s previous paintings.  But the most important piece of data the robot works from actually comes live from the canvas.  My robots paint with feedback loops. So every brush stroke it makes, it photographs and uses that data to decide on the next brushstrokes. This is easiest to understand in my timelapses. For example, watch a timelapse of Portrait 18,384.

Portrait 18,384 by @pindar (,384-13132)
Portrait 18,384 by @pindar (,384-13132)
Portrait 18,384
Edition 1 of 1
Portrait painted with acrylic on canvas by an autonomous creative AI using 4 reference photographs, 24 competing algorithms, 2,298 visual feedback loops, 18,384 brush strokes, millions of aesthetic decisions, and countless calculations.

Notice how the robot changes direction three or four times before finally settling in on a design it liked and finishing up the details. This is because there are dozens of creative AI agents, six of which are visualized in the time lapse, competing and fighting for control of the brush. As the painting progresses, the robot is constantly analyzing it in a creative feedback loop. At the same time it is re-imagining the piece based on what is emerging on the canvas, continuously adjusting its direction and interpretation of the data it began with.  All the creative AI agents fighting to influence the final image live as the robot is painting, so I have no idea what will emerge.

This is really how human artists work, isn’t it, which is my goal. We make marks, step back and evaluate the marks, then make the next marks. And when you think about all the datasets that the robot is using, you realize that the most important one is being created right there on the canvas. It is the progress being made on the actual painting. Am striving for a generative painting system so complex, that true creativity emerges from it.  Some people think I have already achieved it, others think it will never happen.


Interesting.. And @falco how do you select what data sources you want to incorporate in your work? What is the relation between agents and the data sources?


As well as being an artist, I also consider myself a researcher because I do a lot of research on different libraries, web api, sounds and images. I spend a lot of time selecting and ordering data sets. I’m also a big fan of algorithmic functions, so I use a lot of different noise algorithms like Perlin noise, Curl noise, and Simplex noise, to name a few. I process data as color values, so RGB represents the XYZ values of an apparent 3D space. 

So for example the white RGB value is 1-1-1. So, 1 pixel takes coordinates 1-1-1 in xyz space. Imagine this with a million pixels with different color values and different possible positions. I always start from one source and handle the information through a feedback system. I use noise to apply different forces, speeds, and life. It’s interesting how complex algorithms appear to be random, but they are not. So, since the generative system is “alive”, the agents follow one of the possible paths. For me it is a great metaphor in which everything has already been said.

But listening to @pindar and interesting thought occurred to me.  @pindar, considering how you are trying to program your robot to have your creativity and think it is possible, do you take this to the logical conclusion all creative processes to be generative?


Yes I believe all creative processes are generative, so that would put almost everything classified as art in the generative category.  And I am serious, hear me out..

Artists typically gain popularity when they achieve a style that is appealing, and they start reliably repeating the style in their work. Think about how you can identify a van Gogh you have never seen before as a van Gogh. It has the van Gogh look. van Gogh had a process and style that he would repeat over and over again. Sometimes he applied his style to a painting of his shoes, sometimes to a self portrait, and most famously to a starry night.  His process was in his head, and whenever he applied it, he took a scene as input, then applied the van Gogh style to that scene.  For me this is similar to how generative artists put input into their generative systems, and create art from it.

 Looking at your work, I can definitely see the style. You have a great look.

Each of my robots eventually develop their own style. Here are two separate systems, one that works with me called cloudpainter, and another that is completely independent called artonomous.

Robot or human, the steps that go into making more work in that style is the generative system. We all follow patterns… And that is why I am so drawn to your work, @harshitrnnh’s work. Even though it is generated by AI, there is something else there that breaks free from the patterns. Something spiritual. Each of your four work seem to cross over from the mechanic into the unseen. 

Not sure if you can answer this, but ready for a tough question, have you found that there is spirituality in AI?


Very interesting question! In my (limited) understanding and encounters with spirituality, while entities can embody spirituality, there is an alternate aspect of entities being able to evoke a sense of spirituality in people that encounter them. In that sense, I do feel art is spiritual, in the sense that it can allow for the viewer or the audience to have an experience that is spiritual and transcendental, even if momentarily. In that spirit, I totally feel AI art has the ability to have a spiritual impact on it’s audience and my attempt is exactly at creating such art. 

 For example, in the Machinic Situatedness series, where I work with Buddha figures, the attempt is to create a dreamy spread of pixels that resemble Buddha, without creating exact contours, evoking a sense of the machine dreaming of Buddha, and in turn drawing the audience to feel some sense of that too.  


I can see the juxtaposition of the traditional and the modern in your work, which is interesting and apparent in each piece. Is this focus on dualities intentional?


The focus on confusing these dualities is definitely intentional. To a large extent, it’s a direct result of my upbringing in India and in a somewhat traditional family (like a lot of families in India) and then being introduced to computers and digital technology at some point. I’ve grown up with Indian mythological stories and a sense of spirituality embedded in them, which both fascinates me and serves as a large part of my value system and outlook in life. On the other hand, I am equally fascinated and excited by computational technology to allow us to manifest various imaginations, and now to have an opportunity to work with AI which, in some ways, offers us a creative partner, and whose ‘intelligence’ we have the ability to sculpt. I therefore try and combine the traditional visual systems that I’ve grown up around with current technologies I’m fascinated in, thereby reimagining visual traditions with modern day technology, thus actively engaging with it rather than passively preserving it.



And on the question of culture, I can not help but notice that we are three artists, on three continents, with many different influences. But each of us are speaking with our art in the same language. @pindar makes portraits, I make generative abstracts, and @harshitnnh appears to be somewhere in between so to speak.  I have really enjoyed talking with you two and enjoyed learning about your work in this show.  Thank you…


Definitely, thank you as well.


It has been a pleasure…

Author profile

SuperRare is a marketplace to collect and trade unique, single-edition digital artworks.

Leave a Reply