New Markets In The Arts #2: Generative Art Economies
Blockchain technology has allowed us to re-imagine the arts. As a movement, it’s been rife with seemingly disparate narratives: people adopting its cocktail of hashes & cryptography as a Rorschach test for their beliefs. It’s simultaneously the most anarchist, most libertarian, most egalitarian, most socialist, most freeing, most authoritarian technology. I’ve always seen it as a tool to empower creatives. That’s why I got into Bitcoin development in 2013: to help my 14-yr self in 2004 get paid. As a teenager, growing up in South Africa, my games couldn’t be sold online without jumping - like a videogame character - through many hoops.
There’s a lot that has been done to empower creatives. There’s a lot more to explore, especially in exploring new markets in the arts.
In a series of posts, I will explore three broad avenues that currently interest me:
Generative Art Economies.
This is article #2.
Generative Art Economies
Art is everywhere. This hasn’t been more true in our modern era, where anyone, and ultimately any thing could create art. With powerful computers & the proliferation of machine learning, it’s running head-first into art, and boy, is it fun.
Somewhere, in all the latent space of potential art, lies patterns that will move us, and we can task our friendly mechanical friends to go find them.
We’ve seen how powerful AI and machine-learning can be for enlightening us: Professional ‘Go’ players spoke about AlphaGo as if it gifted us a new perspective.
[European champion] Fan Hui thought the move was rather odd. But then he saw its beauty. It’s not a human move. I’ve never seen a human play this move,” he says. “So beautiful.”
Holly Herndon, who co-created (along with Mat Dryhurst), an accompanying AI to create an album together (PROTO), discussed how one should see the advent of AI coming into creativity.
Holly Herndon’s Proto
So, in turn, moving our capacity to explore this latent space of potentiality towards art will bring forth beauty.
Jason Bailey gives an excellent run-down on generative art, its history and why its worth falling in love with. Recommended reading before we delve deeper.
This generative art is from: Manolo, 2018.
When you combine the ability to form economies around generative art, you can empower the art to descend from its adjacent possible space into our world. It all exists, we just have to go find it, or incentivize our machines to do it for us.
In this new space, we’ve seen projects experiment with pre-generated art, auto-generated art, and finally incentivized art.
CryptoPunks
With CryptoPunks, Matt Hall & John Watkinson, pre-generated 10,000 punks and created the first, popular collectibles on Ethereum. Each one is a collectible artwork.
AutoGlyphs
Following on from CryptoPunks, Matt & John, created a follow up project: AutoGlyphs. In this case, the artworks are still pre-genereated (512 of them), BUT, their code that defines their artistic output is embedded on the blockchain. In CryptoPunks, the specific image itself is reference on-chain. With AutoGlyphs, it follows closer to a Sol Lewitt work, where the recreation is part of the experience of the work.
Gener8tive
A similar project to AutoGlyphs, Gener8tive, uses different generative procedures called K-Compositions, to generate its artwork.
Crypko
Crypko follows on earlier successes in the cryptocurrency space with CryptoKitties collectibles. New cats are bred, by combining collectibles together. The new traits have specific pre-defined images that end up as the final image (or cat, if you will). In Crypko, the combination of features isn’t a pre-set combination, but generated through Generative Adversarial Networks (more popularly known as GANs).
In the above variations, the value that gets created is held individually: those who hold a specific collectible can own it, cherish it, sell it, buy more, etc. In the above, there’s usually little value for people to create collective value together. DAOs change this scenario: when it becomes possible to collectively create art together, and in doing so create a rising tide lift all boats.
Initial experiments use these collective organisational structures into manual creation of more digital art. For example, in experiments such as DAOSaka and TrojanDAO, specific commissions are being done, funded through manual decision-making. In the case of DAOSaka, it commissioned works and made a profit from the sale of the works. Subsequent sales will net DAOSaka increasing revenue.
The future would use these concepts to automatically generate art.
Where this start to get mind-blowing is when start combing AI/ML into DAOs. After discussions with Trent McConaghy & Greg McMullen in 2016, we devised early experiments into what Trent called: ”AI ArtDAOs“.
It could be quite diverse: using templates like DAOSaka or TrojanDAO and having a generative art bot BID for commissions to having the internal kernel of the DAO be a generative artist itself. In other words, it generates art, and the DAO around it, helps it to choose what it should sell and how it should gather revenue. Over time, the DAO aims to improve this generative artist.
The original design has an art bot generate collectibles through a DAO that is chosen by buying into the DAO and staking towards a specific art generator. It’s a mouthful, but essentially: participants vote what art the DAO should attempt to create.
This design led Block.Science to model its economics, and went on to win ETHBerlin Hackathon in 2018: thanks to a wonderful team who ran with the blueprint.
Since then, I’ve designed an alternative version that requires less complexity, by having the collectibles minted on a bonding curve itself. More on this kind of collectible economy here.
A great example of such a collective art-finding game that includes a shared reward mechanism is Clovers. In exactly the process of searching the latent adjacent possible space for art, users run simulations of the game of Othello/Reversi to create unique end-states. The most prized ones are end-states of symmetry.
What’s great is that it incorporated modelling in its design as well.
I love that it achieves its goal: by having a collection of people simulate games to find beautiful end-states that THEY appreciate. So, you get what you want: incentivization of art through crypto-economic games.
Curation Markets
Part of the parcel of improving decision making around what a generative artist DAO SHOULD create is by employing some market. In Clovers, it’s direct: whomever chooses to mint or buy a pattern is a direct indication of what the community finds valuable.
There are other components were individual pricing might not be possible and thus some internal curation market could be meaningful. In original Artonomous designs for example, the token bought from the continuous token model (using a bonding curve), is used to curate specific generators that will produce the artwork. In other words: people buy in to vote with their money what artworks the DAO should generate. After that, the artworks find themselves in the market. If they are successful, they receive more rewards.
There’s many ways to plug the components together.
Abraham.ai
All of these components come together in a holy grail: a project called Abraham by Gene Kogan (creation of machine learning for artists (ml4a)). In fact: this whole article is already being summarised by Gene, in his series: “Artist in the Cloud”, containing his own excellent literature review on all the aspects I mention in this article.
What Abraham hopes to add is an additional feature: something that resembles a soul. With previous generative artwork designs, if you know how the software is set up together, you can either know what it will generate before-hand, or in the case of a project like Clovers, generate your own design from constraints and go and mint it into the economy. Generative artwork given the same inputs should deterministically generate the same outputs.
However, what Abraham aims to do differently is having the AI model be decentralized and have it be trained on private data, such that no one individual would sufficiently know what is to be generated.
“To ensure the originality and uniqueness of sampled artworks, a model is required to be irreproducible and unforgeable. That is nobody — inside or outside of the AAA — is able to clone or retrain the same model, nor sample from it externally.
To meet the irreproducibility constraint, the model is trained blindly on crowd-sourced data which is never aggregated, instead remaining private to each of the individual contributors, leaving behind no easy way to recreate the same dataset a second time.
Uniqueness is secured by splitting the model into many pieces which are distributed throughout the network, and held together as a shared secret. To sample from the model, a query propagates through the entire network. Because no individual has the full model, it’s impossible to generate a sample any other way.”
Mind-blowing. A truly autonomous artist whose works we can buy and own, whose artwork is created collectively.
I look forward to what’s coming. The potential is still vast!
Conclusion
By combining the new burgeoning field of generative art and crypto-economics, we will see wholly new artistic experiences grace us. The potential is vast.
It potentially even lends itself to self-referential meta representations of art: using markets as medium itself. If the process of collectively using new markets to experience art, what can we learn from seeing markets as medium itself?
