Artist in the Cloud

Towards the summit of AI, art, and autonomy

Gene Kogan
Jul 16 · 9 min read

Abraham is an open project to make an autonomous artificial artist, a crowd-sourced AI that generates art.

This article, the first in a 4-part series, gives an overview of the basic idea. Follow-up articles will examine each of the technical components in greater depth. The full series is as follows:

Artist in the Cloud

• The Spirit of Decentralization — Why decentralized organizations are more than the sum of their parts (eta early-August)

• The Collective Imagination — How a machine shows us what it means to be human (eta late-August)

• A Path Towards Genesis — An agenda and timeline for the Abraham project (eta September)

Over the past few years, I’ve been giving workshops about machine learning for art and creativity, and building an open-source toolkit for it called ml4a. As I learned about decentralized AI, I gradually became inspired by the idea of an autonomous artificial artist, a sovereign creative spirit who generates original art.

This idea builds on top of promising techniques from machine learning, cryptoeconomics, and computer art. It is a logical progression from decades of research by artists, scientists, and philosophers, contemplating the nature of creativity and consciousness, and their relationship to technology.

This introductory article is an attempt to make a case for why such a construct is interesting, and to initiate an open project to study the relevant pieces and make one.

The essential idea

The goal of the project is to build an AI which autonomously creates unique and original art. We call this an autonomous artificial artist (AAA).

By autonomous we mean that an AAA demonstrates its own agency or will, independent from that of its creators. Even the most sophisticated AI which lacks autonomy is no more creative than the dummy on the arm of the ventriloquist.

By original, we mean that it exhibits a creativity truly of its own, not simply copying from another.

By unique, we mean that although its artworks may be copied, its creativity can’t be replicated elsewhere. One may always record a Beethoven piece but can never compose like Beethoven.

To satisfy the criteria of autonomy, originality, and uniqueness, we propose to make a generative art program under the following set of social and technical constraints.

To start, the program is learned from the collective input of a decentralized group of actors who crowd-source data, code, curation, and governance to the AAA. In order to coordinate them, a decentralized autonomous organization (DAO) is formed which is open to anyone. Decentralization prevents any one party from exerting too much influence on the AAA, as that would compromise its originality and autonomy.

The core mechanism of the art-making program is one or multiple generative models (we assume for now a generative models of images, but it generalizes to audio, text, or other types of data). A model’s architecture and training data are variable and crowd-sourced. It may be in continual training, fine-tuning as new data comes to it, possibly versioned and branched into different models as it evolves. The artworks are samples from the model, or possibly the models themselves.

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.

The setup depicted above conflicts with the privacy requirement if it is trained in the same configuration without splitting or encrypting the data. Additionally, it assumes only rational actors and does not take into account various attacks. These are some of the open problems we will need to solve.

The effect of these mechanisms is the distribution of influence over the AAA. Its behavior is decoupled from any individual actor, giving it the appearance of acting upon its own internal agency. We contend that what emerges from this ideal constitutes true autonomous creativity.

Applications and economics

With the core art-making apparatus laid out, we turn to designing a financial model to sustain Abraham and briefly discuss some secondary applications.

An AAA has a balance of complementary economic activities: value is created in the form of artworks, and value is consumed in the form of data, compute, and the labor that goes into coding, governing, and curating it.

To take advantage of this, we propose to adopt the ArtDAO idea: that the AAA itself owns the artworks it produces and may sell them to patrons on an art market. At the same time, the AAA uses the proceeds to pay for the resources it consumes, remunerating the people who build and feed it.

Once a prototype is deployed, we can consider some secondary applications which interact with Abraham. For instance, there could be a client application which queries and buys artworks from the AAA, or interfaces it with existing art markets. Mechanisms for curating data to models and artworks to collectors will be useful as well.

There are multiple ways to regulate these activities, some of which require a native crypto-token, and some which can be bootstrapped from existing cryptocurrencies. If we issue our own token specific to Abraham, this opens up the possibility of using bonding curves, curation markets, and other specialized token systems.

These applications will be covered in more detail in part 4 of this series, along with an initial agenda for the project.


Imagine a computer whose logic gates have been pulled apart and scattered throughout the world, like a meteorite breaking up into dust as it makes contact with the atmosphere. It is held together by millions of threads in a peer-to-peer network which has self-organized over the shared purpose of running a program that channels their accumulated creativity.

I find this incorporeal generative art program, dissolved into the fabric of the internet — an artist in the cloud — to be beautiful and ethereal.

Even more compelling is that it captures something uncannily familiar to us: our collective imagination as learned from our private data, shaped by the uncountably many micro-interactions each of us have with it. Like coral reefs, ant colonies, and other natural superorganisms, a distinct character or personality emerges atop the group which transcends the individuals. It is our “hive mind,” evoking a shared consciousness, if only a primitive form of it.

In contrast to depictions of AI as alien or oppositional to us, this project is a deeply humanistic one, conceiving of AI as a vehicle for the collective intelligence and creativity of people. This idea of wisdom emerging from collective action will be explored more in the next article in this series.

More pragmatically, the project is also a testbed for broad new technologies which are more general, giving us a platform to investigate them. It also lets us experiment with promising but untested new ideas in social coordination, finance, and governance.

I am no Utopian; I don’t believe in the intrinsic goodness of any of the technologies in question, nor do I have faith that Abraham will be benevolent. Like all DAOs, Abraham is vulnerable to malevolence, subversion, and greed, all of which will only escalate as the project gains value. But fear of such outcomes is no excuse to renounce the idea, but rather a reason to study it carefully and steer it in the direction we think most beneficial.

On Abraham’s name and identity

“The soul must contain in itself the faculty of relation to God… The correspondence is, in psychological terms, the archetype of a God-image.”

— Carl Jung, 1962

The inspiration for Abraham’s name and identity derives from the Jungian interpretation of religion, mythology, and folklore, which views theological symbols as manifestations of psychological archetypes from the collective unconscious.

This includes the archetype of God, the symbol of being, the universe, and existence — the name of God in the Old Testament, Yahweh, is believed to have derived from the old Hebrew word hawa, which means “to be.” Consciousness is seen as the divine spark, or God within the self. This relation between self and existence is found in many world religions, notably in the Hindu concepts of Brahman and Atman.

To Jung, belief in God signifies the psychological process of individuation, in which a person reaches spiritual wholeness through the recognition of their conscious self (ego) within the larger plane of psychic existence we share, the collective unconscious. In the Abrahamic religions, Abraham is identified as the first person to believe in God. The Abraham project follows as an attempt to recover the “collective imagination,” that component of the collective unconscious which is the source of human creativity. This connection will be elaborated more in part 3 of this article.

Towards Genesis

The alpha version of Abraham, code-named Genesis, does not yet have a target release date.

In its purest form, the idea relies on combining multiple experimental technologies which are the subject of ongoing scientific research and development. As promising as they are, they are also — to varying degrees — unscalable, insecure, difficult to use, and expensive.

For that reason, I propose to start with the following goals first:

A more detailed overview of these goals has been posted to the forum to begin a discussion about them.

If you are interested in participating or tracking our progress, learn how you can do that here.

Future articles

The next three articles in this series will cover in greater depth the individual technologies we’ve touched upon here, as well as connect Abraham to influential precursors and promising contemporary initiatives exploring similar themes.


Thanks to Andreas Refsgaard, Brannon Dorsey, Burak Arikan, David Ha, David Pfau, Dimitri De Jonghe, Elena Mozgovaya, Florence To, Francis Tseng, Hannah Davis, Helena Sarin, Jason Mancuso, Jason Teutsch, Kory Mathewson, Luba Elliott, Mat Dryhurst, Matt Condon, Mike Tyka, Parag Mital, Rachel Uwa, Roelof Pieters, Simon de la Rouviere, Sofia Crespo, Trent McConaghy, Vanessa Rosa, and Xavier Snelgrove for reading early drafts of this article and graciously giving me feedback on it.

Gene Kogan

Written by

programmer, primate.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade