Table of Contents

About

The following namespace contains some original designs created by Wizardry and Steamworks using various online AIs. We consider the idea to be the art whilst the rendition is up to the generator.

Organization

The gallery is accompanied with a data-card deck in order to keep track of the art being rendered such that each rendition shall have a metadata associated with it.

Identifying an artwork is done by the invariant NFT defined as:

\begin{eqnarray*} H_{t} &=& hash(q) \\ \end{eqnarray*}

where:

called the thematic NFT, and the perceptual NFT, defined as:

\begin{eqnarray*} H_{p} &=& pHash(I) \\ \end{eqnarray*}

where:

This leads to a tree-like topology where an artwork is localized first by its thematic hash $H_{t}$ and then sub-sorted by its perceptual hash $H_{p}$. Whilst the thematic hash $H_{t}$ will localize the conceptual idea behind the AI-render, which might be unique to an artist, the perceptual hash $H_{p}$ will be able to identify variants of the artwork generated by the AI engine based on the unique thematic hash $H_{t}$.

It is then easy to:

NFTs then become an indicator of uniqueness (and symmetrically similitude) where the value of each individual artwork is indexed by its thematic hash $H_{t}$ (which we consider to be the only original contribution of the artist) with the rendered variants as possible branches of perceptual hashes $H_{p}$.

Note that or simplicity, the thematic hash $H_{t}$ could be extended, perhaps as:

\begin{eqnarray*} H_{t} &=& hash(q) ^ p\\ \end{eqnarray*}

where:

in case the AI generator presents additional options on how to render the art; but for all intents and purposes, the parameter $p$ should be absorbed in the context of $H_{t}$ in order to

One can observe that $H_{p}$ falls within the context of a given AI-generator, such that the only context-free uniqueness lies within $H_{t}$ that becomes an "artistic invariant" within the terms of definition of AI-assisted art.

It is worthwhile to notice that in that case, the extra input parameters $p$ are not context-free, but rather seem bound in the context of both $H_{p}$, as a property of the AI generator, as well as within the context of $H_{t}$ as a thematic idea of the author. In order to resolve the issue, the extra input parameters $p$ should be always absorbed within the context of $H_{t}$ thereby making the input as part of the idea. For instance, instead of using, say, the "graffiti" tickbox of the AI generator, an author should preferably make the "graffiti" style part of the input query, ie: "draw me a cat using graffiti style rendering". Naturally, "graffiti style" would still have the locality of the AI generator but it has already been defined that $H_{p}$ is local to the AI generator itself, such that the term "graffiti" and whatever it implies will just make all resulting $H_{p}$ local to the AI engine (as it should) whilst still preserving the uniqueness of the artist as a choice of style.

There would be some interesting applications; for instance, in legislation terms, yet analytically, one could define a license that allows people to reuse some artwork as long as the derivative work is a certain similitude close or away depending on the wishes of the original author (this covers, for example, instances of stock photos that are not supposed to be resold as-is, but rather as part of a derivative work).

Index