stencilblog

[stencilblog]

The very first pictures by the stencil

To start, I decided to use basic geometric shapes (rectangles so far) to build the pictures from. Each rectangle has 9 attributes: X and Y coordinates (on a 100x100 matrix), height, width, angle and hue, saturation, brightness and opacity for color.

At this point, Stencil is basically a random generator with manually adjustable min and max values to optimize the attributes. The program is also capable of visualizing the results in .svg files.

I’m not sure if it counts as an ‘Artificial Intelligence’ yet, but it’s more than enough to start with. Random has an important role in almost all abstract art, but controlling it properly is essential. Consequent and conscious selection means taste and style. Teaching these skills to the AI is one of the main goals of this project.

Here’s a compilation of pictures created with 10 rectangles on default settings.  (x: 1-100; y: 1-100; heigth: 1-100; width: 1-100; angle: 1-360; hue: 1-360; brightness: 0-100; saturation: 0-100; opacity: 0-100)

(Pictures were selected selected out of ~500)

def_01

def_02

def_03

def_04 

These pieces show a consequent style, but the next step should be playing around with the values manually, to see, how changing them can affect the pictures radically and result in whole dirrefent styles.

(x: 1-100; y: 1-100; heigth: 40-200; width: 5-10; angle: 0-0; hue: 1-360; brightness: 0-100; saturation: 0-0; opacity: 0-100)

But the real goal is to make the AI (based on human feedbacks) learn basic interralations between values, to optimize attributions and algorythms by itself.

For a simple example: the bigger the rectangle is, the smaller the opacity should be. (To avoid too large surfaces covering the picture.) Color harmonies are also easy to express by mathematical phrases by limiting the hue scale to certain sections. Composition is an extremely important and complicated factor, even in this abstract, geometrical situation. I wonder how the computer will try to draw up ‘rules’ according to different human feedbacks.

We are figuring out how the feedback platform should be built, so everybody can get involved in teaching the neural network.

K

[stencilblog]

AI Art examples and thoughts

It was quite obvious that I’m not the first one to come up with the idea of art created by Artifical Intelligence. So I did an online research and found many inspiring art projects in connection with AI. My main source of information is https://aiartists.org/, a great site dedicated to different AI art projects.

Sougwen Chung is an award-winning artist who uses hand and technologically-reproduced marks to explore communication between people and machines.  https://aiartists.org/sougwen-chung

It seems that there are two major way to AI art experiments. Some artists focus on some kind of human-AI collaboration to help their own art, and others try to make the AI creative by itself. Both way seems interesting, but I’m definetely excited about the second one. 

A common way of making the machine self-sufficing is building a database of actual photos and paintings to “get the visual inspiration” from. In this case, the AI is trying to mimic and mix the parts of actual pictures.

Robbie Barrat: AI Generated Nude Portraits 

Using an implementation of Progressive Growing of GANs and a corpus of thousands of nude portraits scraped primarily from WikiArt, a neural network was trained to create nude portraits. The resulting portraits are not realistic. The machine failed to learn all of the proper attributes found in nude portraits and instead has fallen into a local minima where it generates surreal blobs of flesh. Is this how machines see people? https://aiartists.org/robbie-barrat

David Young: Flowers AI / Machine Learning generated images. 2018-2019.

Using AI & machine learning the computer was trained with photographs that Young took of flowers at his farm in Bovina, NY during the summer of 2018. It then generated its own images. Young notes that what emerges is not accurate, but the work isn’t asking for accuracy — it’s asking for the machine to build its own unique vision of the natural world. https://aiartists.org/david-young

But in these cases, the AI is still not entirely independent, because the building units of the final artificial product are coming from a human source.

In my opinion, the AI art should be much closer to abstract, which means the visual blocks the machine uses to construct the masterpiece from, should also be abstract. Like simple or amorph vectorgraphic shapes.

The most inspiring examples I’ve found are the works of Tom White:

"My work focuses on how machines see the world. I have created a drawing system that allows neural networks to produce abstract ink prints that reveal their visual concepts. Surprisingly, these prints are recognized not only by the neural networks that created them, but also universally across most AI systems which have been trained to recognize the same objects. Neural networks have been trained to recognize thousands of individual items from everyday life, such as “Rabbit”, “Banana”, or “Killer Whale”. Using the drawing system I have provided, the neural networks directly express in simple ink drawings their own versions of these categories – creating abstract shapes that convey their understanding of the world.“ https://aiartists.org/tom-white

Tom White: Cello  https://www.instagram.com/dribnet/

Although I find very excited about how these pictures represent different objects from the real world, my personal goal is to create non-figurative, constructivist, suprematist pieces.

The very first visual achievements of the Stencil project are arriving.

K

[stencilblog]

Human Art Inspirations

In these early stages, it seems impossible to separate my personal taste from the AI style development, so I think the best idea is to not even try at this point.

So I’ve been collecting a bunch of pictures from actual artists, which I found useful for the project. The most inspiring and suitable style seems to be suprematism. It is an art movement from the 20th century focused on using basic geometric forms, such as circles, squares, lines, and rectangles. I find these pictures extremely immersive and relatively easy to reproduce by the AI.

I’m especially interested in the art of László Moholy-Nagy, Lazar El Lissitzky and Kazimir Malevich. Moholy-Nagy:

Lazar El Lissitzky:

Kazimir Malevich:

I’ve been also looking for talented contemporary artists interested in suprematism:

https://www.instagram.com/kolorama_studio/ https://www.instagram.com/pules.art/ https://www.instagram.com/delrayjack/ https://www.instagram.com/yshiddi/ Please help me finding more and more inspiring pictures and artists and feel free to share your thoughts.

The following post is going to be a quick research in connection with similar AI art examples. Later, K

[stencilblog]

Can art be created by Artificial Intelligence?

Stencil is a starting media art project experimenting with digital fine art created by a custom AI. Our goal is to find and break the limits of optimizing artificial sensibility to art, style and taste, making Stencil an evolving, naive and mostly “open-minded” constuctivism artist.

Stencil also means an open and online community in which members are taking part in the developement process by giving feedbacks, sharing ideas, discussing problems and solutions. In the near future, every interested will be capable of using the software and create pictures individially. We are also going to create platforms to share and discuss consequencies. The Instagram profile is also under construction, but already avaible on http://instagram.com/stencil_ai . The common hashtag for the project is #stencil_ai

This Tumblr blog is the main platform for discussing difficulties and achievements in connection with development. I’ll try my best to keep the feed up-to-date with actual questions, possible answers and illustrations.

One of the main goals is to make this project open and viral, so please feel free to write comments and give feedbacks every possible way. Later, K