# Beyond AI — Visual Imaginative Synthesizers By [avive.eth](https://paragraph.com/@avive) · 2022-08-09 --- ![Keep it Jungle — A visual Synthesis](https://storage.googleapis.com/papyrus_images/f170ec088aac21e04c08a43a3e2851a01e2cde5d30cd99391ca5469642ed67eb.png) Keep it Jungle — A visual Synthesis _Visual_ — outputs images. _Imaginative_ — able to produce creative output, and in some cases original output. _Synthesizer_ — takes input, preprocessed data together and creates the output. **Intro** --------- Bleeding-edge AI projects such as [DALL-E](https://openai.com/dall-e-2/) and [MidJourney](https://www.midjourney.com/) are all the rage in the modern digital art world. They are best described as concrete software implementations of a new class of systems we should name \*\*Visual Imaginative Synthesizers, \*\*or just **VIS**. This is preferable to calling them Artificial Intelligence Systems (AIs), as our proposed name is more precise regarding what they do, how they do it, and what we can reasonably infer back regarding the characterization of the creation process by an examination of the output by itself. **Methodology** --------------- Not all visual outputs of generative processes are creative, nor all creative visual outputs original, but some can be. We can define \*\*original output \*\*(or even original art) as an output of a creative process that has its own evident **distinct style**. A style that is not as evident or as distinct in the output of other processes. By employing the humanistic activity of ‘art-critiquing’ on these systems’ outputs, we argue that the output is extremely creative and in some cases even original. Now, it is reasonable to infer the existence of an imagination by considering this evidence of creatively-produced, and in some cases highly-original digital images. Original output necessitates creativity which necessitates imagination. Describing these systems as **intelligent** is unnecessary and should be avoided as a similar argument can’t be easily defended. If we agree that intelligence is the ability to make models, and that imagination is a form of intelligence defined as the ability to create imaginative models, then imagination is the intelligence at work here and other forms of intelligence are not that relevant. Describing these systems as **alien** does not get us very far. By alien we typically mean _out of this world_. However, these systems are very earthly. Our evidence suggests that they were created by humans via software on planet earth, and we can readily explain their existence on earth without having to appear to us from an alien world. Describing these systems as **artificial** is unnecessary if we call them synthesizers, because all synthesizers (besides biological ones that exist in organic lifeforms) are artificial and man-made. **Visual Imaginative Synthesizers** ----------------------------------- We should call these systems synthesizers because they operate just like a synthesizer and produce output just like a synthesizer, so we basically just use a _duck-typing_ line of argumentation. A synthesizer is a system that generates outputs using provided input from various input sources, adjustable process parameters values, and typically a random noise generator. Note that the input source can be an AI system or another VIS, so no human in the loop is needed. AIs using synthesizers to create original art, oh-boy…. Music synthesizers are not necessarily imaginative in the sense that they are mostly specifically designed to follow quite closely human operator prompts, even though some utilize random noise generators as source material to process and build upon. In fact, most such synthesizers are basically ‘just’ musical drones that follow a human operator tightly or loosely but still follow it as the main input to consider in synthesis. At least when on default settings. In the end of the day, they are designed to be studio or live music instruments. **So what the hack is a VIS?** ------------------------------ A visual imaginative synthesizer (VIS) composes and outputs a set of new original images from: 1. An input set of pre-processed input visual images (upper-bounded by all existing digital images created up to some recent point of time before composition-time). The set is typically the training data set used for the neural network component of the synthesizer. 2. A set of **descriptive visual-art history image intents** such as: * Scene content such as background, objects, people, shape, foreground, frame. * Paper type, colors and hues. * Movement, symmetry, perspective. * Paint brush type . e.g. oil painting, crayons, pencil, water colors. * Artistic style — an established artist’s style, a historical art-school style such as pop-art, constructivism, surrealism or renaissance. * Rendering style such as flat or three dimensional. * A reference partial visual image — typically a PNG with one or more transparent regions * Rendering resolution. * Output aspect ratio such as 16:9, 4:3. **Playing a visual synthesizer** -------------------------------- The image intents are provided to the synthesizer as **a prompt** that is generated in a creative process now called **prompt programming**. The programming can be done by a human operator — a visual designer or an artist, or by an AI system designed to output prompts. These prompt are basically written in a new kind of _a natural-language derived programming language_ which has a specific dialect for each different VIS implementation. For example, DALLE wants you to add _Digital Art_ to your prompt if you want your image to be digital art and doesn’t support aspect ratios, while MidJourney is only designed to create digital art and supports many different aspect ratio in its prompts. --- *Originally published on [avive.eth](https://paragraph.com/@avive/beyond-ai-visual-imaginative-synthesizers)*