Beyond AI
The Three Realms
Human imagination is limited. Even dreams, which are made from fragments of our memories, are limited. If we could head out into the infinite sea of possibility, eventually we would reach another limit: the limit of human comprehension. In between these two limits lies an unknown, but potentially knowable, realm. Let’s call it Realm Ψ.
Artificial Intelligence
Many truly beautiful artworks have been created with AI. However, AI relies on training with data from humans, so the same human limit applies.
The Paradox
Computers have no imagination whatsoever, and so, paradoxically, computer algorithms can access Realm Ψ. In fact, there are many examples of extraordinary beauty (which had not been imagined previously) generated by computers, such as the Mandelbrot Set.
The Visual Evolution Engine
This is the rationale behind the Beyond AI Project—to allow computer algorithms to run free, and see where they lead. Outside the house, they must be accompanied by a responsible adult. Obviously.
Beyond AI is powered by the Visual Evolution Engine software (VEE). The heart of VEE is successive operations with two-dimensional Fourier Transforms of images, and there are a number of software add-ons:
- Wavelet operations,
- Adaptive Layered Image Synthesis (ALISA) used to create figures,
- Cyclic Nonlinear Desaturation (CYNDE) which generates nonlinear effects in images, and
- Ultracubism, which embodies a 3D effect.
As well, external Stable Diffusion software is used to enhance the textures in images. (The calculations are performed locally on a high speed graphics card.) This is part of AI, but the role of the prompt text is almost negligible, and accessing Realm Ψ is not at risk.
In addition, various photographs may be combined to produce a particular initial image prior to calculation. In many cases, the edges of such images are simply cut out at the pixel level and the match with the background is satisfactory. However, with foliage and undergrowth, it’s sometimes necessary to fill a narrow band between initial photographs using standard AI to achieve a seamless transition. Once again, originality and Realm Ψ are not at risk.
Finally, many hundreds of Photoshop™ operations are typically performed, mostly automated, but there is also a great deal of creative input to generate a final artwork.
