mythographics

Norse Yggdrasil tree with Bifrost rainbow bridge

https://openart.ai/create?ai_model=openart/OpenArt_SDXL

The idea with this article was to see how interesting (or not) the Large Language Models (LLMs) introduced by companies like OpenAI, actually were in creating images of the Jungian archetypes - how efficient LLMs were in representing the universal subconscious - which is kind of what they actually pretended to do: after all these LLMs were derived from the sum total of our knowledge and experience as stored digitally on our Web/Net. If OpenAI and similar companies had succeeded in this task of collecting and collating ALL our digitally stored and archived knowledge - all our histories, all our academic papers, all our encyclopedias, all our newspapers, TV programmes, films, novels, plays, all our art, all our graphics, comics, illustrations etc, then Chat GPT and similar queries and prompts should be able to provide unique syntheses of these Jungian archetypes. So these are some of the results:

Norse god Loki, shape-shifter, trickster god, full-length portrait

Odin and Yggdrasil the World-Tree...

Heimdall Bifrost guardian

Another version of Heimdall, guarding Bifrost - the Rainbow Bridge to Asgard - the home of the Gods..

Another archetype Maiden in the style of R.B. Kitaj

So, this image is nothing like any of Ron Kitaj's work that I have seen (and he's one of my favourite painters from the earlier 1960s - his work is painterly, figurative, based on deep drawing, graphically inventive and explorative. This ChatGPT image has reverted to a comic-strip characterisation - a very disappointing parody of Kitaj's magnificent work - published as prints and paintings in several large monographs... What algorithm is at work here?

Ronald Brooks Kitaj: The Red Banquet 1960

As you can see, the 'real' Kitaj is graphically and painterly inventive, leaves his compositional working processes as integral records of his creative practice, is fond of the blend of painting and drawing

'Maiden' archetype in style of Paolozzi

And the same with this image. Paolozzi was the inventor of 'Pop Art' - in his Bunk! collages of the  late1940s. And in the Sixties he became a master of graphic collage, reproduced in his highly successful serigraph prints by Editions Alecto:

Eduardo Paolozzi: As is When 1965 Paolozzi takes a brilliant intellectual approach - to his sculpture and his graphics - and this series of prints based upon the life and ideas of Ludwig Wittgenstein - deep philosophy at this time.

in the style of Joan Miro

This too is not even a decent pastische of Joan Miro's imagery, though it refers to some elements of Miro's Constellation: The Morning Star of 1940 (below)

Introduction

_{This will be an ongoing series of visual experiments exploring the potential of ChatGPT and other query engines for Large Language Models (LLMs), with my observational and art-critical feedback on their effectiveness in trawling, collating and interpreting  the vast extent of visual content and aesthetic treatments in the LLM. These 'bots' are essentially using mankind's entire visual collection of art, media and graphics - stuff that is our world's cultural heritage, and belongs to all of us - so I think its important that they perform at least on par with the skills of individual human creative talents. On these initial experiments (above), I suspect that they do this rather poorly, and I want to discover why. The provision of Large Language Models and interfaces like Chat GPT require large resources of Data storage and generative-processing - making them carbon-intensive, so I'm not sure that this will be carried much further by me. I don't fly much anymore either.}

_{to be continued...}