I'm glad we are having this discussions. From my understanding, ‘AI Art’ can be reframed as ‘Art generated by Applied Statistics’. It uses statistics to create an image result that is most likely to please the largest amount of users. Hence, In my experience, it is unlikely to challenge biases. It will, more often than not, rely on stereotypes and generalisations for it’s process. When I input ‘Fierce Female Buddhist Deity’, I receive results that, despite my wish, clearly show some variation of friendly Guan Yin, with soft porcelain features, maybe with red Vedic-style clothing, or at maximum flames present (see result below). Even when I ask for ‘bone ornaments’, ‘human skin garments’, ‘wrathful enlightened demeanour’, some censor in the system prevent the full manifestations these.
It seems that the algorithm has made up it’s mind that ‘Buddhist = Peaceful’, ‘Goddess = Stereotypically Attractive’. And the data in the statistics it uses to make that assumption, come from the Internet’s own exportation of the unfair biases in our global culture. It is often a mirror of our culture’s most dualistic, limiting and consumer-friendly side. Both Puritanical and Exploitive forces are deeply seeded in the visual culture of the web. I believe advertising trends have a huge role to play in this.
In order to receive better results, just like in all art, many try to work against normativity. I have generated results that closer resonate with my perceptions of the Dakini, by trying hard to use prompts against the AI's assumptions of what 'belongs together'. Knowing that the straight-forward prompt will only generate exotified and typified results, I attempt to confuse it with unexpected and odd elements in the language of the prompt.
For example, I draw on other artists who have some 'fierce feminine' elements in their work. If AI is, inevitably, a thief, then I try to tell it what I myself have tried to emmulate in my art. the prompt ‘ferocious being from a surreal scary dream, angry grin, Cthonian, Lina Iris Viktor colour scheme, Byzantine Icon of a Vajrayana Chod Charnel Ground Practitioner, Toshio Matsumoto historical film still of statue with fangs, Sevdaliza Music Video in which a dakini portrays Julia Ducournau’s ‘Raw’’ creates something only marginally better than usual (result below). It looks like an occult movie poster, but shows me a new direction to explore and challenge. All in all, it is a strange and often frustrating tool, but like any attempt to change a cultural bias, worth the effort of those who pursue it.
My visual taste is as much a product of the cultural market as anyone else's, so an aim in this investigation is to question that too at every step. At the same time, in the history of Buddhist art, many instances can be found when images popular in a non-Buddhist cultures, such as the depictions of Vedic and Bonpo gods, are assimilated into Buddhism, so that their Iconic status in people's minds can be subverted into a tool of Dharma. Popular Forms and Symbols from of the time were used towards the hopeful end of moving beyond all Form, beyond all Symbols. Maybe AI, the tool that has the strongest grasp of which images are holding control over our cultural consciousness right now, will help us not be trapped in the limits of the past, but fold the chaotic modern world into the Dharma too.