Your recent social feeds have probably been flooded with images generated by AI systems like DALL-E , Midjourney and Stable Diffusion.
When these systems first made their debut, there was shock, awe and fear as creative practitioners questioned if they would soon be replaced.
(Also read our great story on the implications of cheaper, faster images)
Ayaz Basrai, co-founder The Busride Studio, has been experimenting with both DALL-E and Midjourney and suggests an interesting mental switch for creators – from AI to IA, or Artificially Intelligent to Intelligently Augmented.
Says Basrai, “Exposing myself to Midjourney made my focus sharper and forced me to
question what I could do better. Producing recombinant, stunning and tangential output is now super simple.”
How ‘Directable’ is the AI?
For commercial use of these systems to become a reality, they must be able to interpret and fulfil prompts far more accurately than they do now.
It took designers Anusha Nawalgaria and Ritik Sareen more than a hundred prompts to create the images for the packaging of Dragão, a fictional ‘beirao and liquor chocolate brand handmade in Portugal.’
History teacher Samir Krishnamurti and designer Anoopa John tried using DALL-E to generate images to accompany Krishnamurti’s lessons. They focused on periods with very little recorded visual history like the reign of Alexander the Great. While DALL-E generated familiar images of Alexander atop his horse, it blocked prompts with the word ‘revolution,’ saying they violated policy.
While they are far from perfect, even today the AI systems offer capabilities that can be used to generate say, a storyboard in seconds. LA-based director, Paul Trillo’s 30 second video for fashion designer, Shyama Golden, showcases hundreds of outfits and was created using DALL-E.
Stable Diffusion: The Open-Source Game Changer
In mid-August, AI tools company, Stability.ai released a sophisticated AI model called Stable Diffusion, which is available to the public through a safety-centric, open-source license. Immediately developers started building their own apps on Stable Diffusion’s backbone – the popular Y Combinator-backed Photosonic is said to be one of them.
Stability.ai also launched a site called DreamStudio, which can be used to generate 200 images for free, after which you pay $10 for the next 1000 images. Founder EMostaque tweeted that they would soon offer audio, 3D and video!
Our Future Aesthetic
There may be another implication that no one is talking about yet. Basrai points out that images from these platforms embody an other‑worldly, mostly symmetrical, fractal‑like quality. As their use increases, this aesthetic is bound to find its way into our art, fashion, architecture and design, much as Instagram and Pinterest influenced the now ubiquitous, pastel, minimalist look.
The fact that these AI systems are biased depending on the datasets on which they are trained, is a whole other story.
The Existential Question
Basrai says laughingly that he was looking for ways to have fun with prompt and “For anyone growing up in the 80’s, the existential question ‘Choli ke peeche kya hai?’ has assumed a Neil DeGrasse Tyson sort of gravitas.”
His first prompt to Midjourney was, “Who is behind the Indian dancer on a dark stage?”
Successive iterations revealed that it was in fact, Bruce Wayne as Batman ; Dressed up for a Navratri event, but Batman nonetheless.