BitDepth#1387
MARK LYNDERSAY
IN THE last six months, the art world and photographers have been tossed into the spin wash of artificial intelligence (AI) technology with the emergence of multiple computing clusters that turn text and image input into output that looks very much like art.
There are two main ways this is being done across a growing number of image-generation portals.
In some, you enter a text prompt, a description, collated keyword style, of what the image should contain or be influenced by. That might include locations, art styles, popular people and other descriptions that collectively influence what these computers synthesise.
In others, you upload an image and the software goes freestyle on it.
I opened a half-dozen of these tools, many of which demand payment up-front, and ended up trying the popular MidJourney portal and Lensa app.
Lensa, from the company that created a briefly popular prism art filter app a couple of years ago, is a general purpose, AI-powered image enhancement tool. For a fee, it will also upload a batch of your selfies to an AI server to create fanciful avatars.
Three hundred photos and US$7 later, my collection is an agonising gallery of all the ways that AI can get human features and limbs wrong. For some reason, Lensa feels rather overwhelmingly that I should also wear open-collar suits and a variety of hats, from baseball caps to smart looking fedoras.
They are all appalling in one way or another.
MidJourney is accessible through bots running on a Discord channel. When you sign up, you get 25 free chances to create something digitally amazing.
As with Lensa, I tossed various photos of myself at the software with a range of prompts and the results were equally unusable, but more interestingly so.
MidJourney works on the submitted image using the text and from my results, there is no question that understanding prompt text is crucial to AI art.
I asked two local designers and artists about work they have produced using AI.
Ian Reid, art director of Reid Designs, is refining his prompting.
"You will no longer need to pay a mas designer to render costume ideas," Reid explained. "I will do that for you by being the prompt engineer. Right now I am learning everything there is to know about prompting so I can get what is requested.
"I am interested in rendering images that clients request that we can't find in [stock image agency] Shutterstock. Young Indian woman from Trinidad using a phone, for example. My goal is to see if I can slot into a niche where I can provide a service to clients who can't find a solution anywhere else."
But where do these new images come from? Image AI computing clusters are trained by analysing a selection of images submitted to it. The most powerful AI tools have been trained by scraping publicly available images across the internet, assimilating them by the billions.
It's here that the my
