Intelligent Artifice

Learned Color Spectacular 2022 #5

Finding the exact phrase a generative a.i. engine needs to produce something interesting, much less thought-provoking, takes some work. Put an adjective out of place or use a word the program doesn't recognize and the results might be, well, less than stunning. But the engines are getting smarter and the technology is getting better. Brace yourself, because a.i. is going to change a lot of things in the next couple of decades.

Ready or Not

What do we mean by generative a.i.? What are we really talking about? It’s easier to describe the current state of a.i. by what it’s not: it’s not Mr. Data, nor is it Hal 9000. It’s not even Bomb #20¹. What it is, or, more properly, what they are, are very sophisticated programs designed to take a lot of information, learn from it, and then use that learned information to complete specified tasks. But it’s easier to show rather than tell and we can do that by starting with this list of prompts:

• a red dragon in the style of alphonse mucha

• Art Deco snoopy

• high quality photo of a buddha smoking a hookah

• a hand drawn sketch of a schooner

• A synthwave style conquistador on a donkey above the reflecting water of the sea, digital art

• 60s surf style poster of a rhino

• 60s retro rocket launch

• a spaceship orbiting saturn in the style of John Berkey

• realistic antique watercolor painting of an artificial intelligence

Each of them is a a simple string of descriptors surrounding a basic noun. So, in the first one, we have the noun "dragon" and the descriptors "red" and "style of Alphonse Mucha." The a.i. engine will then take this string and begin to draw that picture². It does so by first learning what all those words mean by analyzing millions of images that correspond to those words in some way, and then creating an image until it succeeds in creating one that matches the criteria set by the prompt. How well it does so...well, that's where the fun comes in.