I Recreated Amazon’s AI ‘Fallout’ Show Art In 10 Minutes, Which Is A Problem

Amazon, despite having a promising-sounding live-action Fallout series on their hands, stumbled into controversy this week when they debuted the very first promo art for the project, an image with an AI-generated background.

This comes in the midst of the Writers and Actors Guild strikes where AI is a front and center issue. Not in terms of artwork, perhaps, but in terms of AI scripts and AI-generated performances. But artists are part of production, too.

There was at least a little debate about the AI image, first whether it was AI (it is, very clearly) or whether that was okay. But I think the main issue people had here was that not only was it using AI, which employs questionable copyright already, but this is Amazon, a trillion dollar corporation that couldn’t pay one artist some amount of money to make one promo image for this enormous show not using AI.

The problem is that studios are seeing just how easy this is to do, how many people don’t notice, and are running with it. But the argument that “AI art is hard too” just is not true, as someone who has used these tools and in this case, recreated essentially this exact advertisement in 10 minutes. 20 total, as half the time was spent just trying to find similar fonts and getting the colors right, non-AI stuff.

This is the image I came up with, which actually has a lot less screwy AI stuff than the original one, which involved three legged women and windowless buildings. Though, hilariously, both of ours have a red car in the foreground that looks like it was designed backward.

Again, this took…literally four sets of Midjourney prompts. What I ended up with was:

“1950s los angeles street, 1950s cartoon, postcard, advertisement, washed out, faded color, idyllic, —ar 16:9”

I don’t know what the original Amazon AI person put in, but it had to be pretty close to that. Then I applied a transparent “grime” filter in Photoshop to give it a weathered look, and that was it.

You can see how this is problematic, even if this is one piece of art for one project. It’s the fact that it was done at all by such a massive studio, and literally anyone could have made this in a few minutes. I am neither an artist nor a graphic designer and I was able to clone this with even a cursory knowledge of how Midjourney works.

I don’t know if the Writers and Actors Strike would cover something like this, but the lessons here apply broadly. The whole point of AI is corner cutting and eliminating jobs or reducing pay because it is “easier” to do it with AI. And yet from the studio perspective, there is no reason not to do this because of the time and cost savings, hence why something like this would have to be written into a contract. There have not been any definitive lawsuits saying something is or is not violating some sort of copyright policy with employing AI art, which is all they’d care about. Although there may be a reverse UNO here where you cannot copyright AI-generated products themselves, which could spell trouble for studio plans.

All of this is leading somewhere bad, and unless it’s stopped early, it’s going to change the entire industry in not-great ways. Amazon still has not responded to request for comment about the image in question.

Source link