Why Is Corridor Digital's So-called 'AI Anime' Video So Controversial? Here's How The 'Rock Paper Scissors' Animation Became One Of The Biggest Viral Debates Surrounding AI
"Did we just change animation forever?" read the clickbaity title of a recent and controversial video by digital effects studio Corridor Digital. Many online literally and figuratively replied, "I sure hope not."
Two weeks ago, Corridor released what they billed as an 'AI Anime' short film, boasting in subsequent videos that they'd used numerous AI tools to animate an overly dramatic game of Rock, Paper, Scissors.
The video struck at the heart of the AI art debate that's been raging for months as artificial intelligence appears to be nearing the capacity to create work comparable to that of a mostly unpolished (or mediocre at best) artist, and it attracted very vocal critics and defenders.
Here, we'll run through just what made Corridor's short so controversial and what the larger debate is revolved around.
Corridor's short film first attracted a wave of critics concerned largely with its quality and ethical implications.
"ANIME ROCK PAPER SCISSORS" was created by using Stable Diffusion, Dreambooth and Davinci Resolve to animate over real actors playing the scene in front of a green-screen background, similar to how rotoscope films are made. The various AI tools were fed screenshots of the anime Vampire Hunter D: Bloodlust in order to make the style consistent, and the backgrounds were developed in Unreal Engine.
The resulting film crossed into the uncanny valley for many viewers, and quickly, those with an eye for animation were able to identify why: The film is rife with basic animation errors.
At LEAST check if the damn eyes colour is consistent pic.twitter.com/YhLlXoOD8H
— ⭐️Emma Marie⭐️ (@_Emma___Marie_) March 1, 2023
From frame to frame, the characters' features seem to change nearly constantly, while their hands are often inhuman blobs. Characters' emotions don't seem to read on their faces, and their intense screaming match (typical of an anime battle) finds their mouths open eerily too wide.
While mistakes like this may be expected from a tech demo showing the current capabilities of AI animation, others took issue with the murky ethics behind the video's creation. As mentioned before, the video was trained on Vampire Hunter D: Bloodlust, and while the video itself may fall under the safety of free-use on the basis of being a sort of fan art, if another animator wanted to pull a similar trick for money, they would certainly be opening themselves up to copyright lawsuits.
Others noted that should AI evolve to the point where it can make less noticeably terrible animation, major studios may look at AI animation as a cost-effective measure to get past hiring entry-level animators for gruntwork jobs (or even more expensive experts). YouTuber Geoff Thew noted that a corporation like Disney could get around the copyright issue by feeding an AI its vast catalog of IP, allowing an AI to emulate a "Disney style" in a hypothetical future film instead of a real animator.
All of these arguments are familiar to anyone who's been a part of the AI art debate that's raged in several fields over the past year, boiling down to a single point: Not only does AI art arguably look worse than human-made work, the dearth of quality will likely mean little to executives looking to maximize profits, and soon consumers will get worse products while actual animators get fewer jobs.
The flip side of the debate came from people largely interested in the potential Corridor's experiment could represent. While even Corridor's most ardent defenders acknowledged that the animation's quality wasn't up to snuff, they contended that it was a matter of time before the technology "got there," so to speak, and it could be game-changing in certain ways.
Some argued that the technology could "decentralize" the animation industry, meaning AI could help small creators create productions on the level of Disney at a fraction of the cost. Some of those small creators also voiced that competent AI would be a boon to them, especially since they can't hire the large teams required to make quality animation currently.
As for the use of training the AI on existing work (Vampire Hunter D: Bloodlust in this instance), the ethical debate surrounding this is far from hashed out as to whether this falls into plagiarism or not.
Defenders of this point to things like artists covering songs, people mimicking the style of a successful artist without completely copying them (such as a painter learning cubism without stealing the work of Pablo Picasso and Georges Braque), as well as if this fell into parody territory (which is a whole 'nother debacle). In the end though, and completely unsurprisingly, nobody could seem to agree on this point either.
As It Stands Now
Both the defenses and criticisms of Corridor's short hinge on a future where AI one day does have the capability to make a competent animation, and while AI's capabilities did grow from "terrible" to "nearly there" quickly, AI has been stuck at "nearly there" for a significant period of time.
Some defenders give AI a year or two to craft quality animation, while others contend it will never get there because AI is not capable of an original idea. It can only approximate a human being's IP to the best of its abilities, and, if Corridor's "ANIME ROCK PAPER SCISSORS" is anything to go by, it arguably can't recreate the emotion, artistry or soul of that living person. At least, not yet.
For more information, check out the Know Your Meme entry for Corridor's AI Anime Controversy.
Additional comments have been disabled.