By Daniel Anderson.
Recent advancements in machine learning algorithms have allowed for the quick production 
of documents, audio and images. AI generated content has garnered attention from both the media and the public for its role in high profile cases. New York lawyer Steven A. Schwartz used ChatGPT to cite falsified cases that the language model had generated. Pablo Xavier’s AI generated image of Pope Francis wearing a puffer jacket went viral on social media. Spongebob characters performing various songs went viral on Tiktok. A video posted by user “bobbylashton” reacting to the AI generation received 3.2 million views and 518.4k likes at the time of writing. With technology such as Adobe Firefly and DALL-E 2 becoming more prevalent; the copyright of the resources AI uses has come into question.
“Artificial intelligence is going to penetrate every element of our lives,”  said Matthew Waddell, a Calgary based digital artist who specialises in AI generated work. Matthew and Laura Anzola used ChatGBT and stable diffusion for their work at Contemporary Calgary. “It’s going to be hard not to use it”
AI content generators create content by drawing from references in databases. According to a paper published by Cornell University, AI models use a form of  “deep neural network” called “convolutional neural networks” inorder to produce these images. A deep neural network is a category of machine learning algorithms aimed to mimic the information processing of the brain. A convolutional neural networks are deep neural networks designed specifically for image processing. The AI “learns” from a database of images curated by its operators. The contents of image databases that A.I models use are unknown. It’s uncertain that no copyrighted material is used to train the AI model. 
“Most of the images that have been fed to databases aren’t under any copyright,” said Laura Anzola. Anzola is a Calgary based artist who uses technology to produce her pieces. Media company Getty Images has filed a lawsuit against Stability AI, the creators of Stable Diffusion, for using Getty’s content to train Stable Diffusion. This was discovered after the AI repeatedly replicated the Getty Images watermark in it’s generations
Due to copyright concerns, content created by AI is already being regulated. Video game digital distributor Steam, owned by Valve, is unwilling to publish games with AI generated content.  Reddit user “potterharry97” had their game rejected from publishing on the steam store due to it containing AI generated content. Valve cites the lack of clear copyright on AI generated assets as the reason why the game was pulled. “As the legal ownership of such AI-generated art is unclear, we cannot ship your game while it contains these AI-generated assets” said Steam support.
Machine learning algorithms have been used to great success before in the past. Sony’s “Spider-Man: Into the Spider-verse” (2018) used a similar machine learning algorithm to current A.I models during the production of the film. The film makers developed a line drawing system to emulate the style of an illustrator. 
 “When you look at traditional CG, a lot of that (emotion) tends to get softened as you’re just moving geometry around," said Justin K. Thompson, Production designer for ImageworksVFX.
ImageworksVFX trained a machine learning algorithm to map out where these line drawings would go on a character's face for the production of  Into the Spider-verse. The machine learned from the studio’s own database of facial expressions for the line work. -The results from the algorithm were hand adjusted by a human animator before release. Into the Spider-verse earned $ 384.3 million USD at the box office.
“We’re still at the point where we still have the distinction of what’s human made and what is AI made” said Anzola.
Copyright, in Canada, is the exclusive legal right to produce, reproduce, publish or perform an original literary artistic, dramatic or musical work. It gives creators the legal right to control how it’s used in order to protect its value. Youtube faces similar copyright controversy from a form of reaction video. These videos consist of a creator watching another creator’s content in its entirety upon its initial viewing. These videos when uploaded are argued to be able to replace the original as a substitute. 
“This is in contrast to those who use portions of another’s work to create something distinct and apart from the original” DarkviperAU, Youtube creator with 1.34 million subscribers. “This group does not take all value from the original therefore what they create does create a substitute for it”
In the US case of Matt Hosseinzadeh v. Ethan Klein and Hila Klein (2017), the US court found that videos can be used if the contents are transformative for the purpose of criticism. It’s yet unclear if AI generations fall into the category of transformative criticism, to be art separate from the original pieces or if it’s able to replace the original work. The U.S copyright office has already determined that AI generations do not have copyright protection.

You may also like

Back to Top