The Case for Illustrated Visuals in a Digital Time

The way images are consumed has always been linked with its corresponding media technology. For print media, a sea change was prompted by the iPad in 2010, compelling much of the publishing industry to overhaul its longstanding revenue models. More recently, the primacy of broadcast television has been usurped by online media consumption. In particular, mobile web has surged to lead internet traffic worldwide.

“Visual search and recommendations in Google, Yandex, YouTube, Instagram or Pinterest expose us to endless images and video, while websites of major museums, invite us to browse hundreds of thousands of digitised artworks and historical artefacts.”

Lev Manovich, 2017

The rise of media tech and digital platforms has spurred innumerable visual works made to exist as screen-based image data. And while some artwork may not be natively digital, this too is frequently documented, adjusted, and/or published digitally. Once within the mediasphere it is of little consequence how an image has been created; if it can be viewed on a pixel display, it is subject to the conditions imposed by the hardware and software technologies of the day. Unlike the single artefacts of the 20c., visual culture is now available at a massive scale in digital form.

Adapting to Media Tech

To paraphrase Professor Emeritus Alan Male, illustrated visuals are not made to be encountered in their original state. They are contextualised, and communicate to their audience largely through contemporaneous media. From the engraved images of late 19c. newspapers, to the offset-lithography used for printing ephemera, to 8bit computer graphics seen in early video games. Commercial practitioners have implemented techniques to make their visuals fit the media. Ergo, the skill profile of practitioners has adapted to the changes in media. As with other creative domains, the early internet beckoned greater computer literacy, setting the stage for use of today’s digital media platforms. While many illustrators might use traditional methods for creating images, the dissemination of visuals happens on screens.

“In a post‐digital world, surely we are beyond the debates about digital and analog in the production of illustration: all illustration becomes digital in some shape or form.”

Roderick Mills, 2019

In 2008 former AOI Director Paul Bowman highlighted numerous shortcomings within the Illustration scene. This included what he regarded as the parochial stance taken toward emerging media. However much truth there may have been in this view, it bears mentioning that wider broadband adoption had only begun. Since then hardware limitations have diminished, and evermore online space has been opened up for multimodal visual content. Illustration may have been incubated through commercial printing, but it could not have been insulated from the far-reaching changes that swept through the communications industries. The role of contemporary illustrated visuals now lies with their emergent roles in today’s post-digital culture.

Is Copyright Now Under Threat From A.I. Machine Learning?

The emergence of machine learning has given many illustrators and visual artists cause for concern. The techlash has been especially pronounced where artwork is used by A.I. companies to train image making models. This also marks a turn for illustration markets which can now potentially be serviced by A.I. generators.

Unlike the availability of creative software, A.I. does not remove barriers to entry, it removes the need for practical expertise. Text and Data Mining is essential for A.I. development, and governments have granted copyright exceptions to make those inputs available. As a result, text-to-image generators like Stable Diffusion, Midjourney, and DALL·E, are being trained using copyrighted images harvested from the internet.

The Fallout from Machine Learning Outputs

The likelihood is that the A.I. generated outputs are where most first encounter this emergent technology. Though not protectable by copyright, an illustrator’s style is often a crucial part of their stock-in-trade, which has been made easier to coopt with little recourse. The difficulty of enforcement is partly why spawning.ai has been developed by digital artists Matt Dryhurst and Holly Herndon. Rather than react via copyright warnings and takedowns, this digital service allows artists to opt in or out of datasets used in A.I. training.

From a legal perspective, if author attribution can be reassigned by this technology, the bearing on copyright ownership is significant. Litigation involving entertainment companies repeatedly shows us how copyright can be weaponised.

First, can you copyright the output of a generative AI model, and if so, who owns it? Second, if you own the copyright to the input used to train an AI, does that give you any legal claim over the model or the content it creates?

Andres Guadamuz speaking with James Vincent – The Verge

Unsurprisingly the Society of Illustrators has disallowed A.I. work from their annual competitions. Their rationale is that A.I. has been trained using copyrighted images, which they view as “the cornerstone of the illustration community.” Similarly, the Association of Illustration has urged the Intellectual Property Office to reconsider the copyright exception granted for A.I. training.

This exception would mean that creators and rights holders of images would not be able to prevent their work from being scraped for data unless it was protected behind a paywall, which clearly would not be practical for image makers who want potential commissioners to see their work in the easiest way possible (or stock libraries who want to licence images).

The Association of Illustration

Platforms Abetting Machine Learning

Though authorial works are protected by copyright, when original content is published to an online platform those rights can be held hostage. Artist gallery platforms Artstation and DeviantArt have drawn the ire of their user-base for permitting image scraping. The community reaction has led to them introducing opt-in ‘No-AI’ tags, preventing content being used to develop A.I. generators. Meanwhile, A.I. generated images have become viral on social media.

Are A.I. Generated Images the New Stockart

Much concern related to A.I. generated images is around the erosion of illustration’s market value. Once prompted, an image generator can render countless images, making the task one of curation and adjustment. It is disheartening that a person’s creations can be used to undermine the viability of markets for that very work. Author and art journalist Zachary Small writes in Artnet;

Recent developments in machine-learning programs have turned A.I. into an impressive artistic tool capable of outpacing—and underpricing—human artists, touching off an earthquake in creative circles. Anxieties are highest among graphic artists and commercial illustrators whose livelihood is connected to their ability to turn out content to clients’ specification.

Zachary Small – Artnet

The notion that A.I. is a technological tool frames its comparison with other historical technologies. Engravers were made redundant electrotype plates, and illustrators gradually lost the advertising market because of advancements in photography. But the execution of artwork has always been pegged to human authorial intent, which is not the case with A.I. Conceptual artists can often provide a reasonable argument for the superordiante role of creative authorship. But as conceptualism has shown, this does not bode well for human craftsmanship, nor the autographic works of hand and mind working as one.

The Un-new Truths of Our Derivative Culture

In spite of how societies glorify creativity, derivative works are much the norm in our post-digital culture. This coupled with algorithmic bias means little of what we encounter is truly original. Be this as it may, published content once had to be made by people. As emphasised by the AOI, an illustrator’s contribution is not merely the image creation, but also suggesting ideas and exploring a brief to bring about best results. Though cases of progressive work using A.I. are likely, one can imagine how generators trained on chirographic man-made images shall mostly amplify the dominance of derivate images.

Might Legislation Catchup to A.I.

The intersection between A.I. and copyright law is rapidly evolving, and is where many expect to see landmark cases in the coming months. The use of A.I. does recast the role of authorship, and often makes a claim for copyright protection less accessible. However, in spite of its shortfalls, copyright law may be one of the best available deterrents against A.I. enabled plagiarism. The stakes are high. If the cultural products we consume are A.I. generated, the potential for controlling human perception may become centralised in a way that won’t be easily unspun.