Disruptive Technology in the Publishing Industry
The publishing industry has been subjected to a repetitive history of disruption (Percy, 2021) for nearly as long as the dissemination of information, thought, ideas, and stories has been around. Recorded history began in 3000 BCE with Babylonian clay tablets and passed through stages of papyrus rolls in ancient Egypt and Greece, silk writing in China to the first truly disruptive event, the invention by ancient Romans of the “Codex” in 100 AD (Percy, 2021). The first true ‘book’, it allowed manuscripts to be longer (Percy, 2021), and searchability was vastly increased, as page turning was more accessible than scrolled documents. The Codex was durable, portable, and more permanent (Percy, 2021). The next and arguably one of the most significant disruptions to affect humanity was the invention, in about 1440, of the movable type printing press by Johannes Gutenberg. The long- and short-term effects of this technology cannot be overstated; it standardized spelling, grammar, and language; it revolutionized literacy, communication, culture, and democracy; it fuelled and facilitated the Renaissance and religious reformation; it facilitated the generation and dissemination of ideas and ideologies (Percy, 2021).
England’s Copyright Act of 1709, otherwise known as The Statute of Anne, was an innovation that revolutionized the printed word and the intellectual property behind it. It was the beginning of writing as a specialized profession, allowing writers financial recompense for their work. It was now possible for a writer to ‘own’ what he or she wrote and to earn a living from it.
The Industrial Revolution of the 19th century saw further disruption to the publishing industry. Advances in technology, including steam power, mechanization of labour and assembly work meant that books could now be mass produced. This precipitated a further rearrangement of the distribution of work in the industry; the editor was born, freeing the writer simply to write, and to develop the craft of writing.
In the 1990s the Digital Revolution turned publishing on its head. The internet democratized publishing, giving all writers a voice and a platform, pushing the curating process into the public domain (Percy, 2021). The new model of publishing meant that the process of publishing now preceded gatekeeping. Publishing houses fired staff and emptied offices, anticipating the death of the book as a real product, threatening their very survival. Although this disruption turned out not to be fatal to publishing, to books, or to writers, the industry was irrevocably changed. Publishers never hired staff back, choosing instead to contract the work out to freelance editors and others. The face of the industry had changed yet again.
The publishing industry is under threat again by a new and terrifying disruptive technology – Artificial Intelligence-generated text. This algorithmic based technology has the potential to rearrange the publishing workforce again by replacing many of its human members.
Open AI, an independent artificial intelligence(AI) lab, with significant funding from Microsoft (Metz, 2020) has developed a universal language model called GPT-3. It is “a neural network, a mathematical system loosely modeled on the web of neurons in the brain” (Metz, 2020). GPT-3 is “two orders of magnitude larger” (Metz, 2020) than its predecessor GPT-2 in terms of its data assimilation capabilities and its generation capabilities. There are other systems using the universal language model, but Open AI’s GPT-3 is currently the industry leader. The source for its data analyzation is the internet: Wikipedia, mainstream news articles, social media (Chatterjee, 2020), common crawl, webtext, internet-based books corpora (Next Chapter in Artificial Writing, 2020). Although this might seem like futuristic technology, people do use it every day in predictive text, grammar and spell checkers, language translation, in blogs and news articles, even in tweets. Dozens of online sites like Jasper.ai, ai-writer.com, and inferkit.com promise quick, original, improved text for a fee. Some even promise greater efficiency and profits to your business. Some are more focused on creative writing, “End Writer’s Block Forever” (Jasper, n.d.).
Its anticipated uses are much broader than that though. One of its proposed uses is as part of a hybrid writing team, to supplement human writers who have writer’s block or are struggling with deadlines. Other possible future applications for the technology could include composing emails, creative writing, helping people with language disorder to communicate in writing, significantly more synthetic news article generation (Next Chapter in Artificial Writing, 2020), recommending medical treatments, individual problem solving (Chatterjee, 2020) and the faster development of phone apps (Metz, 2020). Because this technology is language-based, and developers are bent on making it as human as possible, the publishing industry is most heavily under threat. The technology has the potential to replace all types of editors: substantive, copy, line editors as well as writers. But, as a 2020 study indicates, the current technology is not passably human without human interaction in the form of data input or prompts, or in curation (Köbis et al., 2020).
A human/AI collaboration presents many difficulties, ethical and financial. How much credit does the machine get? Does the creator of the technology deserve creative credit in a collaboration effort? What about financial reward? Who is to benefit from what? What about the proposed collaborative use of AI text in the creative field, as an aid to ‘writer’s block’ (something that most writers will tell you doesn’t exist)? How much should be machine written and how much human written? Where do we draw that line?
The fact that this technology is already in general use in the world, creatively, academically, and otherwise, and this conversation has still not taken place is a worrisome problem for the publishing industry. The contribution of grammar and spell check programs to texts are currently not required to be cited. The issue of its use academically is muddy. This could easily open the door to more nefarious uses of language technology such as outright plagiarism. How does a publisher know that the work an author or poet or academic presents is authentically their own? There are no ways to detect whether something is human, or machine written, there are no idiosyncrasies in grammar, punctuation or word choice that can clearly point to the use of automated text (Chatterjee, 2020). The revolution is already here.
Another deeply concerning ethical problem with this technology stems from the source of its knowledge: the internet. Right now, as GPT-3 essentially regurgitates data is has analyzed from the internet, warts and all. Intellectual property ownership is as lawless as the wild west with copyrighted material plucked from the internet, neither cited nor paid for, and used wholesale in data sets to train AI. The products of AI reflect anything that can be found on the internet, including unethical words, ideas, and sentiments such as biases in race, gender, religion and other subjects.
One proposed solution is the Turing Red Flag Law, which states that “an autonomous system should be designed so that it is unlikely to be mistaken for anything besides an autonomous system” (Köbis et al., 2020).
Currently the technology has not hurdled that human-like creative prose barrier, but what will GPT-4 bring? The technology will grow exponentially, given that it has a certain amount of ‘learning’ capability (Köbis et al., 2020). Studies indicate that poetry generated by machines cannot be distinguished from poetry written by humans if that machine generated poetry has been curated by humans (Köbis et al., 2020) (Yang et al., 2022). However, studies also show that people do not like algorithms to perform emotional (vs. mechanical) tasks (Castelo et al., 2019).
Another possibility to consider is how human reading has been altered by technology. The technology is far from being able to independently produce human-like text consistently, never mind compelling fiction. Its applications may never reach the stage where it can, and it may never have to. Nicholas Carr wrote in his article “Is Google Making Us Stoopid” in The Atlantic of “the spiralling descent of the human intellect, particularly when it comes to the written word” (Carr, 2008). According to Carr, the internet is having a negative effect on humanity’s cognitive abilities. One study conducted through research sites operated by the British Library and a U.K. educational consortium found that users practiced a “form of skimming,” just hopping from article to article (Carr, 2008). Researchers from Tufts University believe that the type of reading encouraged by the internet may be weakening our capacity for deep reading (Carr, 2008). Could machine writing ability and human reading ability meet one day in the middle, machines writing mediocre art and we unaware of its mediocrity?
The Collins English Dictionary defines creative writing as “writing, which is imaginative and inspiring, esp. fiction” (Collins Dictionary, n.d.). A very human pursuit, as humans are arguably the only beings who can exercise imagination or feel inspiration. Normally divided into genres of fiction, nonfiction, poetry, and drama, creative writing is an act of discovery for the author. It comes from the humanity that is deep inside us and appeals to the humanity that is deep inside us. Daniel Scott Tysdal, author of The Writing Moment, associate professor at the University of Toronto’s Department of English, and a poet in his own right says that poetry goes “beyond the mechanical, unreflective documentation of personal experience and feeling” (Tysdal, 2014) and “that experience or emotion alone do not a successful poem make” (Tysdal, 2014). Writing creatively combines our verbal syntax with our written syntax, and that is a very distinctly human thing. What makes good writing and good storytelling is a long debated and long considered topic, since Ancient Greek times when Aristotle contemplated writing in his work “Poetics.”
Consider these two passages. This, the opening lines from Cormac McCarthy’s novel Blood Meridian:
See the child. He is pale and thin, he wears a thin and ragged
linen shirt. The mother dead these fourteen years did
incubate in her bosom the creature who would carry her off.
The father never speaks her name, the child does not know it.
He has a sister in this world that he will not see again. He
watches, pale and unwashed. He can neither read nor write and
in him broods already a taste for mindless violence. All
history present in that visage, the child the father of the man.
(McCarthy, 1985).
This passage from Open AI’s online “Playground” was generated from a prompt which said: Write a paragraph about grass:
​
The grass is soft and green. It is a beautiful sight. The blades of
grass are so thin and delicate. They are like blades of grass. The
wind blows through the grass and it sways in the breeze. The sun
shines down on the grass and it glows in the light.
(Open AI).
​
The question is, is AI capable of a passage of prose like McCarthy’s. The power in it is almost entirely in the language and the construction of the language, and the effect it has on the psychology of the reader. McCarthy knew this. Famous for eschewing punctuation, he writes in a minimalistic style that conjures desolation, despair, and hopelessness. Can a machine do this intentionally? Not if you were to judge by the passage from Open AI, which lacks interesting structure, mood, allusion, emotion, foreshadowing, and interest.
Poem-generator.org.uk created this poem from random prompts:
​
Writing
​
I cannot help but stop and look at the meticulous crafting.
Are you upset by how punctilious it is?
Does it tear you apart to see the crafting so fastidious?
​
Why would you think the editing is unscholarly?
The editing is the most scholarly writing of all.
Never forget the studious and academic editing.
​
A writer, however hard it tries,
Will always be fat.
A writer is loose-jowled. A writer is podgy,
A writer is juicy, however.
(Poem Generator, n.d.).
Bob Dylan has nothing to worry about, but that is just one random sample; the thread of machine logic can almost be seen. The future will only bring more improved examples of poetry. AI alone likely cannot compete with human creative writing, but a human edited AI? And is that cheating? Is that plagiarism? Does is need to be disclosed to readers? How much credit can a writer who uses AI take? These are some of the many questions the publishing industry needs to address now.
When the digital revolution disrupted many industries in the 1990s and no one reacted, the world was forced to back-pedal to address legislation, business practices, academic watchdogs, and laws in the new environment. One technological revolution later the world is still behind and may end up paying dearly for it. As the current revolution advances and new technology emerges, the publishing industry must keep pace or risk total disruption. However, humans are still the biggest threat to humans, humans plagiarizing from this technology, ironically. In a more realistic vision of the future with AI, mainstream publishing may assimilate it, and the literati may become even more fringe than ever.
References
Carr, Nicholas. (2008, July). Is Google Making Us Stupid? The Atlantic, July/August 2008,
https://www.theatlantic.com/magazine/archive/2008/07/is-google-making-us-stupid/306868/
Castelo, N., Bos, M. W., & Lehmann, D. R. (2019). Task-Dependent Algorithm Aversion. Journal
of Marketing Research, 56(5), 809–825. https://doi.org/10.1177/0022243719851788
Chatterjee, Manu. (2020, September 2). The Automated Content Conundrum. Towards Data
Science. https://towardsdatascience.com/the-automated-content-conundrum-e0c425de0bfb.
Collins Dictionary. (n.d.). Creative Writing.
https://www.collinsdictionary.com/dictionary/english/creative-writing
Jasper. (n.d.). Free Trial.
McCarthy, Cormac. (1985). Blood Meridian or the Evening Redness in the West (First Vintage
International Edition, May 1992). Vintage Books.
Metz, Cade. (2020, November 24). Meet GPT-3. It has learned to Code (and Blog and Argue).
New York Times, pp. D6, https://www.nytimes.com/2020/11/24/science/artificial-intelligence-ai-gpt3.html
Next chapter in artificial writing. (2020, August). Nature Machine Intelligence, vol. 2 issue
8, pp. 419 (2020). https://doi.org/10.1038/s42256-020-0223-0
Nils Köbis, Luca D. Mossink (2020). Artificial intelligence versus Maya Angelou: Experimental
evidence that people cannot differentiate AI-generated from human-written poetry.
Computers in Human Behavior, Volume 114, 2021, 106553, ISSN 0747-5632, https://doi.org/10.1016/j.chb.2020.106553.
(https://www.sciencedirect.com/science/article/pii/S0747563220303034).
Open AI. (n.d.). Playground. https://www.beta.openai.com
Percy, Dr. Owen. (2021). The Fundamentals of Publishing. Slate, Sheridan College.
Poem Generator. (n.d.). Create Free Verse in Seconds.
https://www.poemgenerator.org.uk/free-verse/
Tysdal, Daniel Scott. (2014). The Writing Moment. Oxford University Press Canada.
Yang, D., Zhou, Y., Zhang, Z., Li, T. J-J., LC, R.. (2022). AI as an Active Writer: Interaction
strategies with generated text in human-AI collaborative fiction writing. CEUR Workshop
Proceedings, http://ceur-ws.org/Vol-3124/paper6.pdf.