John Warner’s negative view of artificial intelligence (AI) shows already in the table of contents with chapter titles such as “Automation, Not Intelligence” (chapter 1), “Stop, Now, Before It’s Too Late!” (chapter 2) and “Only Humans Write” (chapter 4). This book, then, is not a factual discussion on AI and writing – it is rather a passionate plea for the preservation of a cultural technique which he sees menaced by current technological developments. From the perspective of politically oriented technology assessment (TA) this raises an eminent question: Which political decisions are necessary to preserve crucial cultural skills like writing and reading given current societal transformations?
In the book’s first thematic section, Warner focuses on language models behind ChatGPT and Co., while throughout the book no differentiation is being made between the terms ‘large language model’ (LLM), ‘artificial intelligence’ and ‘generative AI’. This tendency to generalize also applies to the second section of the book, in which the author describes what ‘writing’ means to him. Warner, when talking about ‘writing’, means above all literary or creative writing (which he has been teaching for years at American universities) as well as academic writing. In the book’s third section, which deals with the background to writing (including reading, financial aspects and the increasing automation of text production), Warner’s specific perspective on writing as a novelist, university lecturer and journalist becomes obvious. In the fourth section, he then focuses on three keywords that he believes provide a remedy for dealing with automated text generation in the future:
The underlying thesis that serves as a unifying thread throughout the book is: “ChatGPT cannot write. Generating syntax is not the same thing as writing” (p. 8). Warner supports his aversion towards AI with evidence on a technological side such as climate damage (due to high energy and water consumption, p. 34), socially unacceptable circumstances (exploitation of workers who have to check large data sets for ‘security’ and, among other things, watch countless violent videos, p. 33) or racist bias in AI technology (p. 36). These quite objective parts of the book are contrasted by Warner’s very emotional statements that are, at best, personal, if not unprofessional: “What is going on that so much has been invested – financially, psychically, socially – in something that can plausibly be compared to a children’s toy?” (p. 27) The part in which the author devotes himself to human writing processes and everything that goes with them is permeated by a very idealistic concept of ‘writing’: “What I want to say about writing is that it is a fully embodied
Warner is right: Programs like ChatGPT which compose texts on the basis of LLMs cannot have their own idea of rhetoric, let alone intention (p. 70). On the other hand, he fails to recognize the potential of clever prompts, which can lead to texts in various styles – thanks to human input, but still created by the technological system itself. Creative uses of ChatGPT in (human) writing processes – which are certainly being discussed in the literary scene, journalism, and academia – seem to be irrelevant to the author. From a constructive TA perspective, this stresses the importance of accompanying the changing use of technologies scientifically, for example in different parts of the educational system.
Warner’s main aim is to illustrate what happens when people have texts (of whatever sort) merely ‘spoon-fed’ to them instead of experiencing the writing processes themselves. In some areas it seems unproblematic to rely on the support of generative AI programs – for example as a basis for research in some fields. However, the thought process underlying most writing processes can lead not only to changes in the text, but also in the attitude of the writing person herself towards the topic. By his mantra-like repetition ‘writing is thinking’, Warner is therefore basically saying: When people stop writing texts themselves they stop thinking. This is why the author “[…] realized I had to stop thinking about AI and start thinking about humanity” (p. 226). Yet, Warner does not deal with the philosophical question why writing is genuinely human and why a world with mere automatically produced ‘content’ would be so disastrous. Neither concepts like dignity nor thoughts about what is changing in humans in an increasingly automated world are of his interest.
When people stop writing texts themselves they stop thinking.
Contrary to the reader’s assumptions, Warner does not seem to struggle with surviving in a world of text automation as an established author. Rather, he is deeply concerned about basic communication skills: “We don’t only need to teach children to read, we need to
When Warner isn’t recounting anecdotes from his private life and professional career – which are sometimes quite interesting – he talks about everything
Another study by Hoffmann and Schmid (
Concerning the usage of AI in academia, Guest et al. (
Warner warns against getting used to average quality.
A more detailed and fact-based analysis would have been desirable in order to draw fruitful conclusions for TA. Surely, the book is not supposed to be a scientific study, however, more insights into what Warner’s students think of AI and writing would have been interesting as well as a differentiating perspective on the various usage of generative AI tools, as offered by Sundar and Lee (