What happens when a veteran mathematics textbook author and a leading AI scientist exchange ideas on the future of textbook publishing? In anticipation of his namesake keynote at TAA’s 2025 Virtual Textbook & Academic Authoring Conference, Michael Sullivan connected with Gerald Friedland—textbook author and AI Scientist at Amazon AWS—for a thoughtful conversation on artificial intelligence, authorship, and the evolving landscape of academic publishing. Friedland will expand on these ideas in his session, “Artificial Intelligence and the Future of Textbook Publishing: Opportunities and Challenges,” on Friday, June 6, from 9–10 a.m. ET.

Sullivan: “Most textbook authors engage a publisher and assign the publisher virtually all rights to their work. As a result, they must rely on the publisher to protect their rights from any infringement, such as AI. What advice do you have for these authors?”

Friedland: “I am not a lawyer and ultimately a law professional should answer this question. As far as I know though, book contracts can be terminated. This would result in the publisher not being able to sell any further copies and the copyright goes back to the authors. If the book is valuable enough, this would lead to a renegotiation of terms. My publishers have been very protective of their licensed material though, so this has not been a concern for me. What caught my attention was my publisher approaching me about further rights to be able to share the content with AI providers and the consequent uncertainty with signing such a contractual amendment.”

Sullivan: “What are the ethical implications for authors to use AI in their writing? How should citations, both author and machine, be handled?”

Friedland: “It depends on how AI is used. Using AI from scratch in the form of ‘Write me a chapter about X’ bears the risk of the content coming straight from memory and therefore being effectively a copy of somebody else’s original work. Also, in my experience, GenAI does not yet produce high enough quality for such output to merit commercial publication. Using AI to edit and correct text that came out of an author’s brain is fair game and shouldn’t be an issue. Similarly, creating translations or summaries from originally-authored text.”

Sullivan: “What do you view as the most significant ways that AI learning materials improve learning outcomes for students?”

Friedland: “With AI it is possible to adopt the language of a certain domain and skill level. I can tell AI to ‘act as a university professor of math’ or I can ask ‘explain this to a 10th grader’. This allows a very quick adaptation of the language of a given text to the reader, based on their background. In general, I see text transformations like editing, translation, and reader-level adaptation as the main drivers for AI-use in textbook authoring.”

Sullivan: “I thought it appropriate to ask AI what I should ask an AI specialist. ChatGPT gave me nine questions. I thought these two are particularly interesting and pertinent: 1) ‘In your opinion, how can AI be used to augment human capabilities rather than replace them? Are there particular areas where this is especially promising?’ and 2) ‘How do you think we should balance the rapid advancements in AI with ethical considerations, especially in areas like bias, privacy, and accountability?’

Friedland: “1) First came the text editor replacing the typewriter and correction became easier (do people still remember Tipp-Ex?). Then came the spellchecker, allowing easy auto correction. Later came the grammar checker (like Grammarly). Now we have AI and we can transform and adapt entire paragraphs and sections using different criteria, for example by adapting it to a different audience or simply just translating it. In short: AI will allow us to focus even more on the creative process. AI automates possible adaptations of the text, but only once the original text is out of the author’s brain. I believe translation and summarization are the most promising areas.

2) Unfortunately, I do not have a solution for bias. In fact, AI favors the most frequent occurrence. That is, it will always have a mainstream bias. When it comes to privacy, we already see that models like DeepSeek run on individual laptops without Internet connection. Furthermore, Amazon AWS has solutions for the private use of Large Language Models. The question about accountability is the easiest one: Word or even Latex was never accountable for the formatting of the document, it’s the author’s responsibility. In the same way,  AI will not be accountable for your text. Accountability only works between humans and therefore we will always have to have an author who is responsible for the content published.”

---

Please Note: Gerald Friedland’s responses are opinions of his own and not representative of his employer.

---

You can learn more about the possibilities of AI at Friedland’s keynote presentation “Artificial Intelligence and the Future of Textbook Publishing”, at the 2025 TAA Virtual Conference on Textbook and Academic Authoring, which will take place June 6-7.. Register here.