Bill Tomlinson, Andrew Torrance, Rebecca Black and I wrote a paper that was accepted to the UMKC law review, a top 10% academic law review journal about what the future of academic publishing might look like if LLMs like ChatGPT were embraced as an academic publishing tool.
What does the future of academic publishing with ChatGPT hold?
In the paper we discuss the potential impact of artificial intelligence (AI) on the scholarly process and the artifacts it produces. We argue for a shift from the traditional “early-binding” process, where ideas are fully developed and finalized before being written, to a “late-binding” process, where ideas are dynamically written at the moment of reading. This shift would allow for more dynamic, personalized, and interactive scholarly works.
We highlight the historical context of scholarship, from ancient civilizations to the present day, and how advancements in technology, such as printing presses and computers, have transformed the scholarly process. We emphasize that AI, particularly large language models like GPT-3, offers new opportunities for scholars to engage with knowledge in ways that were previously impossible.
AI can revolutionize the scholarly process by enabling scholars to interact with knowledge in real-time, refining their ideas as they work. This dynamic and fluid process, known as “late-binding” scholarship, allows for the creation of works that can be continuously updated and reimagined based on new findings, alternative theories, and specific audience needs. It also leverages AI systems’ computational power to produce comprehensive and tailored works.
Late-binding scholarship advantages include enhanced personalization, improved speed and efficiency, interdisciplinary collaboration, and greater accessibility to knowledge. We compare this new form of scholarship to dynamic HTML and Jupyter notebooks, highlighting the potential for dynamic content and engagement.
However, there are challenges posed by late-binding scholarship, particularly in the realm of copyright law and issues related to authorship, ownership, transformative works, and compulsory licensing.
In conclusion, we propose a new form for scholarly works, characterized by key components such as a text abstract, hyperlinks to related works, novel data and metadata, algorithms or processes for analysis, and references to AI models for rendering the canonical version. This form enables both the rendering of the canonical version and the possibility of dynamic AI reimaginings of the text. Thus new paradigm of scholarship can lead to more effective, timely, accessible, democratized, and evergreen knowledge production.