For several reasons, it is clear that large language models (LLMs) like OpenAI’s GPT-GPT should no longer be feared by academics. Artificial intelligence, machine learning, and natural language processing are just a few of the areas of research where LLMs have proved to have much promise. Researchers can advance their knowledge and provide novel solutions to challenging issues by utilizing and investigating the possibilities of LLMs.
LLMs can also be useful teaching tools. They can create excellent teaching materials, respond to inquiries, and give each student specific feedback. For students who live in isolated or underserved locations, integrating LLMs into the classroom can improve the educational experience and make information more accessible.
LLMs can also be used for artistic endeavours, including producing music, stories, and art. In addition to opening up possibilities for human-machine creative collaboration, this may also be a source of fresh artistic inspiration. Last but not least, LLMs have the power to eliminate barriers between people with various degrees of literacy, infirmities, and language proficiency. They can help those who struggle with dyslexia, vision problems, or communication difficulties so they can access information and engage in society more completely.
Due to the creative ability of Chat-GPT, academic institutions, particularly in Australia, the US, and the UK, have taken it upon themselves to educate their students on ways to cite Chat-GPT’s articles in their academic writings. In short, some citation styles have begun to develop guidelines for citing Chat-GPT and other generative AI. This is a massive development because content from generative AI was initially considered a non-recoverable source. After all, it could not be retrieved. This is because if different users give Chat-GPT the same prompt, it produces a unique response each time.
However, recently created third-party solutions like ShareGPT and AI Archives enable sharing of Chat-GPT output by creating an archive of the content and a unique link. These instruments may help document the use of LLMs like Chat-GPT. Additionally, Chat-GPT just added the ability to share links, but it’s crucial to remember that if a discussion is deleted from an account, the link will also be broken.
It is standard citation practice to cite all external sources for information. For instance, anything that is not one’s own idea is not generally known and/or is a source from which one has gathered information. For instance, if an assignment or article calls for the citation of an AI source, all of the tool’s content included in the work must be referenced. Academic dishonesty may occur from failing to cite non-original, externally sourced material. Therefore, those references should list each source’s clear and correct information and the parts of their work where they were utilized.
Citations, while acknowledging that the writer or the person citing is not the idea’s owner, simultaneously but implicitly recognizing that the person cited is the work’s author. For instance, when I paraphrase or quote an idea in my academic essays, I am accomplishing two goals: first, I am admitting that I am not the owner of the idea, and second, I am identifying the person I quoted as the owner or creator of the idea. Putting it into the context of citing Chat-GPT articles, when we cite such writings, we are implicitly acknowledging that Chat-GPT is the author of such writing. Can Chat-GPT, however, be considered authors?
Authorship is commonly referred to as making a significant, direct, and recognizable intellectual contribution to a piece of writing that will be or has already been published and taking responsibility for the study’s findings that the research product describes. There is no doubt that Chat GPT makes a substantial contribution to the article, for they are collaborative contributors to the creative process. Only significant contributions should be credited for writing an article, including conceptualization and design, data analysis and interpretation, and writing the paper or critically reviewing it for crucial intellectual content. However, is their contribution significant enough to be regarded as authors?
There are two ways to consider whether Chat-GPT can be considered as authors. The first way will be to look at what they possess. Language models like OpenAI’s GPT-GPT can generate text that seems authored by a human, but they are not “authors” in the traditional sense. These models are trained on vast amounts of text data and learn patterns in language to generate coherent and contextually relevant responses. While the output may seem impressive and well-written, the models lack the true understanding, consciousness, and intentionality that human authors possess. They do not have personal experiences, emotions, or opinions and cannot have their unique voice or perspective. Instead, they primarily mimic the patterns and style of input data they were trained on, making them more like sophisticated text generators rather than authentic authors.
The second way to understand whether Chat-GPT could be considered as authors will be to evaluate whether they can fill the role associated with authors, particularly the roles they assume after writing a creative work. These roles include taking responsibility for their writings and taking charge of the intellectual property of their articles.
This raises a serious ethical dilemma. On the one hand, one recognizes that they have to reference works or ideas that are not theirs, even if they originate from LLMs like Chat-GPT. On the other hand, one is attributing the title of an author and its responsibilities to LLMs when their position as authors is questionable, if it at all exists. Therefore, the pertinent question will be, should we stop citing LLMs’ articles and run the risk of students not citing their sources, or should we continue citing LLMs and run the risk of acknowledging them as authors?
My suggestion will be that perhaps now is not the right time to start educating students on how to cite Chat-GPT. This is because Chat-GPT is not an author, per se. And since they are not authors, they cannot own the intellectual property of their writing or take responsibility for the effects of their writing. Instead, academic institutions should significantly impact students’ knowledge of how LLMs are developed and used. The academic community can actively strive towards understanding the function and the degree of contribution of LLMs during the writing process by interacting deeper with LLMs and conducting research in this area. They do this to ensure these models are used responsibly while defining their ethical requirements and constraints.
Paschal is a postgraduate student at the University of Johannesburg, and a researcher at the UJ Metaverse Research Unit. He can be reached via email: [email protected]