By Viktor Daropoulos
Introduction
ChatGPT is a large language model (LLM) that was released from OpenAI in November 2022 and quickly gathered attention due to its ability to generate human-like responses that can be difficult to distinguish from real human speech in a wide range of topics. The user is presented with a terminal-like display where a prompt is entered and the chatbot responds. Apart from responding to general questions, ChatGPT is able to translate, summarise, restructure text or even write code.
ChatGPT for academic research
Due to the ability of the bot to provide human like responses, it has already been used in academia as a tool to compose abstracts and introductions [1] causing a lot of debate in the academic society with regards to ethical concerns. For example, many journals have completely banned the use of large language models such as ChatGPT or have requested from authors to disclose their use [2]. The debate is certainly going to continue since it’s not clear at the moment how to respond to the introduction of advanced chatbots capable of humal-like text generation. Efforts are being put forth for the development of tools capable of detection of text generated by chatbots, such as “GPTZero” [3] and research papers are already being published in that direction [KGW+23]. Nevertheless, it clear that with the introduction of competitive bots such as the one expected from Google [4] the race on the production of advanced A.I language models is just getting started. Despite the ongoing debate and the ethical concerns about the use of systems like ChatGPT, there are positive aspects as well. For example, tools like ChatGPT can be used as an assistive tool for a researcher during his research. One aspect of research is reading papers, and in many cases the concepts presented in a paper are not clear or difficult to understand. In those scenarios, tools like ChatGPT can be used to explain a concept more clearly or even assist a researcher in scientific writing, something very beneficial especially for those whose native language is not English.
References
[1] ChatGPT listed as author on research papers: many scientists disapprove.
https://www.nature.com/articles/d41586-023-00107-z.
[2] As scientists explore AI-written text, journals hammer out policies.
https://www.science.org/content/article/ scientists-explore-ai-written-text-journals-hammer-policies.
[3] This 22-year-old is trying to save us from ChatGPT before it changes writing forever.
https://www.npr.org/sections/money/2023/01/17/1149206188/this-22-year-old-is-trying-to-save-us-from-chatgpt-before-it-changes-writing-for.
[4] Google unveils its ChatGPT
https://edition.cnn.com/2023/02/06/ tech/google-bard-chatgpt-rival/index.html.
[KGW+23] John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein. A watermark for large language models. arXiv preprint arXiv:2301.10226, 2023.