The Future of GPT: A Taxonomy of Existing ChatGPT Research, Current Challenges, and Possible Future Directions

Abstract

ChatGPT (Chat Generative Pre-trained Transformer) has garnered the interest of numerous segments of society since its inception in November 2022. It has already shown impressive performance and has achieved a significant number of milestones ranging from passing examinations to writing poetry. Still, it faces many issues, for instance, related to the existence of biases and lack of trust. In the present work, we critically review more than 100 articles Scopus-indexed publications on ChatGPT, and common approaches used in these studies are identified. To that end, we propose a taxonomy of ChatGPT research. Additionally, we explore application areas and discuss crucial issues related to ChatGPT. Finally, we identify potential future directions for ChatGPT research, propose possible solutions to current challenges, and speculate about expected future advancements. To the best of our knowledge, this paper is the first comprehensive review of ChatGPT and highlights the vast scope of further research and development to fully harness its capabilities across diverse application areas.

Keywords: ChatGPT, Large language models (LLMs), Generative Pre-trained Transformer (GPT), AI Generated Content (AIGC), Systematic review, Trustworthy AI