OpenAI, a research and development company focused on artificial intelligence (AI), recently introduced ChatGPT, a large scale language model (LLM) designed to have human-like conversations. In just five days, over one million users have utilized ChatGPT to answer complex questions or generate short texts. However, detecting plagiarism in texts generated by ChatGPT can be difficult compared to manually created texts.
A recent study in the Frontiers in Public Health journal examined the evolution of LLMs and how ChatGPT may impact future research and public health. The study seeks to facilitate discussion on the function of ChatGPT in medical research, given the concept of “AI-driven infodemic”.
Over the past five years, LLMs have experienced exponential growth, thanks to self-attention network architectures, called transformers. In 2018, two innovative models, Generative Pretrained Transformer (GPT) and Bidirectional Encoder Representations from Transformers (BERT), were developed using a combination of supervised fine-tuning and unsupervised pre-training. The GPT model rapidly evolved, leading to the creation of GPT-3, which is 100 times larger than GPT-2 and contains 175 billion parameters.
However, many LLMs, including GPT-3, tend to produce biased text with false facts, reproducing the bias because they predict the next text element based on available data on the internet. OpenAI developed ChatGPT with 1.3 billion parameters trained using Reinforcement Learning from Human Feedback (RLHF), addressing ethical problems and establishing standard rules of its application. The accuracy of ChatGPT has improved, but all limitations must be considered, especially in medical research.
ChatGPT can be used by researchers to produce significant scientific papers, generating titles, writing drafts, or expressing complex scientific concepts in simple, grammatically correct English. However, manipulating LLMs to produce texts related to controversial topics or misinformation is possible, which can endanger people’s safety. Therefore, LLM detectors should be improved to identify fake news and ensure compliance with author guidelines.
During the COVID-19 pandemic, information spread through social media, creating a phenomenon known as an infodemic, significantly impacting medical decision-making in prevention and treatment strategies. The authors predict that AI-driven infodemic outbreaks will pose a significant public health threat in the future.
In conclusion, ChatGPT, an AI chatbot that can have human-like conversations, raises concerns regarding an AI-driven infodemic in public health. It is essential to address ethical concerns and establish standard rules of its application to minimize the risks.