THE EFFICACY OF DETECTING AI-GENERATED FAKE NEWS USING TRANSFER LEARNING
Issues in Information Systems · 2023
This research investigates the detectability of AI-generated fake news by a fine-tuned BERT transformer that does text classification.The performance of the model is evaluated by testing it on GPT-generated (Generative Pre-trained Transformer) news articles as well as real-world news articles from a Kaggle dataset.The study focuses on how fake news articles can be created by AI tools and how effective a finetuned, open-source large language model is at detecting the GPT-generated fake news.The model was fine-tuned with Kaggle real-world articles and GPT-generated articles from OpenAI ChatGPT and Davinci.The use of state-of-the-art tools like the BERT pre-trained, large language model transformer were found to effectively classify GPT-generated articles but were less effective at classifying fake news articles that were not GPT generated.