AI-generated disinformation more convincing than human-generated content, study reveals
Originally Published 2 years ago — by The Verge
A new study comparing tweets written by humans to those generated by OpenAI's GPT-3 language model found that people were more likely to find AI-generated tweets convincing and had a harder time discerning between AI-generated disinformation and tweets written by humans. The study focused on science topics like vaccines and climate change, which are often subject to misinformation campaigns. The research highlights the potential for AI language models to be used to spread disinformation and emphasizes the need for improved training datasets and critical thinking skills to counteract the spread of false information.