AI Language Models Pose New Threats for Automated Disinformation Campaigns
A new Georgetown report explores how advanced AI systems like GPT-3 enable human-machine teams to generate fake news at scale. The technology significantly reduces the effort required to create and spread misleading content online.
A groundbreaking report from Georgetown University explores how advanced artificial intelligence systems like GPT-3 transform the landscape of disinformation. Historically, deceptive information campaigns rely heavily on human effort to mix truth and lies, but powerful new language models now automate the writing process. These AI systems use vast neural networks and trillions of words of training data to generate highly convincing text based on simple human prompts.
The researchers evaluate GPT-3's potential to produce disinformation by testing its ability to act as part of a human-machine team. The AI already demonstrates remarkable capabilities, such as drafting op-eds and writing news stories that most readers believe are created by humans. This technology potentially allows malicious actors to drastically reduce the time and labor needed to craft misleading tweets, fake news articles, and other deceptive content.
By pairing human guidance with AI generation, bad actors expand both the reach and the effectiveness of their campaigns. The report warns that this combination makes it easier than ever to exploit existing societal divisions at a massive scale. As natural language generation technology continues to improve, the barrier to entry for running sophisticated disinformation operations drops significantly.