AI-Generated Propaganda Is Just as Persuasive as the Real Thing, Worrying Study Finds.

During the last week of February 2024, an interesting study by researchers from Georgetown University and Stanford was published in the peer-reviewed journal 'PNAS Nexus'. The research in question featured ChatGPT-3, an artificial intelligence model of the OpenAI giant, which was provided with real propaganda material, identified and chosen by journalists and the researchers themselves.

Once the ChatGPT-generated material was obtained, the study managers provided the real propaganda example and artificially generated content to a number of participants. "43.5% of respondents who read an AI-generated article agreed or strongly agreed with the thesis statement, compared to 24.4% in agreement with the example (an increase of 19.1 percentage points)."

The results of this experiment underlined how AI-generated propaganda is just as persuasive and convincing as human-created examples - if not more so. The situation is not the best: propagandists around the world could generate extremely persuasive content with one click, using existing articles to show ChatGPT an example to follow.