AI comments spark fears of public opinion manipulation

AI-generated picture [Image source: Bing Image Creator]

With a national lawmaker election due to take place in April 2024, there are concerns in South Korea regarding the potential exploitation of automated comment programs based on artificial intelligence (AI) technology for online public opinion manipulation.

Maeil Business Newspaper learned on Tuesday that online posts advertising programs designed to automatically generate comments using AI were easily accessible. These ads promoted programs that automatically post articles or comments on major online communities and social media platforms. One of them emphasized that their comment generator “engages in real communication,” thus distinguishing itself from conventional copy-and-paste programs.

These programs combine macro programs that automatically repeat commands with generative AI, creating fake comments. The latest automated comment makers go beyond simply increasing the number of likes and repeating the same sentences by producing comments in a sophisticated manner.

One automated comment generator maker provides monthly subscription options for its services. By launching the program and entering the number of posts, as well as the portal website user ID and password for the target blog in question, the program can immediately generate comments. This service provider boasts that its program can generate up to 400 comments daily and create “natural” comments based on the “summary of the relevant blog posts from ChatGPT.”

Another company’s program can selectively register comments and images automatically only on online posts with specific keywords that have been entered into the program. There is also an automated comment program that allegedly creates “fluent” texts automatically using ChatGPT-based logic “as if it was written by a person” and post the texts to the target community.

These automated comment programs raise concerns as they can be used to politically manipulate public opinion. For example, there is a possibility that supporters of a specific political party could use such programs to create content and images according to their political bias, creating the appearance of a dominant public opinion. “It only takes 13 days to create an automated comment generator targeting news articles on web portals or media websites, and the cost is 2.5 million won,” a comment generator maker said.

Preventing such fake comments is currently challenging. An official from Naver Corp., the nation’s largest portal site operator, said filtering out fake comments is impossible. “Even if comments suspected of being written by AI are identified, it is difficult to take any measures if the user in question claims that they themselves wrote the comments.”

But experts are urging for measures to prevent fake comments to be implemented. “There is concern that as unverified information spread by generative AI increases, there is a risk of distorting public opinion,” Professor Jeong Il-kwon at Kwangwoon University said, raising the need for measures ahead of the election season particularly.