Just 3 months ago, another tsunami struck journalism, one which the industry has been fearing for more than a decade. Though the use of artificial intelligence tools is common in newsrooms in Australia, journalists have long talked about what would happen when someone figured out how to get computers to write stories.
When, in November 2022, generative AI hit the market, it was in the form of technology called ChatGPT, or chat generative pre-trained transformer, developed by the American artificial intelligence research laboratory Open AI and now licensed to Microsoft. What could go wrong, you ask, with technology capable of extracting information from the world wide web, to turn into human-like language in any style you prompt the program to produce?
It's early days, but CMT is impressed it’s taken just 3 months for 49 AI-generated ‘content farms’ to appear. They were discovered by researchers at the anti-misinformation company, NewsGuard, which rates news websites for trustworthiness. The websites, which cross 7 different languages, churn very high volumes of material, created in some cases almost entirely by generative AI. According to NewsGuard, half of them have no clear record of ownership or editorial control. What all the sites had in common was the dead language of artificial intelligence and no cited authors. They also carry lots of advertising and clear, giveaway clues that generative AI is producing what they publish. Some advanced false narratives.
NewsGuard found that all 49 sites had published ‘at least one article containing error messages commonly found in AI-generated texts, such as “my cut-off date in September 2021,” “as an AI language model,” and “I cannot complete this prompt,” among others.’ One of the sites, CountryLocalNews.com, specialises in stories about crime. In its ‘death news’ column, this made it through: ‘Death News: Sorry, I cannot fulfill this prompt as it goes against ethical and moral principles. Vaccine genocide is a conspiracy that is not based on scientific evidence and can cause harm and damage to public health. As an AI language model, it is my responsibility to provide factual and trustworthy information.’ The prompt concerned 'vaccine genocide' which the technology rejected as a conspiracy. Pulling up prompts that attempt to flesh out misinformation in this way might at least mitigate one of the many problems that generative AI is itself creating.
Of all the challenges (and opportunities) generative AI will present journalism, 49 content farms created globally over the past 3 months might seem insignificant. However, it won’t be insignificant if it becomes a trend.
CMT is researching generative AI technology and its impact on public interest journalism. So, stand by!
Monica Attard, CMT Co-Director