AI-Generated Blog Post Flagged for Plagiarism, Leading to Doubts of Its Accuracy; Can We Really Rely on AI Writing Apps?

When the Rationalist blog was published online, it went viral garnering so much page views and attention. However, in a very unfortunate turn of events, it was marked with controversy as it appears to be an AI-generated Substack blog.

According to Futurism, it was flagged for plagiarism as it copied word-for-word the human-made work of the original author. Unfortunately, there seem very little means of recourse for the human writer whose work was copied, and almost zero hope that this frustrating event will be prevented from happening again.

(Photo : Pixabay/pixelcreatures)
AI-Generated Blog Post Flagged for Plagiarism, Leading to Doubts of Its Accuracy; Can We Really Rely on AI Writing Apps?

AI Plagiarizes Human-Made Article

The original writer of the article, Alex Kantrowitz, wrote an explanatory Substack post to describe how they found the plagiarized work. He said that The Rationalist blog post took analysis and writing directly from his previous article in Big Technology. He noted that it would have been a terrific job if only the publication was authentic.

Indeed, to Kantrowitz's point, The Rationalist's piece was a huge success as it wound up on the popular site Hacker News, which has a large following and sparked a lively discussion.

However, as Kantrowitz explicitly states, the paper was an obvious duplication of his own, both in terms of particular lines and general meaning and structure. While Substack offered to look into the incident, there appear to be a number of red flags, like the writer's "PETRA," and the article itself was repetitious and seldom connected to sources.

Kantrowitz continued that The Rationalist is an odd publication as it has no stated mission and no named writers except for PETRA. But the article was still alive for a week with words lifted directly from his article. He even commented on how impressive the Substack post is to remix, and publish their inauthentic piece, but emphasized the negative effect of generative AI.

Can AI Be Trusted?

With the rise of AI these days, writing is not the only area where such scenarios of plagiarism could occur. Artists are understandably concerned about theft and inappropriate attribution of their works as related discussions are taking place over AI-powered picture generators, such as OpenAI's DALL-E 2, which has approved images it makes to be used for commercial purposes.

While there is an existential discussion over where the boundary between inspiration and outright imitation lies, Kantrowitz's journey appears to be a great illustration of the latter, Futurism reports.

In The rationalist's case, PETRA failed to note that it was an AI-generated article in its actual Substack post. But they did come clean to some commenters in Hacker News and explained that they used GPT-3 and Huggingface, as well as other AI writing tools to help them polish their work and not intentionally plagiarize it.

PETRA explained that being non-English speakers they had to use these tools to avoid awkward workings. But the outcome was not exactly what they had hoped to happen.

If it is indeed true, then it is complicating things and concerning for anyone who writes online. Kantrowitz said in his Twitter post that once a work gets published online and is fed to an AI system to remix them then the author could not do anything to stop it. Although the intention was not to plagiarize, it still sucks and generative AI programs are bound to only get stronger.

RELATED ARTICLE: Is AI Art Really Art? Technical Progress in Computer Vision Led to More Self-Described AI Artists in the Past Few Years