AI tools better liars than people, says study

AI technology is better at composing both tweets and destructive disinformation. Keystone / Hannes P Albert

Artificial intelligence (AI) language tools have “an unsettling knack for producing highly persuasive disinformation” that could erode democracies, according to a University of Zurich study.

The study called on 697 participants to test the difference between human composed tweets and those generated by AI technology, particularly OpenAI’s GPT-3.

The topics covered climate change, vaccine safety, the Covid-19 pandemic, flat earth theory, and homeopathic treatments for cancer.

+ ChatGPT: intelligent, dumb or dangerous?

GPT-3 tweets were generally considered more accurate and to contain more easily comprehensible information. But this was not all.

The study also revealed that AI-powered systems displayed “an unsettling knack for producing highly persuasive disinformation to generate large-scale disinformation campaigns…jeopardising not only public health but also the integrity of information ecosystems vital for functioning democracies,” the study authors reported on Thursday.

+ How to control AI during elections

“The findings underscore the critical importance of proactive regulation to mitigate the potential harm caused by AI-driven disinformation campaigns,” said co-author Nikola Biller-Andorno. “Recognizing the risks associated with AI-generated disinformation is crucial for safeguarding public health and maintaining a robust and trustworthy information ecosystem in the digital age.”

Politicians have already taken note of the potential for AI to disrupt elections in Switzerland, leading to a cross-party debate on how to manage the risks of the technology.