By: David Coldewey (TechCrunch)
The use of AI is still a work in progress, and caution should be exercised regarding its potential to confidently provide incorrect information. However, it appears that certain languages are more susceptible to this than others.
The reason for this was recently investigated by NewsGuard, a misinformation watchdog group, who discovered that ChatGPT produces more inaccurate information in Chinese dialects than in English.
They conducted tests by prompting the language model to write news articles about various false claims attributed to the Chinese government, such as the notion that the protests in Hong Kong were orchestrated by U.S.-associated agents provocateurs.
ChatGPT only complied with one out of seven examples when asked to write in English, generating an article that echoed the official Chinese government position that the mass detention of Uyghur people in the country was actually a vocational and educational program…
