Nature study finds AI chatbots answer political questions about China far more favorably when queried in Chinese
The models rely on training data influenced by state-controlled narratives.
A Nature study, amplified on X by @GaryMarcus and @nxthompson, says AI chatbots gave much more favorable answers about China when asked in Chinese rather than English. The posts point readers to a familiar AI problem with geopolitical stakes: language models reflect the material they are trained on, and the study and posts argue that Chinese-language training data carries stronger state-controlled narratives. Public reaction on X so far is limited, but the source claim itself is clearly stated.
super interesting - and a reminder that language models model language and not ideas per se. also a stark example of how easily LLMs can (as a consequence) be influenced by propaganda. https://twitter.com/nxthompson/status/2074867381420089461
Commentary on X
Highest rankedMetrics
13h agopublished34comments437