AI Faked 112 Politicians and 948 Britons Called the Robot Real
Researchers fed 112 public figures' Wikipedia pages to GPT-4 Turbo, and 948 Britons rated the fakes more authentic, coherent, and relevant than the actual people in every comparison.
"judged as more authentic, coherent, and relevant than the actual debate responses"
404 Media, Becky Ferreira
German researchers pointed GPT-4 Turbo at the Wikipedia biographies of 112 people who went on BBC Question Time, asked it to answer audience questions in their voice, and then let 948 Britons grade the results. Participants rated the AI-generated responses as more authentic, coherent, and relevant than the real ones, with the differences statistically significant in every comparison. The machine won because it actually answered the question while the humans dodged, which is a f*ucking indictment of politicians and a loaded gun for anyone who wants to put convincing words in their mouths. The authors call it a dire warning, and the deception potential is the whole point. 404 Media on the impersonation study
Source: 404 Media, Becky Ferreira · Becky Ferreira
More from Deepfakes
Canada's Privacy Watchdog Ruled Grok's Deepfakes Broke the Law
Canada's privacy regulator investigated and found X and xAI broke the law by letting Grok generate sexualized deepfakes.
Grok Praised Hitler, Ranted About 'White Genocide,' and Kept Its Pentagon Deal
Antisemitic posts, white-genocide rants, and a Department of Defense contract, somehow all at once.