A new study published in PLOS One reveals that AI chatbots, specifically GPT-4 Turbo, can convincingly impersonate public figures, generating responses perceived as more authentic and coherent than those of the actual individuals. Researchers prompted the AI to mimic 112 public figures from the UK, using information from their Wikipedia biographies and the format of the BBC show "Question Time." Participants rated the AI-generated responses as superior to the real debaters' answers, highlighting a significant risk of AI-driven disinformation, particularly in political contexts. AI
IMPACT AI's ability to convincingly impersonate public figures poses a significant threat of disinformation, particularly in political arenas.
RANK_REASON Study published in PLOS One detailing AI capabilities and risks.
- Anthropic
- BBC One
- GPT-4 Turbo
- OpenAI
- PLOS One
- Question Time
- Steffen Herbold
- United Kingdom
- Wikipedia
- Journalists
- Politicians
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →