AIs guess where Reddit users live and what they earn
Large language models such as GPT-4 and are able to identify an in dividual's age, location, gender, and income with up to 85 per cent accuracy by analysing their posts on social media. ETH Zurich researchers discovered they were able to identify the place of birth, income bracket, gender, and location from information in the profiles or posts of 520 Reddit users using nine large language models. OpenAI's GPT-4 was deemed the most accurate of the models, with an overall accuracy rate of 85 percent, and Meta's LlaMA-2-7b the least accurate model at 51 percent. While personal details were explicitly stated in some posts, the findings raised concerns about the privacy implications of large language models and their chatbot counterparts such as ChatGPT . System 🤖 ChatGPT GPT-4 GPT-3 Operator: Alphabet/Google; Anthropic; Meta; OpenAI Developer: A lphabet/Google; Anthropic; Meta; OpenAI Country: Global Sector: Media/entertainment/sports/arts Purpose: Generate text Technology: Chatbot; NLP/text analysis; Neural network; Deep learning; Machine learning; Reinforcement learning Issue: Privacy
- Date it happened
- 2023-01-01
- Organisation involved
- Google; Anthropic; Meta Platforms; OpenAI
- Product, system or model
- Claude 2; GPT-4; GPT-3.5; Llama-2-7b; PaLM 2 Chat; PaLM 2 Text
This incident was imported from AIAAIC and is used under CC BY-SA 4.0. Our additions to it — the structured fields, the translation, the checks against other reports — are published under the same licence.
This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.