Large language models (LLMs) like GPT-4 can identify a person’s age, location, gender and income with up to 85 per cent accuracy simply by analysing their posts on social media.

But the AIs also picked up on subtler cues, like location-specific slang, and could estimate a salary range from a user’s profession and location.

Reference:

arXiv DOI: 10.48550/arXiv.2310.07298

  • cucumber_sandwich@lemmy.world
    link
    fedilink
    English
    arrow-up
    6
    ·
    1 year ago

    There was a study on Facebook that showed that they could predict with between 80-95% accuracy (or some crazy number like that) your gender, orientation, politics, and so on just based on your public likes. That was ten years ago at least. What is this even showing?

    Advocates diabolo: that a large language model can do it without extra training, I guess. The Facebook study presented a statistical model on “like space” while this study relies on text alone, a much less structured type of input.

    I’m not saying it’s a good study. Just pointing out some differences.