• pleiades@lemmy.ml
    link
    fedilink
    English
    arrow-up
    2
    ·
    5 days ago

    Very odd. I don’t know how taboo it is to say negative or incorrect things about Mao in China, but perhaps they don’t want their model accidentally doing that?

      • yunqihao [comrade/them]@hexbear.net
        link
        fedilink
        English
        arrow-up
        3
        ·
        5 days ago

        Given the chaos of the cultural revolution and the poverty worship of the gang of four and their supporters definitely possible but honestly unlikely.

        The most likely reason in my opinion is simply due to how LLMs are trained knowing the general vibe of the training data especially in English (full of atrocity propaganda) when it comes to sensitive people or events it’s best to simply censor it so those most likely to fall for said propaganda can’t be made use of and those who wouldn’t can simply do the research. (There’s a similar train of thought with regards to the firewall and the necessity of a VPN to access western “news” discouraging most who could be made use of to even bother while those who make the effort are less likely to fall for it.)