17 ms·
I think we need some LLM that is trained with CCP censorship that suggest you how great leader Xi is and how you suppose to work for glory of The Country and te
by SXX 3y ago
I think we need some LLM that is trained with CCP censorship that suggest you how great leader Xi is and how you suppose to work for glory of The Country and tell you how many social credits you lost or earned on each request.
Might be then people behind LLMs in Meta and OpenAI will see how laughtable and awful this childish censorship is.
- sixothree 3y agoWould that not be the actual result of training an llm with real data?
- thaumasiotes 3y agoYour wish has been granted: https://languagelog.ldc.upenn.edu/nll/?p=60536 https://languagelog.ldc.upenn.edu/nll/?p=60536 In a somewhat surreal turn of events, the author of that post concludes that "Hobbled by such constraints and limitations, China's chatbots are certain to be inferior to those of free societies." (He is prone to slamming the CCP regardless of whether he can make a logical case, which is odd since it's not like it's hard to make justified criticisms. He is also known for his habit of deleting comments if he doesn't like them, which is why that post has only the one comment on it.)
- dragonwriter 3y ago> Might be then people behind LLMs in Meta and OpenAI will see how laughtable and awful this childish censorship is. They are doing it to curry PR and political favor, which can be useful to secure favorable regulation, not because they think it. Whether it is “childish” is irrelevant to the calculations which support doing it.