Because the host you’re using, or the specific model, has those censors enabled to follow state law, just as all US hosted LLMs will have to call the gulf of Mexico the Gulf of America within the year.
Use the uncensored model and run it locally. Or better yet don’t use LLMs whatsoever since they destroy the earth faster than crypto while providing even less benefit.
A) LLMs are not a source. They never will be. They do not have the capacity to say true things any more than any other random number generator.
B) the reason the exact question you asked is censored is not because of Winnie the Pooh, a chapter featured to this day on tv, in theme parks, in children’s books, and across the internet. It is because the only response from the training data the LLM was trained on could generate is disinformation as forbidden by mainland law. No, Chinese users on WeChat didn’t start the meme. This is categorically false, and instead of repeating this explanation constantly to whatever racist fuck tries to justify it, sites and information to the contrary are simply blocked.
It’s less work to fight bigots that way. Imagine if a bunch of people suddenly claimed that a picture of Obama as a monkey was made by black Americans as a total in joke about how he walked one time because that makes sense. Instead of explaining to every person that wonders if it’s true, just don’t even entertain the question.
In the West, smart people call the technique being blocked JAQing off, aka ‘im just asking questions.’ it’s a staple argumentative technique of fascists, and it takes more energy to interact with than just stopping it at the source.
You can disagree with that method, that we should waste time addressing every bigoted question a million billion times, but as we see in the US that simply does not work so you need to present another solution.
Really? It’s a myth? I had no idea! To learn more, I used China-based Deepseek to find more information!
Aaand it broke, oh no!
An LLM fails to do something? Must be a day ending in y.
Why does it exclusively fail when discussing topics I have been led to believe China censors online, but works perfectly otherwise?
Because the host you’re using, or the specific model, has those censors enabled to follow state law, just as all US hosted LLMs will have to call the gulf of Mexico the Gulf of America within the year.
Use the uncensored model and run it locally. Or better yet don’t use LLMs whatsoever since they destroy the earth faster than crypto while providing even less benefit.
Ah, I see, so the China-based model can’t answer any questions about China due to censorship!
Thank you for helping me see just how bad censorship is in China, I had no firsthand experience until now!
🤣
A) LLMs are not a source. They never will be. They do not have the capacity to say true things any more than any other random number generator.
B) the reason the exact question you asked is censored is not because of Winnie the Pooh, a chapter featured to this day on tv, in theme parks, in children’s books, and across the internet. It is because the only response from the training data the LLM was trained on could generate is disinformation as forbidden by mainland law. No, Chinese users on WeChat didn’t start the meme. This is categorically false, and instead of repeating this explanation constantly to whatever racist fuck tries to justify it, sites and information to the contrary are simply blocked.
It’s less work to fight bigots that way. Imagine if a bunch of people suddenly claimed that a picture of Obama as a monkey was made by black Americans as a total in joke about how he walked one time because that makes sense. Instead of explaining to every person that wonders if it’s true, just don’t even entertain the question.
In the West, smart people call the technique being blocked JAQing off, aka ‘im just asking questions.’ it’s a staple argumentative technique of fascists, and it takes more energy to interact with than just stopping it at the source.
You can disagree with that method, that we should waste time addressing every bigoted question a million billion times, but as we see in the US that simply does not work so you need to present another solution.