Do people agree / disagree? 
Discussion
That might be true unless your asking about Tiananmen Square or the CCP.
Is there a list of available Chinese LLM’s to test?
Depends on the inputs, no? How do you know it’s being censored if you’re not allowed to think otherwise
Maybe the closed source ones.. The Chinese make many of the self hosted ones though.. It really depends on the model, this could be correct..
I Disagree. While DeepSeek may be “faster” or “learn” more; more isn’t always better. It confuses itself A LOT and contradicts consistently!
If we call blatantly wrong statements “less censorship” sure, but it doesn’t make the output any more accurate.
Interested to see if nostr:npub1yy3uc7y85faxy5cwq8f574yda3j9457p7r9h2ymlkwlx6lp8f4vsy4kawc has a thought on this?
Literally just put out a note on this. DeepSeek is pushing hard and making gains quickly. Not entirely certain about the “openness” of the platform though.
Reposting the article here for visibility.
I just saw that after I tagged you 🤦♂️
Surprised the model is Open Source, which is cool. They do mention censorship around the CCP in the article, but done at higher layer.
I use Nostr. I use Bitcoin. The ability to censor me weakens everyday.
This is something that can be measured. I don’t see this as an opinion kind of question.
Is the guy with the most censored & closed Ai system asking this question?
I can’t take conehead ScamDreesen seriously, even when he poses basic, relevant, semi-thought provoking question.
That shifty snake ass VC grifter always has an angle.
Wouldn’t know 🤷♂️
This is a decent community discussion on the latest DeepSeek release.
I'm sure the Chinese LLMs are censored within China...
So as a follow up, can now confirm that not only is Deepseek censored locally in #China, the same too for servers hosting outside of China.
Sat down this afternoon with my IT guy and compared both personally. Marc Andreessen is completely wrong with his comment. Cc nostr:npub1trr5r2nrpsk6xkjk5a7p6pfcryyt6yzsflwjmz6r7uj7lfkjxxtq78hdpu
Well.. maybe it gets a different point of view..
This says more about ChatGPT than the Chinese models.
never thought I'd read that
there are laws in china that makes LLM development serious. you can't talk against ccp i guess. ask it about tianenmen and taiwan and it may refuse or give a weird answer. but marc is probably talking about wokeness in american LLMs. the corp built LLMs are pretty woke.
and i have a basedness (anti-woke) leaderboard in case anyone is interested: https://wikifreedia.xyz/based-llm-leaderboard/npub1nlk894teh248w2heuu0x8z6jjg2hyxkwdc8cxgrjtm9lnamlskcsghjm9c
deepseek r1, while everyone loves the smartness, i don't like the human alignment. it is worse than most models. and worse than deepseek v3.
idk
..
Can you speak of Tiananmen square ?
Somewhere, somehow, there must be very advanced, and totally uncensored, LLMs. Those who actually rule us (I am not talking about people like Biden or Trump) would demand that such a thing exist, for their own use and benefit. They would also want to prevent public access to such LLMs, because the nature and identity and outrageous crimes of our rulers might thereby be publicly disclosed.
having spent some time unscientifically but persistently comparing LLM censorship in response to the same prompts with CN and US models I'd say that some CN models I've explored have galaxies of off limits outputs. however they are different. so you need to ask the kinds of questions that get these answers. Which means you need to focus on censored themes in China. Like the five poisons...
Thought of this paper: https://arxiv.org/html/2410.18491v1
Of course Chinese AI will be super censored. The stability of their society demands it. Dictators can’t have an open society
