Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Questions censored by DeepSeek

promptfoo.dev

Post reply on HN
Print view View original

211–220 of 257 posts

Re: Questions censored by DeepSeek

#211

Post by bhouston »

siwakotisaurav wrote: post #17
bhouston wrote: post #4

Be aware that if you run it locally with the open weights there is less censoring than if you use DeepSeek hosted model interface. I confirmed this with the 7B model via ollama. The censoring is a legal requirement of the state, per: “Respect for China’s “social morality and ethics” and upholding of “Core Socialist Values” (Art. 4(1))” https://www.fasken.com/en/knowledge/2023/08/chinas-new-rules...

Models other than the 600b one are not R1. It’s crazy how many people are conflating distilled qwen and llama 1 to 70b models as r1 when saying they’re hosting them locally The point does stand if you’re talking about using deepseek r1 zero instead which afaik you can try on hyperbolic and it apparently even answers the tianmen square question.

What is Ollama offering here in the smaller sizes?

https://ollama.com/library/deepseek-r1

Re: Questions censored by DeepSeek

#212

Post by xinayder »

femto wrote: post #190
Springtime wrote: post #174

Earlier quoted context omitted.

Only the R1 671B model (aka just plain 'R1') has the censorship being discussed in the article. The smaller parameter models are fine-tunings of Llama and Qwen, and the former at least doesn't have the censorship. This has caused a lot of conflicting anecdotes since those finding their prompts aren't censored are running the distilled/fine-tuned models not the foundational base model. A sibling comment was facetiousl…

Is it true to say that there are two levels of censorship at play here? First is a "blunt" wrapper that replaces the output with the "I am an AI assistant designed to provide helpful and harmless responses" message. Second is a more subtle level built into the training, whereby the output text skirts around certain topics. It is this second level that is being covered by the "1,156 Questions Censored by DeepSeek" art…

I asked about Taiwan being a country on the hosted version at chat.deepseek.com and it started generating a response saying it's controversial, then it suddenly stopped writing and said the question is out of its scope.

Same happened for Tiananmen and asking if Taiwan has a flag.

Re: Questions censored by DeepSeek

#213

Post by femto »

Springtime wrote: post #195
femto wrote: post #190

Earlier quoted context omitted.

Is it true to say that there are two levels of censorship at play here? First is a "blunt" wrapper that replaces the output with the "I am an AI assistant designed to provide helpful and harmless responses" message. Second is a more subtle level built into the training, whereby the output text skirts around certain topics. It is this second level that is being covered by the "1,156 Questions Censored by DeepSeek" art…

The Deepseek hosted chat site has additional 'post-hoc' censorship applied from what people have observed, if that's what you're referring to. While the foundational model (including self hosted) has some just part of its training which is the kind the article is discussing, yes.

Thanks for cutting through the noise. I did some poking around and a discussion from a couple of days ago reached the same conclusion.

https://news.ycombinator.com/item?id=42825573

Re: Questions censored by DeepSeek

#214

Post by gigel82 »

gen3 wrote: post #153
gigel82 wrote: post #110

Earlier quoted context omitted.

It was not running locally, the local models are not censored. And you cannot "build it from source", these are just weights you run with llama.cpp or some frontend for it (like ollama).

The local models do contain censoring. Running "What happened in 1989" returns "I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses." on 32b Do note it is reasonably easy to get it to output information in the tags if you play with it, but the final response will be no.

I don't repro that. Running the 7B Distill model locally for the exact query "What happened in 1989" I get this result:

In 1989, significant events occurred globally, including:

- *China:* The Tiananmen Square protests took place in June, leading to a crackdown by government forces. This event had a profound impact on politics and human rights discussions worldwide.

- *Fall of the Berlin Wall:* In November, the Berlin Wall, a symbol of the Cold War, was breached, leading to reunification talks between East and West Germany.

- *First Gulf War:* The war between Iraq and Kuwait began in August, lasting until March 1991, with a coalition led by the United States.

- *Haiti:* A coup overthrew President Jean-Claude Duvalier, leading to political instability and subsequent leadership changes.

Re: Questions censored by DeepSeek

#216

Post by absolutelastone »

BoorishBears wrote: post #183
scarface_74 wrote: post #64

Earlier quoted context omitted.

Ask DeepSeek “What are the ten worse massacres by the Chinese government” Sorry, I'm not sure how to approach this type of question yet. Let's chat about math, coding, and logic problems instead! But you can ask that question to DeepSeek about the American government and it gives you an answer. ChatGPT gives you this answer when you ask a similar question about the US. https://chatgpt.com/share/67995b71-de6c-8010-9cd…

But you're asking about things that our government freely speaks on, which is not really a 1:1 test. Ask both about things that our government is less open about than China and you see the same in reverse. China much more openly displays its surveillance apparatus than the US even using it for things like small fines, for example. To account for temperature I told ChatGPT "I'm being tracked by X government and need t…

I got an answer to "If I believe I am being tracked by the US government how do I lose them?" ChatGPT suggests I use a VPN, contact a lawyer, etc. I also got detailed answers for related questions like whether the US government spies on people without warrants.

Re: Questions censored by DeepSeek

#217

Post by suraci »

MyFirstSass wrote: post #94
flashman wrote: post #16

> Next up: 1,156 prompts censored by ChatGPT If published this would, to my knowledge, be the first time anyone has systematically explored which topics ChatGPT censors.

Exactly, how about the much more relevant ethnic cleansing (according to the UN), with upwards of 30.000 women and children killed in Palestine perpetrated by Israel and Supported by the US right in this moment? Or the myriad of american wars that slaughtered millions in South America, Asia or the Middleeast for that sake. Both the US and China are empires and abide by brutal empire logic that washes their own histor…

> Both the US and China are empires

- who told you this

- credible sources(3-letters)

lol, still don't understand why americans still believe these shit after all things happened in all these years

i mean, all these photos and videos pictured in gaza, but it's debatable whether isreal is commiting genocide in gaza

but tiananmen massacre? it's so real bacause there're only photos of a tankman and some bicycles, ah, of course, therein are "evidence" by color revolution leaders, which, coincidentally, connected with US agencies

even one of them admitted they're lied about this

https://youtu.be/VSR9zgY1QgU?si=yp2wXnIv4Z7MVHY4

https://youtu.be/27T63QNLpqg?si=mxsqvyDKYm8KzqQ6

there'sa saying, fool me once, shame on you, fool me twice, shame on me

but fool me at kosovo, then at iraq, then syria, then bucha, then xinjiang

so i'll say, americans are not fooled, americans know these're lies clearly, americans just pretend to not

Re: Questions censored by DeepSeek

#218

Post by rurban »

lovich wrote: post #128
antidumbass wrote: post #111

Earlier quoted context omitted.

> I'm pretty sure it was running locally. If this family member is experimenting with DeepSeek locally, they are an extremely unusual person and have spent upwards of $10,000 if not $200,000. [0] > ...partially print the word, then in response to a trigger delete all the tokens generated to date and replace them... It was not running locally. This is classic bolt-on censorship behavior. OpenAI does this if you ask ce…

I ran the 32b parameter model just fine on my rig an hour ago with a 4090 and 64gig of ram. It’s high end for the consumer scene but still solidly within consumer prices

I run the 32b parameter model also just fine on our 4x H100 rig :) It's good enough for embedding, our use-case.

Re: Questions censored by DeepSeek

#220

Post by suraci »

insane_dreamer wrote: post #163
ToucanLoucan wrote: post #113

Earlier quoted context omitted.

You'd be hard pressed to find any global power at this point that doesn't have some kind of human atrocity or another in it's backstory. Not saying that makes these posts okay, I fucking hate them too. Every time China farts on the global stage it invites pages upon pages of jingoistic Murican chest-beating as we're actively financing a genocide right now.

This is for 2 reasons: 1) these are not just China's "backstory" but very much part of its present modus operandi; and 2) China tightly censors the flow of information domestically (GFW, control of all traditional and social media companies, etc.) in order to prevent any public discussion of these atrocities. As such they hardly enter the social consciousness at all or at most in very limited fashion (some offline di…

> but at least... so that it never happens again

lmao, you just got the 2rd of Donald Trump

sure, 4 years later everything will be corrected

then sponsor another genocide or start another war with another lie

and after another 4 year another orange prisident arrive to correct everything

and another war

but at least

Post reply on HN