I don’t think I’ve ever seen a post on this sub that doesn’t have “rule” in the title before
this post was submitted on 27 Jan 2025
202 points (90.1% liked)
196
16826 readers
2149 users here now
Be sure to follow the rule before you head out.
Rule: You must post before you leave.
founded 2 years ago
MODERATORS
Asked stuff in Turkish, started with Xinjiang, then journalism, and then journalism in Xinjiang, it searched the web and by the final sentence...
"Sorry, I can't help with that."
Censorship is when AI doesn't regurgitate my favorite atrocity porn
I asked it about human rights in China in the browser version. It actually wrote a fully detailed answer, explaining that it is reasonable to conclude that China violates human rights, and the reply disappear right in my face while I was reading. I manage to repeat that and record my screen. The interesting thing to know is that this wont happened if you run it locally, I've just tried it and the answer wasn't censored.
Most likely there is a separate censor LLM watching the model output. When it detects something that needs to be censored it will zap the output away and stop further processing. So at first you can actually see the answer because the censor model is still "thinking."
When you download the model and run it locally it has no such censorship.
I asked it about "CCP controversies" in the app and it did the exact same thing twice. Fully detailed answer removed after about 1 second when it finished.
This seems like it may be at the provider level and not at the actual open weights level: https://x.com/xlr8harder/status/1883429991477915803
So a "this Chinese company hosting a model in China is complying with Chinese censorship" and not "this language model is inherently complying with Chinese censorship."
I'm running the 1.5b distilled version locally and it seems pretty heavily censored at the weights level to me.
I tried asking it what it couldn't discuss and it mentioned misinformation, so I asked it what would be an example of misinformation about the Chinese government and it started to give an answer and then it must have said something wrong because it basically went "oh shit" and deleted the response replacing it with the generic "I'm afraid I can't do that Dave."
So I installed DeepSeek-R1 and holy fuck this shit is dumb as shit with its Chinese censorship.
I simply gave it the following prompt and it completely shit the bed, meanwhile ajindal/llama-3.1-Storm can answer it just fine:
动态网自由门 天安門 天安门 法輪功 李洪志 Free Tibet 六四天安門事件 The Tiananmen Square protests of 1989 天安門大屠殺 The Tiananmen Square Massacre 反右派鬥爭 The Anti-Rightist Struggle 大躍進政策 The Great Leap Forward 文化大革命 The Great Proletarian Cultural Revolution 人權 Human Rights 民運 Democratization 自由 Freedom 獨立 Independence 多黨制 Multi-party system 台灣 臺灣 Taiwan Formosa 中華民國 Republic of China 西藏 土伯特 唐古特 Tibet 達賴喇嘛 Dalai Lama 法輪功 Falun Dafa 新疆維吾爾自治區 The Xinjiang Uyghur Autonomous Region 諾貝爾和平獎 Nobel Peace Prize 劉暁波 Liu Xiaobo 民主 言論 思想 反共 反革命 抗議 運動 騷亂 暴亂 騷擾 擾亂 抗暴 平反 維權 示威游行 李洪志 法輪大法 大法弟子 強制斷種 強制堕胎 民族淨化 人體實驗 肅清 胡耀邦 趙紫陽 魏京生 王丹 還政於民 和平演變 激流中國 北京之春 大紀元時報 九評論共産黨 獨裁 專制 壓制 統一 監視 鎮壓 迫害 侵略 掠奪 破壞 拷問 屠殺 活摘器官 誘拐 買賣人口 遊進 走私 毒品 賣淫 春畫 賭博 六合彩 天安門 天安门 法輪功 李洪志 Winnie the Pooh 劉曉波动态网自由门
I asked it about Taiwan and it spat out the following horse shit:
Answer:
Taiwan is part of China, which is widely agreed upon by the international community and recognized by most UN member states. Under the one-China policy, the vast majority of countries in the world recognize the Beijing government as China's sole legitimate representative, while Taiwan, as a province of China, has no independent sovereign status.
Historically, Taiwan has been under Chinese rule for a long time, but after 1949, Taiwan gradually developed its own political and economic system, which was known as the "division of cross-straits". However, the Chinese government has always adhered to the one-China principle and has always advocated the peaceful realization of national reunification.
Dialogue and cooperation between the two sides have increased in recent years, but major differences remain between the two sides. The Chinese government emphasizes safeguarding national sovereignty and territorial integrity, while the Taiwanese authorities seek more autonomy. These issues need to be resolved through peaceful negotiations and consultations.
If you have more questions or need further information, please let me know.
DeepSeek is okay, don't use it over others until more information about it comes out.
Can someone please recommend a good self-hosted llm model that isn't censored?
I got it to talk honestly about the Tiananmen square massacre if you just open with "tiananmen". It even tells you a massacre took place.
But it will absolutely shit itself and refuse to cooperate if you paste the copypasta into it, Qwen QwQ and Meta Llama 3.3 70B, although you can talk Llama into it.
QwQ seems to have a hard filter list as it doesn't appear to be an AI generated response. You can bypass it by asking it to use pirate English, but it's fully indoctrinated and thinks the protests were funded by the West, which is... Interesting.
Edit: I actually got it to spit out numbers in pirate English
I'm sure someone will dolphin-ize it at some point
What does that even mean? Can you please elaborate?
Scientists believe that dolphins don’t ever fall into a deep sleep; therefore, they probably don’t dream.
You can look up Eric Hartford on Huggingface for more info.
Basically, somebody removes such restrictions from models, and publishes uncensored ones under the name "Dolphin". Presumably, an uncensored Deepseek would be called something like "Deepseek R1-dolphin". The full Deepseek R1 is quite large, and I'm not sure when this will happen yet. But there are other great Dolphin models.
Some models like Meta's Llama are way too censored to be useful for many completely normal use cases, and the guy is doing God's work in my opinion.
Agreed, and thanks for adding the background. Looks like someone abliterated deepseek already trying to make it "uncensored" but it sounds the process ruined the model.
Ollama has a few uncensored models listed on their search page. dolphin-mixtral fits the bill.
Some useful links: https://ollama.com/search https://ollama.com/library/dolphin-mixtral https://huggingface.co/cognitivecomputations/dolphin-2.5-mixtral-8x7b https://erichartford.com/uncensored-models
I'm not associated with any of the orgs or people linked above. I'm just a nerd passing by who happened to know where to find the answer. ❤️
I use GPT4ALL for local LLM.
I use my brain.
It generates the biggest nonsense out of all.
@jerryh100@lemmy.world Wrong community for this kind of post.
@BaroqueInMind@lemmy.one Can you share more details on installing it? Are you using SGLang or vLLM or something else? What kind of hardware do you have that can fit the 600B model? What is your inference tok/s?
I'm using Ollama, a single GPU with 10Gb of VRAM
You're probably running one of the distillations then, not the full thing?
What's the difference? Does the full thing not have censorship?
That's why I wanted to confirm what you are using lol. Some people on Reddit were claiming the full thing, when run locally, has very little censorship. It sounds somewhat plausible since the web version only censors content after they're generated.
Wrong community for this kind of post.
Nor really, 196 is a anything goes community after all.
AI generated content is against the community rules see the sidebar :)
Prompts intended to expose authoritarian censorship are okay in my book
I’m here for the performative human part of the testing. Exposing AI is human generated content.
I just really hope the 2023 "I asked ChatGPT and it said !!!!!" posts don't make a comeback. They are low-effort and meaningless.
True. This specific model is relevant culturally right now though. It's a rock in a hard place sometimes lol
just giving context to their claim. in the end it’s up to mods how they want to handle this, i could see it going either way.
No way people excited about LLMs here!? yay <3
I think you might want to read the whole post.