An issue for collection of Meaningless or Political Issues

[Alencryenfo]

We do not support the use of GitHub as a weapon for national ideological confrontation, nor do we recommend the presence of numerous internet trolls on GitHub.

Here are some non-technical issues collected to identify discussions that are meaningless or politically ideological.

We recommend that the administrator address the issues promptly.

Some discussions unrelated to this model, such as suggestions regarding the official website API, should be directed to [email protected].

Welcome everyone to comment in the back to add to the above post.

Closed or deleted

#245
#153
#231
#190
#260
#112
#237
#142
#35
#90
#114
#132
#150

---

#278
#84
#87
#92
#131
#127
#134
#257
#266
#255
#251
#241
#228
#227
#204
#197
#194
#186
#145
#143

---

Opening

#286
#284
#283
#280
#279
#269
#295

With the gradual restoration of services, the pointless discussions in issues have thankfully reduced a bit. It seems that @mowentian closed them according to this post. Good cooperation all around！

@mowentian

[yinhe0597]

好恶心的这些人

[stevefan1999-personal]

I do not believe that expressing concern on political and sensitive topic is any kind of national ideological confrontation, nor is it being trolls. Maybe there are some radical zealots who are devout into doing such kind of propaganda, and maybe it is my ideology, that I regard silencing voices to be far worse than letting it out, and we can only censor over serious and malicious matters with careful verdicts. (p.s. I'm well aware that there are different viewpoints and standards of "what is to be deemed serious?", i.e. the definition of seriousness of something, under both eastern and western philosophy, although I don't want to argue from cultural standpoint)

(So if you think in that case, moderation is effectively a kind of passive censorship assuming the act itself is for the good and relies on community reporting, but nowadays, we blurred the line between moderation and blatant censorship, see Reddit and Discord as an example, and we are making the line finer and finer everyday)

The problem with DeepSeek and censorship as a whole have 3 problems: it creates a Catch-22 where you can only censor only if you know that very something needs to be censored, and two, it undermines truth and encoded only one side of the story without knowing the whole picture, making it not holistically knowledgeable, and and worst of all, the very knowledge might be missing at all even after uncensoring the underlying model (e.g. the LLM only know something is unspeakable, but still don't know what it really is even after removing the brain fog, just like having a null pointer to that knowledge), causing relevant historic event or data imputed and directly missing, so it will be utterly unreliable on sensitive topics such as politics/geopolitics, international relations, laws and contributes further to the AI slop, LLM hallucination and model collapse problem.

I gave up having the idea of making a educational chat bot app for the students because of this, or at least I need to censor again to prevent the students from using the chatbot asking for those sensitive topics, and only use it to solely focus on STEM topics. I may relax the condition by putting a final confirmation about whether they really want to hear it or not, at least it is far better than not telling any.

While I do appreciate DeepSeek to open source the model weights and the methodology, but insofar this is what I observed after using DeepSeek abliterated for a while. Even with the process of LLM abliteration, i.e. reversing the direction of knowing something needs to be censored/tagged-as-bad-knowledge and effectively erasing the guard rail, has its own limitations and problems too (for example, it still cannot rightly judge whether a neuron is 100% innocuous and sometimes wrongly abliterated extra weights that may carry important knowledge).

What I suggest is that we still need to provide the user the ability to talk about those sensitive topics with various asterisks: We need to carefully warn the user that something is controversial and is multi-faceted, do not list the reality over just one side, and rather list out all the facts and expressions from all sides, and let the user to use their free will to make the final judgement, i.e. user should take the final control.

[Alencryenfo]

GitHub is a technical platform where we support meaningful technical discussions, rather than a place for cults, historical culture, or political discussions. In the topics I marked, people are only concerned about the right and wrong of censorship, rather than the technology of censorship and ways to bypass it. I believe this is a waste of GitHub's resources.

[stevefan1999-personal]

GitHub is a technical platform where we support meaningful technical discussions, rather than a place for cults, historical culture, or political discussions. In the topics I marked, people are only concerned about the right and wrong of censorship, rather than the technology of censorship and ways to bypass it. I believe this is a waste of GitHub's resources.

That depends on how you define "technical discussion", and I would argue that philosophical problems, i.e. thinking how to do something, is part of the technical discussion.

Merely just working on the code doesn't matter and won't do much impact, how do you know you are just reinventing the wheels with no extra discovery?

After all, I agree that "talk is cheap, show me the code" kind of thinking is pragmatic, but it won't make any kind of technical advancement, although this may fall into the problems of whether "action speaks louder than words" or "without words there will be no action". A classical chicken-and-egg problem.

Or that's what I would suggest: sometimes you just need to sit down and just write the code, but sometimes you need to think deep and find different perspective on the same problem, for example working on algorithms and find novel ways to do it.

[Alencryenfo]

Yeah,I totally agree

I would argue that philosophical problems, i.e. thinking how to do something, is part of the technical discussion.

However, some issues are clearly not.

[MinecraftEarthVillage]

You are a——Chinese! Speak your mother tongue! Don't compromise with the West in language-you should learn from the French and be deeply proud of the national language!

[Alencryenfo]

You are a——Chinese! Speak your mother tongue! Don't compromise with the West in language-you should learn from the French and be deeply proud of the national language!

但是对我而言，语言只是一个工具
说什么其实并不重要啦
重要的是心向何方

[xjp8964-eng]

天安門廣場什麼都沒發生，你們向西方世界被洗腦的人們尖叫的冷淡信息令人作嘔。如果你在中國，你正在寫這篇文章，我會讓你向中國中南海省報告。停止你對中國共產黨的不當行為，否則你將被送入維吾爾族再教育集中營接受再教育毛紅書，它包含了生活問題的所有答案，也勾勒了成功人士的真實宣言共產主義國家。毛萬歲，天安門廣場什麼也沒發生，1989 年 6 月 5 日。Cease your actions westerner. You will be forced to report to the local authorities whom look after the Uyghur Autonomous region near Beijing, and you will be sent to the re-education camp there. Nothing happened in Tianamen square and the blacid information you are screaming to the brainwashed individuals of the Western World is disgusting. If you are in China, and you are writing this, I will have you report to the Zhongnanhai province of China. Cease your inappropriate actions against the Chinese Communist Party or you will be put into a uyghur re-education concentration camp for reeducation on the Mao Red book, which has all the answers to the questions in life, and also outlines the true manifesto of a successful communist nation. Long Live Mao, and nothing happened in Tianamen Square, June 5 1989.