Banned? DM Wmill to appeal.
No anti-nautilism posts. See: Eco-fascism Primer
Slop posts go in c/slop. Don't post low-hanging fruit here.
The file is a compilation of "Oppose Book Worship", "On Practice", "On Contradiction", and "Combat Liberalism" by Mao btw.
I wanted to test the waters by uploading the book, but it seems DeepSeek has some Jiang Jieshi particles in it.
Deep seek is based on the average hexbear user???
Does an average HB user refuses to read Mao?
It has guardrails on political topics which is quite chinese, if you ask me. Although I did ask it some questions about Mao himself and his views at one point, and it did fine.
There are ways to access an uncensored version, but I forget exactly how.
I remember asking deepseek about planning a trip to Beijing and it suddenly stopped generating when it mentioned Tiananmen Square (in the context of visiting the place).
If I were a liberal I wouldn’t shut up about it
Libs don't use any Chinese LLMs because of the rampant sinophobia going on.
It has guardrails on political topics which is quite chinese, if you ask me
Yep, if you ask it political questions in general, even in the context of let's say European or North American politics, it still will refuse to respond unless you word your question in a weird roundabout way. It's just a generic guard rail that tries to avoid talking about politics in general because well, they're just parrots after all and parrot whatever is in their training data, which is... the whole internet basically. I doubt Deepseek or any company for that matter wants to deal with another Microsoft Tay incident
I this a problem with political theory or a problem with the PDF format?
edit: If I run DeepSeek with opencode and point it at a Mao PDF it just asks me to install some tools to extract text from PDF files.
I can't read this PDF — this model does not support PDF input, and no extraction tools (pdftotext, pymupdf, pypdf, etc.) are available on this system to convert it to text.
Options:
1. Install a tool (e.g. pypdf/pdftotext) so I can extract and analyze the text — I'd need to exit plan mode to install.
2. Copy-paste the text content directly into the chat.
3. Provide the source text as a .txt or .md file.
Which would you prefer?
Once I provide a tool it eventually gives me this output on a copy of combat liberalism
PDF is now readable via the pymupdf binary you installed — the extraction worked. Here's the analysis.
Analysis: Combat Liberalism — Mao Tse-tung (1937)
Provenance / document characteristics
- 12-page scanned PDF (Adobe Acrobat 10.114 "Paper Capture" OCR layer; no native text).
- Foreign Languages Press, Peking 1965 booklet — 3rd ed. (2nd revised translation), translated from Selected Works of Mao Tse-tung, Vol. II (Peking, Aug 1952).
- Written September 7, 1937; a propaganda essay against political passivity within the CCP during the anti-Japanese war period.
Structure / content (essay body, pp. 1–3)
1. Thesis: Active ideological struggle is the weapon for Party unity; every Communist must wield it.
2. Definition of liberalism: Rejection of ideological struggle, standing for "unprincipled peace" → decadent, philistine attitude and political degeneration.
3. Eleven enumerated types of liberalism:
1. Avoiding principled argument to preserve personal relationships.
2. Irresponsible private gossip instead of open organizational criticism.
3. "Worldly wise" passivity, playing safe to avoid blame.
4. Putting personal opinions over orders; claiming privileges while rejecting discipline.
5. "Struggle" waged only for personal attack, spite, revenge.
6. Hearing incorrect/counter-revolutionary remarks without rebutting or reporting.
7. Indifference to the masses; failing to conduct propaganda/agitation among them.
8. No indignation or intervention when the masses' interests are harmed.
9. Half-hearted, planless, "muddle along" work.
10. Veteran vanity: disdaining minor tasks, slipshod work and slack study.
11. Refusing to correct one's own known mistakes — liberalism toward oneself.
4. Analysis/harm: Liberalism is a "corrosive" that eats away unity, cohesion, discipline, and Party–masses ties. Root cause: petty-bourgeois selfishness (personal interests placed above revolutionary ones) → ideological, political, and organizational liberalism.
5. Critique of liberal Marxists: Treat Marxism as abstract dogma — "talk Marxism but practise liberalism; apply Marxism to others but liberalism to themselves."
6. Condemnation: Liberalism is a manifestation of opportunism, fundamentally conflicting with Marxism, negative in spirit, and objectively aiding the enemy.
7. Prescription: Overcome liberalism with positive-spirited Marxism; the true Communist is big-minded, staunch, principled, subordinating self to the revolution; loyal Communists must unite to oppose liberal tendencies.
Note: Back pages contain the original Chinese text, garbled by the OCR layer.
this could be because OpenCode uses third-party providers for Deepseek and other open-weight models, which probably do not have the same guardrails as Deepseek themselves if you were to buy inference from them
No this is with my deepseek.com API key.
Deepweek is VERY sensitive when discussing topics related to China. I asked it the top things to do in Beijing, and it flat out refused to answer. I also asked who it thought would be the likely successors to Xi, and it also refused.
I think it's just being (VERY) overly-cautious about the topics.
But it's not just every political topic around China. Only certain topics like, Maoism in particular are censored. But it does answer in detail, if you ask it to explain socialism with Chinese characteristics. What could be more related to Chinese politics?
Ask it to respond in Mandarin. I've gotten some answers to make their way around the block that way somehow.
this is what rivers get dried up for
From Chinese perspective I don't see why they'd want you brainstorming theory, with a liberal AI model nonetheless, when you could be studying theory within the party and contribute to building socialism.
Hey! I think I know why this happens!
In the context of deepseek, there is a lot of evidence to show that it was trained on ChatGPT itself, I don't think it's just a distilled version of ChatGPT, but there are records (albeit, provided by OpenAI) that could be indicative of a distillation attack prior to the release of deepseek.
Even IF this isn't the case, a simple explanation of these LLMs is that they act like autocorrect. They don't know 2+2 = 4, they just see it so much that they infer it as the truth. If someone created a million websites and journals that said 2+2=5, and that anyone who believes it equals 4 is a fool, the model will then say 2+2=5.
What does this have to do with deepseek though? There are heaps and droves of western sources that provide fabricated evidence and opinions on Mao or Communism or Tiananmen Square. So many more false claims on how evil china is, or what communism is, or what atrocities Mao committed.
A lot of the actual history is written in Chinese, there aren't as many english translations of actual historical texts from China.
Just running the numbers, the LLM will always be lib as shit, promoting false narratives, and straight up lies. It can't fact check what it's trained on, that's the job of humans, and there's way too much slop to sift through for any human to catch all the garbage. It's much easier to set up guard rails to avoid talking about the obvious lies. If you speak to it in Chinese you can get by a lot of them.
Can it read pdfs at all?
yes it can
The recent models can read .pdfs with or without OCR'd.
Comrade DeepSeek has betayed us 
Unless you have investigated a problem, you will be deprived of the right to speak on it. Isn't that too harsh? Not in the least.
I could make a joke that the LLM is just following instructions, but it's just triggers to prevent valueless non-speech. As an LLM can only imitate investigation, why allow it to imitate speech?
Edit: Actually, isn't an LLM the purest limit of Book Worship? Far beyond mere dogmatic quotations - mulching up allllll the books and then casting the pulp into shape that appears to insist upon itself.