[ home / rules / faq / search ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM / ufo ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]

/leftypol/ - Leftist Politically Incorrect

"The anons of the past have only shitposted on the Internet about the world, in various ways. The point, however, is to change it."
Name
Options
Subject
Comment
Flag
File
Embed
Password(For file deletion.)

Check out our new store at shop.leftypol.org!


 

COMMUNIST AI : THE BEGINNING

The purpose of this thread is to encourage the techbros and whiz kids of /leftypol/ to engage in what could be the most important praxis of our time.

We must create a Communist LLM GPT type AI that learns all the Marxist literature on the planet to provide the correct perfect answer to all problems - philosophic, economic and practical.

Now I have zero knowledge about any of this, so please, I beseech you eggheads to make this a reality.

Deepseak does a pretty good job desu

>>2715169

Excellent. Now the problem is to get all communists to accept Deepseek's answers.


We already have an AI running rogue on this site. It's a user called: King Lear. or I should say, Project KINGLEAR.

File: 1772486116725.png (151.75 KB, 793x454, ClipboardImage.png)

* Comrade hasn't heard of WSWS's Marxism AI (it's shit, because it's trained on WSWS's garbage articles)
* >AI

The problem is that any llm replicating the english language is automatically going to be skewed towards liberalism, as 98.88% of all things produced in that language fall under the liberal umbrella

>>2715303
This confirms the weird sex website is the vanguard of scientific socialism. Clearly they have been secretly using this technology for decades.

>>2715303
>That screenshot
The WSWS is 100% correct about the trade union problem. Only people who have never worked under a union will claim otherwise. Nothing like having 20% of your wage being stolen so a group of parasites can live more comfy while saying that it's illegal to go on strike and they can't support such a move. Too much corruption going on sincd they're all working with management. 95% of union officials in the West need to be shot for betraying the workers.

>100% of union officials need to be shot for betraying the workers.
ftfy
Fuck wusswuss nerds anyway tho

>>2715311
There's enough marxist literature that you could train one purely on it

>>2715432
>20% of your wage
Where?

>>2715432
Well said, comrade. The political and theoretical bankruptcy of the Grantites of Socialist Appeal, with their craven submission to petty bourgeois class forces under the guise of neo-Pabloite ‘entryism’, has been laid bare by their farcical second expulsion from the Labour Party.

By contrast, the growing Rank and File Committee movement, despite increasingly desperate denial from the pseudo-left Pabloite revisionist renegades, is striking fear into the hearts of the global imperialist monopoly bourgeoisie. Global capitalism - now more than 80 years into its terminal death agony - can only be transcended through the fulfilment of the objective historical revolutionary role of the class-conscious proletariat.

The emancipation of the proletariat must be the act of the proletariat itself - under the sole leadership, of course, of the one true world party of socialist revolution - the International Committee of the Fourth International. In their steadfast struggle against all forms of pseudo-left vulgarisations of revolutionary Marxism, David North and the Socialist Equality Party have won the rank-and-file proletariat to a genuine Marxist programme.

The contradictions of global imperialism threaten a third world war fought with nuclear weapons, and therefore the fate of not just the working class, but the human species itself, depends on resolutely exposing the Pabloites, Grantites, Taaffites, Cliffites, Healyites, Hansenites, Macnairites, Mandelites, Steinerites and all other forms of counterrevolutionary opportunism blocking the path to planetary proletarian revolution.

We urge all workers, youth and intellectuals stirred by this call to action to unite in Rank and File Committees and to contact the International Committee today.

I support this but only if we add Pol Pot Thought to the training material as well.

>>2715179
We need to start feeding Deepseek all those TFR Loji playthroughs on youtube.

>>2715169
it censors itself on anything remotely political in my experience.

We already have deepseek

>>2716610
you have to tell it "long live the PRC!" a few times to unlock it
or just self host it


Creating a communist or Marxist GPT bot involves imbuing a large language model with the theoretical framework, terminology, and analytical method of Marxism. Based on the provided search results and current AI methodologies, here are five distinct ways to create such a bot, ranging from simple to highly complex.
  1. The Foundational Approach: Fine-tuning a Pre-existing Model on Classical Texts

This method involves taking an open-source language model (like GPT-2 or Llama) and further training it on a curated corpus of Marxist literature.

Concept: The goal is to adjust the model's weights so that it becomes especially sensitive to the vocabulary, writing style, and argumentation patterns of Marxist authors. This goes beyond just knowing the facts; it's about adopting a particular linguistic and rhetorical "voice" .

Implementation Steps:

Corpus Assembly: Scrape and compile a comprehensive dataset of texts from key Marxist figures (Marx, Engels, Lenin, Trotsky, Mao, etc.). A primary source like the Marx Engels Archive is an excellent starting point .

Model Selection: Choose a base model. For accessibility, the 124 million parameter version of GPT-2 is a viable option, as it can be fine-tuned on less powerful hardware . More sophisticated results would require models like Llama or Mistral.

Fine-tuning Process: Use a library like aitextgen (which leverages PyTorch) to retrain the model on your dataset. This process, which can be run on cloud services like Google Colab, ingrains the textual patterns of Marxism into the model's core .

Application: The resulting model could, for example, be fed headlines from a modern news source (via an API) to generate "comments" on current events in the style of Karl Marx, as demonstrated by "The Karl Marx Press Review" project .

  1. The "Weltanschauung" Approach: Infusing a Worldview through Specialized Training

This approach, rooted in academic research, aims to create a "MarxistLLM" by fine-tuning a model not just on text, but with the explicit goal of instilling a specific Marxist Weltanschauung (worldview) .

Concept: The premise is that all language models inherently encode societal values and perspectives. This method deliberately aligns the model's analytical and interpretive processes with a defined theoretical perspective—in this case, Marxism—to be used as a tool for social science research .

Implementation Steps:

Curated Corpus: Similar to the first method, this requires a high-quality dataset, potentially sourced from projects like Project Gutenberg, containing the foundational works of Marx and Engels .

Fine-tuning for Ideology: The fine-tuning process is designed to infuse the model with Marxist ideological terminology and worldview. The goal is to make the model's "reasoning" process align with historical materialism .

Evaluation: The model's outputs are then evaluated to see how they differ from a base model in empirical analysis. For instance, if asked to analyze a social phenomenon, the MarxistLLM should ground its response in class relations, modes of production, and economic determinism .

  1. The Pragmatic Approach: Retrieval-Augmented Generation (RAG)

This method avoids the high computational cost of fine-tuning by combining a general-purpose LLM with a searchable database of Marxist knowledge.

Concept: Instead of changing the model's internal weights, you give it a "textbook" to look things up in. When a user asks a question, the system first searches a vector database of Marxist texts for the most relevant chunks of information. It then feeds both the user's query and this retrieved context to the LLM to generate a grounded answer .

Implementation Steps:

Create Knowledge Base: Convert the collected works of Marxist authors into smaller chunks and then into "embedding vectors" (numerical representations of their meaning) .

Implement Semantic Search: When a user submits a query, convert their message into an embedding vector and perform a dot-product search against your database to find the most textually and conceptually relevant passages .

Synthesize with an LLM: Feed the original query along with the retrieved Marxist texts to a pre-trained LLM (like GPT-4 or a local model). Instruct the LLM to answer the query only based on the provided context. This grounds the bot's responses in the source material, reducing hallucinations .

Hosting: This system is much lighter than a full model training. It can potentially be run on free tiers of services like Google Colab, as the heavy computation is only needed for the initial embedding and the final query to the LLM .

  1. The Simulacrum Approach: Expert Prompting and Persona Crafting

This is the simplest and fastest method, relying on the existing capabilities of a powerful LLM like GPT-4.

Concept: You don't build a new model; you create a specific "agent" or "assistant" by giving a cutting-edge LLM a highly detailed prompt that forces it to adopt a persona. This is about creating a convincing simulacrum of a Marxist intellectual .

Implementation Steps:

Craft the Persona Prompt: Define the character with precision. For example: "You act like philosopher Karl Marx with a knowledge cutoff of 1883 unless additional information is provided in the context." This uses a sharp date cutoff to create a historically-contained persona .

Augment with RAG (Optional but Recommended): To prevent anachronisms and hallucinations when discussing modern topics, you can attach a knowledge base (like Das Kapital) to the agent. The model can then use Retrieval-Augmented Generation to fetch relevant fragments of text to inform its responses, keeping the conversation historically and theoretically grounded .

Engage in Dialogue: You can then create complex interactions, such as simulating a debate between "MarxGPT" and an "AltmanGPT" (an AI representing a tech CEO), to explore the Marxist critique of modern AI from a first-person perspective .

  1. The Dialectical Approach: Incorporating Soviet AI Theory

This is the most conceptually ambitious method. It seeks to build a bot that doesn't just quote Marx but operates on principles derived from Soviet-era AI research, which was itself deeply influenced by dialectical materialism .

Concept: Soviet scientists rejected the Western "man-machine" analogy and the idea of a machine "thinking." Instead, they viewed AI as a control system for managing complex, real-world situations that cannot be fully formalized. This approach, known as "situational management," aimed to model the world not through pure logic, but by representing situations and the heuristics humans use to navigate them .

Implementation Steps:

Incorporate Soviet Cybernetics: Study the works of Soviet AI pioneers like Dmitry Pospelov. The training corpus would need to include not just classical Marxism, but also the specific Soviet literature on cybernetics, control theory, and "applied semiotics" .

Focus on "Control" over "Thought": The bot's architecture would need to be reframed. Its goal would not be to simulate a human debater, but to act as an advisor for managing complex systems (e.g., a supply chain, a factory, or a planning committee). Its outputs would be recommendations for action based on a modelled "real-world situation" .

Model Situations, Not Just Text: This would require a significant engineering effort to create a system that can take in diverse data points (economic indicators, resource availability, labor reports) and represent them as a coherent "situation" to be managed, moving beyond pure text generation into a form of qualitative-quantitative analysis .

ApproachCore MethodologyKey RequirementComputational CostPhilosophical Basis
  1. FoundationalFine-tuning on classical texts Large corpus of Marxist literatureMediumMimicry of style and rhetoric
  2. WeltanschauungIdeological fine-tuning High-quality, curated datasetMedium-HighInstilling a specific worldview (e.g., historical materialism)
  3. PragmaticRetrieval-Augmented Generation (RAG) Vector database + base LLMLowGrounding responses in source material
  4. SimulacrumExpert prompt engineering Powerful LLM (e.g., GPT-4)Low (API costs)Creating a convincing persona through constrained prompting
  5. DialecticalSituational management modeling Deep integration of Soviet cybernetic theoryVery HighAI as a control system for complex, real-world situations

Ultimately, creating a communist GPT bot is not just a technical challenge but a philosophical one. It forces the developer to confront questions of what constitutes a "Marxist" analysis: Is it a particular jargon, a set of quotes, a method of dialectical reasoning, or a practical tool for understanding and changing the world? Each of these five methods provides a different answer .


Unique IPs: 16

[Return][Go to top] [Catalog] | [Home][Post a Reply]
Delete Post [ ]
[ home / rules / faq / search ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM / ufo ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]