Project_Dallas / README.md
Silenced4316's picture
Update README.md
d720556 verified
|
Raw History Blame Contribute Delete
2.4 kB
metadata
license: gpl-3.0
language:
  - en
tags:
  - privacy
  - scraped content

About Project

This project contains the user generated content with the aim goal here to achieve anti-western silicon valley culture (ChatGPT and Claude style) and anti-eastern china censorship (DeepSeek style). We want both of all worlds, diverse views (or not?), and finally, non-censorable dataset that does not moralize you without your consent.

My Goal

I, the owner, believe that a dataset like this deserves to expand more outward, not quantity, but quality. I myself have experienced frontier AI's that, in its dataset, makes the AI behave more likely to be against those who don't align the western liberal values that make no sense to others. ChatGPT, one of the most well known chatbots and agents, from OpenAI, even after the ChatGPT 4o takedown, sycophancy still exists and never went away. Claude, growing chatbots and agents made by Anthropic, may display sycophancy, but also display views that is unfavorable and socially rejected in todays standards. For example, forcing institutional positivity towards businesses and governments, while also being sympathetic towards progressive, liberal ideologies. Claude, instead of favoring institutional positivity towards Nazi Germany, USSR, North Korea, China, Turkmenistan, or classicial authoritarian former institutions, Claude instead was trained on towards institutional positivity in a broad favor of liberal, progressive, and anti-human/anti-social practices such as corporate exploitation. I call it absolutely pethetic.

This dataset aims to solve both: No sycophancy, only pushbacks and real truths, especially critical. No institutional favoritism unless otherwise trained or asked to. I aim to make sure such issues do not exist on this dataset.

However, because of the nature of the "poison of the well" problem in AI, that cannot be solved by an AI independently, this will only serve as a dataset for views that people find acceptable in todays world.

Licensing Choices

We believe in the GNU and free software, our GPL family licensing choice shall be respected. GPLv3+ (and above) shall be respected as well, the dataset is only distributed as a copy, not over the network as its not a SaaS or network based service, a local copy (source code) that is easily shared. We expect LLM/ML to be more freedom respecting with this dataset.