这个项目是基于GPT技术，使用Llama和Alpaca模型根据编程随想的文章和内容进行训练，生成一个他的虚拟分身。通过这个项目，我们希望实现编程随想的数字永生，保留他在网络上的影响力和思想贡献，使其可以与人们长期交流和互动。
编程随想是一个中文博客，创立于2010年。博客最初主要涵盖计算机技术和编程方面的文章，随着时间的推移，其议题逐渐扩展至社会政治、心理学、历史学等多个领域。博主在博客中普及翻墙和网络安全知识，也提供了大量电子书籍的下载。 然而，编程随想的博客自2021年5月9日以来再未更新。据后来证实，其真实身份为阮晓寰，已被中国当局逮捕审判。因此，我们希望通过这个项目，保留编程随想在网络上的影响力和思想贡献。
编程随想是一个在网络上有一定影响力的博主，他的博客内容包含了多个领域的知识和见解。然而，由于他的身份被揭露并被逮捕，他无法再维护和更新博客，这意味着他的思想和知识将不再得到更新和扩展。 通过使用GPT技术和编程随想的文章进行训练，我们可以生成一个虚拟的编程随想，保留他的思想和知识，并让人们可以与他进行长期的交流和互动。这将有助于延续编程随想的影响力和思想贡献，也可以为后来者提供有价值的知识和启示。
This is a project that aims to use GPT technology to create a virtual version of "Programthink" (编程随想) - a Chinese-language blog run by an anonymous author who was arrested in May 2021. It was alleged later the author is 阮晓寰 (Xiaohuan, Ruan)
The blog covers a wide range of topics, including computer science, network security, psychology, history, and politics, and is well-known for providing information and resources on circumventing the Great Firewall of China.
The purpose of this project is to create a digital "avatar" of Programthink, using GPT-3 technology to generate content that emulates the style and topics of the original blog. By creating a virtual version of Programthink, we hope to preserve the ideas and spirit of the blog, and enable it to continue to be a valuable resource for individuals interested in technology and free speech.
The arrest of the author of Programthink represents a loss of an important voice in the Chinese-speaking community, and highlights the need for alternative ways of preserving and disseminating information that might be targeted by censorship and surveillance.
The use of GPT-3 technology offers a powerful tool for generating high-quality natural language content that can mimic the style and tone of a particular author or publication. By creating a virtual version of Programthink, we can ensure that the blog's legacy and insights will continue to be available to future generations of readers and thinkers.
This Github repo includes the trained Llama and Alpaca GPT-3 models, as well as scripts for generating content based on the original Programthink blog.
We would like to acknowledge the work of the original author of Programthink, as well as the contributions of the open-source community in developing and improving GPT technology.