I believe that "Zhihu" used to be one of the most valuable websites in terms of content in the Chinese internet. Although its quality has inevitably declined and self-censorship has become increasingly severe in recent years, its foundation of knowledge Q&A and the essence of a community for seeking knowledge still make it outstanding in Chinese content. Compared to WeChat, it is still a world of difference. However, this deteriorating trend is difficult to reverse.
In today's Chinese internet, there are fewer and fewer websites that can be read, which is a very regrettable thing. There are many reasons for the decline in good Chinese content. On one hand, there is an increasingly strict censorship system, and on the other hand, there is an overflow and pollution of paid marketing content. WeChat public accounts may now dominate most of the Chinese internet content, but it is even more strict and unfair than government censorship, filled with marketing content and false information. Why can "Zhihu" still maintain a certain level of information in such a big environment? I think its organization as a "knowledge Q&A community" has inherent advantages in terms of information.
The knowledge Q&A community is an excellent form of internet content
A question-and-answer community is a combination of a knowledge base system and a virtual community. It has been proven to be a very good form of content organization.
Quora is one of the pioneers of this form (although not the earliest), while Zhihu basically started by copying Quora.
Stackoverflow is a question and answer community focused on software technology.
Apart from the question-and-answer format, forming a community around content links is very similar in terms of information structure. Sites with this structure started with Del.icio.us, Digg, and now the typical representative is Reddit.
News Picks from Japan is not a knowledge Q&A platform, but rather closer to Reddit.
Why does the knowledge Q&A community have these advantages?
The knowledge Q&A community is a typical "social network" with content as its main theme, but with strong social characteristics. Users can participate extensively, answer questions, supplement content, express different opinions, and evaluate other people's views, which greatly promotes the democratization of knowledge and prevents information silos. For example, despite the serious erosion of the Chinese internet, as long as there is a certain filtering ability, valuable content can still be obtained on Zhihu. On the contrary, on portal websites, media sites, and WeChat public accounts, there may be no value at all.
Blogs and portal/news websites were once very popular, but have gradually declined in the past 10 years. Knowledge Q&A communities and media websites have several notable advantages compared to blogs and other platforms:
- Market effect：The output of a blog is often one-way, and what to write depends on the author's plan and mood. On the other hand, a Q&A community is more like a "market," often with a higher demand for content and a greater supply, forming a positive feedback loop.
- Cumulative effect：The blog itself has "social networking" characteristics, but it mainly focuses on individuals, with weak social features. This often leads to the voices on the blog being only those of a person and their "friends," making it difficult to find more voices on the same topic.
- Diversification of perspectives.：The "community" characteristic of a knowledge community determines that its content is diverse, so here you can gather more information, hear different voices, suitable for studying a new technology or observing a social phenomenon.
- More targeted themes：Compared to social networking sites like Twitter and Facebook, knowledge Q&A communities often have more substance and are not completely irrelevant.
The combination of knowledge-based question-and-answer community and large language models is innate
The question-and-answer community and large language models are a perfect match. It is said that OpenAI's earliest GPT model, GPT-2, was primarily trained on content from the Reddit community that had received at least two upvotes. Additionally, much of the information on Stackoverflow serves as the foundation for the programming capabilities of current large language models. The content of knowledge question-and-answer communities can provide a foundation for large language models and can also have more points of integration with them (what are these points of integration? This is an exploratory space).