How Recommendation Algorithms Are Causing People to Lose Their Active Search Skills
Recently, I’ve frequently encountered discussions about AI search, such as people’s declining active search abilities and the decreasing amount of content on Chinese search engines. Therefore, I wrote this article to share my search experiences.
01
Currently, Google cannot retrieve content from Zhihu, and searching for Chinese content feels like navigating a desert — there’s a lack of genuinely valuable content, mostly just basic definitions and a few Chinese websites. As a result, I haven’t browsed Zhihu for over a year. Since I lost access to my registered account due to an old phone number, I chose to let it go.
After a year, I realized that not using Zhihu didn’t make much of a difference; other internet platforms provided enough information.
This is partly due to the “island” phenomenon on the Chinese internet. Apart from browser plugins like seek AI, seek All supports the following platforms:
- New-generation AI search engines like Perplexity, Kimi, Mito, Consensus
- Conversational AIs like ChatGPT, Zhipu, and Wenxin Yiyan
- Traditional search engines like Google, Bing, DuckDuckGo, Yandex, Yahoo, 360, and Baidu
- Independent content social platforms like Xiaohongshu, Facebook, Bilibili, X, YouTube, Product Hunt, Zhihu, Weibo, and Sogou WeChat
These platforms can serve as reference information sources for active searches, allowing you to switch between platforms to obtain the most comprehensive content.
In fact, platforms like Zhihu and Xiaohongshu selling data to new-generation search engine companies might be a good thing. Now, using Tiangong, you can search not only for web content but also for Xiaohongshu content. If you want to search for Zhihu content, you can use Mito.
From my search habits, this hasn’t made much of a difference. These new AI search engines, like Google before them, serve as my first step in basic searches. The second step involves platform-specific searches on Xiaohongshu, WeChat, etc., based on keywords.
This step is usually unnecessary unless I’m delving into a completely unfamiliar field. Besides, most of the content I’m interested in is in English. For Chinese content, platform-specific searches suffice, as these major platforms also support web versions.
Sometimes, I solve the first step directly with ChatGPT. OpenAI’s new product, SearchGPT, is currently in the Waitlist phase. Last year’s discussions about hallucinations in large models haven’t been as relevant this year, thanks to plugins like Web browser.
For those with strong active search abilities and curiosity, the decreasing content on Chinese search engines doesn’t make much of an impact. Expert searchers have their own unique workflows and techniques.
Here are three AI information search experiences I’d like to share:
Searching Based on AI Papers:
Reading papers feels a bit like opening a mystery box to me, as I’m a beginner with no prior experience. I started from scratch, downloading all the top-ranked or recommended papers, piling them up, and randomly reading them daily. If I find a particularly promising direction, I’ll search for related content on Google and WeChat.
I’ve been fortunate — just as I started understanding Graph RAG, I came across a paper on GNNs, which led me to search and learn about GNN and GAT as subtopics. For professional AI technology, searching on WeChat often yields valuable information. Even if it’s not direct information, it might lead to information sources, such as articles summarizing GNN papers, which then direct me to download papers from arXiv.
For AI papers and foundational AI knowledge, WeChat Reading is another useful tool. You can search for books or sections containing keywords. For example, I once searched for gradient descent, but ultimately gave up because WeChat Reading only supports mobile searches, with no web version.
Following Active AI Tech Bloggers:
Most people get the latest AI news from bloggers they follow. Many tech bloggers, whether on BiliBili, video platforms, or Weibo, update daily with the latest AI news. These accounts are very active and focused on vertical markets, sharing the latest AI products and technologies. You can follow these bloggers, treat them as information sources, and further explore the original sources.
If you dig deeper into how these tech bloggers obtain their information, you’ll often trace back to platforms like Musk’s X.com or popular AI websites abroad.
Searching for Chinese content on Google can be challenging since domestic social platforms don’t share their content. However, for English content, there’s an abundance of resources — Instagram, YouTube, and various websites. I mainly use Google for searching English content.
From what I’ve seen, AI news bloggers primarily operate on WeChat Public Accounts, video platforms, and Jike. BiliBili video bloggers tend to focus on brief paper reviews and product deployments. If you’re looking for this kind of content, BiliBili is a better choice.
Jike is a niche app, initially popular among marketers, investors, and product managers. Now that AI is trending, many AI professionals have joined, and even traditional industry professionals have started exploring AI.
Personally, I’m not exactly sure how I keep up with the latest AI products. I don’t actively follow these tech bloggers — it might be because I’m immersed in the AI industry, where people around me are always sharing the latest AI developments.
Or, I might just search randomly, focusing on what I’m writing about or trying to learn more about. For instance, I’m currently interested in Graph RAG and Graph Transformer, so I’m searching for related papers and products. As these products become more popular, I naturally come across the information. I also frequent tech forums where people occasionally post AI-related content.
After reading the latest AI news, it’s advisable to search for the original sources and further explore how others interpret the AI product or technology. This will provide a more comprehensive understanding.
Searching Based on Current Needs:
This is the simplest and most basic approach. A couple of weeks ago, I wanted to improve my prompt skills, so I searched for the most authoritative and professional content on prompts. Having a search theme leads to faster progress.
Epilogue
Searching for knowledge is part of the learning process. We consume the results of others and refine them through output into our own insights, much like photosynthesis and respiration in plants — consuming and producing oxygen while accumulating nutrients for continuous growth.
by: pamperherself
