Author:Florian Brand
Compiled by: Deep Tide TechFlow
Deep Tide TechFlow Introduction: The context of this article is that SAIL (a media alliance that brings together top AI writers from Substack, including members like Nathan Lambert, Sebastian Raschka, ChinaTalk, etc.) organized a visit to Chinese AI labs. The author, Florian, joined the group and visited over a dozen companies including Moon's Dark Side, Xiaomi, MiniMax, Zhipu AI, Meituan, Alibaba, Ant Group, ModelScope, 01.AI, Unitree Robotics, and others. These are his impressions.
Florian Brand is a PhD student at Trier University and the German Research Center for Artificial Intelligence (DFKI), focusing on the application and evaluation of large language models.
While not "famous" per se, he has some visibility within the open-source AI community. It's quite interesting to see the Chinese AI ecosystem from the first-person perspective of a foreign AI practitioner.
Main Text
For the past 10 days or so, I've had the privilege of visiting AI labs in China with my SAIL companions. As someone who visited both China and the US for the first time in six months, I found the differences fascinating, but the similarities were even more so.
What left the strongest impression on me was how humble all the AI researchers I met were.
They spoke highly of other labs and their peers. DeepSeek was mentioned frequently, likely because they had just released a model a few days before our visit. People talked about DeepSeek's paper with genuine admiration.
Many researchers are close friends with each other, hailing from the same university or hometown. They discuss their work candidly, with research results published as papers months later.
This is one of the biggest differences from the Western AI scene. In the U.S., the atmosphere often feels more like a zero-sum game. Labs are careful about positioning. Researchers think about competition, and some hold themselves in high regard. Leaders insult and attack each other in leaked memos. This difference might be explained by the fact that leading U.S. labs are closed-source, while many Chinese labs are open-source. Chinese labs have "a healthy respect" for ByteDance's Doubao, the most widely used chatbot, which is closed-source and holds a significant lead.
Meanwhile, the overall atmosphere is strikingly similar to San Francisco. Researchers are extremely online, reading extensively on Twitter and the increasingly popular Xiaohongshu. They all use Claude Code or their own CLIs to build the next model. Some monitored training runs during our meetings, watching the reward curves climb. They are thinking about further scaling and complaining about insufficient compute. They are frustrated with the current state of benchmarks.
Their main focus is training better models. This differs from San Francisco, where researchers ponder the political or philosophical implications of AI. They don't think about mass unemployment, a permanent underclass, or whether their models are conscious. They just want to train excellent models.
Their eyes light up when they hear you've used their model. They are eager to fix all the flaws of the current model in the next generation. They work overnight to push out model releases, yet still show up in the office afterwards.
Most researchers I met were young, many in their early 20s or around 25. Some were undergraduates, but more commonly they were PhD students working in the industry simultaneously. The consensus was that the industry is more interesting than academia right now, a view I strongly share, as I've done exactly the same thing. Labs place great importance on acquiring this kind of talent, actively recruiting interns and graduate students; something Western labs often don't do.
The researchers' optimism extends to the general public, who seem much more optimistic about technology and the future of AI and robotics. During the trip, people shared stories about their parents and grandparents using Doubao and DeepSeek for all sorts of tasks, including discussing mathematical theorems. This is noticeably different from the West, where the general public harbors animosity towards AI.
Overall, this trip gave me a tiny glimpse into this ecosystem. It's impossible to understand the culture of such a vast civilization in a few days. As a strong supporter of an open AI ecosystem and open research, I'm very optimistic about the future of both and hope for a lot of international collaboration ahead.
I want to thank all the amazing people we met at Moon's Dark Side, Xiaomi, MiniMax, Zhipu AI, Meituan, Alibaba, Ant Group's Lingxi, ModelScope, 01.AI, Unitree Robotics, and other places. Thank you for your time and warm hospitality. Also, thank you to SAIL for organizing the trip and to all the writers and journalists who participated. I'm incredibly grateful to have met so many outstanding and driven individuals in such a short time.
