A Western Scholar’s Field Notes from Visiting Chinese AI Labs: Humble, Open-Minded, and Focused on Building Better Models—Not Philosophy
TechFlow Selected TechFlow Selected
A Western Scholar’s Field Notes from Visiting Chinese AI Labs: Humble, Open-Minded, and Focused on Building Better Models—Not Philosophy
The overall atmosphere is surprisingly similar to that of San Francisco, with researchers being highly online—avidly reading on Twitter and Xiaohongshu, the latter growing increasingly popular.
Navigating Web3 tides with focused insightsAuthor: Florian Brand
Translated and edited by TechFlow
TechFlow Intro: This article stems from a delegation organized by SAIL—a media consortium uniting top AI writers on Substack, including Nathan Lambert, Sebastian Raschka, and ChinaTalk—to visit Chinese AI labs. Author Florian Brand joined the delegation and visited over a dozen companies, including Moonshot, Xiaomi, MiniMax, Zhipu AI, Meituan, Alibaba, Ant Group, ModelScope, 01.ai, and Unitree, then wrote this reflection.
Florian Brand is a Ph.D. candidate at Trier University and the German Research Center for Artificial Intelligence (DFKI), focusing on applications and evaluation of large language models.
He isn’t “widely famous,” but enjoys solid visibility within the open-source AI community. It’s also intriguing to read a first-person account of China’s AI ecosystem from a foreign AI practitioner.
Main Text
Over the past roughly ten days, I had the privilege of visiting AI labs across China alongside my colleagues from SAIL. As someone who visited both China and the U.S. for the first time in six months, I found the differences between the two countries fascinating—but even more fascinating were the similarities.
What struck me most profoundly was how humble the AI researchers I met were.
They held other labs and peers in high regard. DeepSeek was frequently mentioned—likely because they’d just released a new model days before our visit—and people spoke about DeepSeek’s papers with genuine admiration.
Many researchers were close friends, having studied at the same university or hailing from the same hometown. They openly discussed their work, and their research findings would be published as papers several months later.
This stands out as one of the biggest contrasts with the Western AI community. In the U.S., the atmosphere often feels more like a zero-sum game. Labs are cautious about positioning. Researchers think in terms of competition, and some hold themselves in especially high esteem. Leaders insult and attack each other in leaked internal memos. This difference may be explained by a simple fact: leading U.S. labs are closed-source, whereas many Chinese labs are open-source. Chinese labs do feel wary of ByteDance’s closed-source chatbot Doubao—the most widely used chatbot in China, which holds a substantial lead.
At the same time, the overall vibe felt surprisingly similar to San Francisco. Researchers are extremely online, consuming content heavily on Twitter and Xiaohongshu (Little Red Book)—the latter growing increasingly popular. They all use Claude Code or their own CLI tools to build the next model. Some monitored training runs during our meetings, watching reward curves climb. They’re thinking about further scaling up—and complaining about insufficient compute. They’re frustrated with the current state of benchmarking.
Their primary focus is training better models. This differs markedly from San Francisco, where researchers tend to contemplate AI’s political or philosophical implications. They don’t dwell on mass unemployment, a permanent underclass, or whether their models possess consciousness. They simply want to train outstanding models.
Their eyes light up when they hear you’ve used their model. They’re eager to fix every flaw in today’s models in the next generation. They push through nights to launch new models—and still show up at the office afterward.
Most researchers I met were young—many in their early twenties or around age twenty-five. Some were undergraduates, but it’s more common for them to be Ph.D. candidates simultaneously working in industry. There’s a broad consensus that industry is far more exciting than academia right now—a view I strongly share, having done exactly the same thing myself. Labs place great value on such talent, actively recruiting interns and graduate students—a practice Western labs generally don’t follow.
This optimism among researchers extends to the general public, who appear more optimistic about technology—and about AI and robotics specifically. During the trip, people shared stories about their parents and grandparents using Doubao and DeepSeek for various tasks—including discussing mathematical theorems. This contrasts sharply with the West, where the general public tends to dislike AI.
Overall, this trip gave me a small glimpse into this ecosystem. It’s impossible to grasp the culture of such a vast civilization in just a few days. As a staunch supporter of open AI ecosystems and open research, I’m highly optimistic about the future of both—and hope for abundant international collaboration ahead.
I’d like to thank all the incredible people I met at Moonshot, Xiaomi, MiniMax, Zhipu AI, Meituan, Alibaba, Ant Lingxi, ModelScope, 01.ai, Unitree, and elsewhere. Thank you for your time and warm hospitality. I also thank SAIL for organizing this trip, and all participating writers and journalists. I’m deeply grateful to have met so many exceptional and ambitious individuals in such a short period.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@xeophon
