This is a well put together post. Always nice to see more perspectives! And I hope you've all enjoyed the trip so far.
You mentioned that China's AI industry's core competency is making money. I don't necessarily agree. It's true that making money is the outcome and a key factor for survival, but the logic may not be that way. Chinese history has taught us a harsh lesson that technology is the ultimate driver of growth (科技是第一生产力) and if you fall behind, you get bullied (落后就要挨打). So I think the initial goal has always been making advanced tech practical and available to everyone. It's almost an instinct now. This also explains why people here are so obsessed with new technology with almost zero apprehension, very unlike the public sentiment in the States.
And a side note, I've spoken to the employees at Unitree. They are actually less enthusiastic of the constant tourists because it interrupts their workflow. So they moved to a new office just to get away from the noise. Guess they're workaholics. Even the founder Wang Xingxing has admitted that he was very surprised by the constant hype around their robots.
I enjoyed your piece overall. Keep up the good work!
curious re: "making money", you don't think it's a goodhart proxy issue??
(countries with less sophisticated capital markets cannot tolerate risk and deviation/variance to the extent of a Google X or spaceX, openAI etc... incubating things insulated from the market a bit and celebrating a la YC just "being a hacker" which leads to money downstream
presumably malthusian conditions culturally - and the USA experienced such in previous regimes - would force this "practicality over vision & ideas" too)
China is still an emerging market. It's unlikely to replicate models like SpaceX or Google X. Even in the US, Google X’s projects have been scaled back by Google, while SpaceX has long relied on NASA orders and government contracts. I think in the future, a lot of nations will adopt this hybrid model: private enterprises will develop new ideas fueled by initial venture capital investment, while the government provides support through contracts. However, within China’s state-led economic model, the role of VC is less prominent.
I think this is a real issue for companies outside the US including China. An early stage deep tech startup can get funding several order of magnitude in the US than in China. For e.g. robotics companies in China with greater technical capabilities are valued at a fraction of US counterparts. This results in Chinese companies almost always playing catch up in paradigm changing techs due to capital constraint incentivizing current market fit and incremental improvements over moonshot type innovation
Wonderful piece! Fascinating to hear how much these companies differ in perspective and outlook. Low PhD salaries are something I heard several PhD students complain about low salaries in China and it makes sense that AI companies fill the gap, one wonders if that will also lead to more practical research.
Fascinating journalism, as a precious ByteDance employee, the posturing makes a lot of sense but I bet there is a fair amount of competition and rivalry that is just hidden rather than openly spoken about.
You literally sat at a dinner between Z.ai and Deepseek researchers - and I assume informative conversations still happened. I think that reinforces Nathan's POV that the labs don't have the type of rivalries that you see in the US.
The compute problem is on the cusp of being solved. As VCs, we have to project slopes out for a year, 5 years, or a decade. There's no doubt where the curve for Chinese labs and access to compute is pointing.
The sense I get is that if I sat between an OpenAI researcher and an Anthropic researcher at a dinner, it wouldn't be that weird. It can both be true that the rivalries seem a little bit more muted than in the US and that they still exist.
Do you have a source on the compute problem on the cusp of being solved? Curious for your reasoning behind this. (Also curious what you mean by solved, here. American companies are also desperate for compute, albeit at a higher absolutely level of supply).
Why are China’s companies so happy to pursue the things that Beijing declares are industrial priorities?
From AI to shares in mineral mines in Africa - often it’s smart, often it’s random, and always companies in my country would have smiled, nodded, and done nothing
Fascinating. The last few months of immersion in intelligence research (working on scicomm for an upcoming issue of The Proceedings of The Royal Society A on World Models) have me convinced that AGI may actually *require* compute constraints (as in, human deliberation is an energy-saving adaptation) — and if those constraints are pushing China into robotics capable of using AI as it is right now, maybe this only strengthens their position to lean into embodiment and real-world problems under the necessary selection pressure while American AI continues scaling in pursuit of emergent behaviors that don't actually emerge from unlimited cash and compute flows. America will get extremely capable System 1 machines, but nothing resembling System 2 human intelligence unless we follow the evolutionary trajectory of social and resource-constrained cognition.
This is a well put together post. Always nice to see more perspectives! And I hope you've all enjoyed the trip so far.
You mentioned that China's AI industry's core competency is making money. I don't necessarily agree. It's true that making money is the outcome and a key factor for survival, but the logic may not be that way. Chinese history has taught us a harsh lesson that technology is the ultimate driver of growth (科技是第一生产力) and if you fall behind, you get bullied (落后就要挨打). So I think the initial goal has always been making advanced tech practical and available to everyone. It's almost an instinct now. This also explains why people here are so obsessed with new technology with almost zero apprehension, very unlike the public sentiment in the States.
And a side note, I've spoken to the employees at Unitree. They are actually less enthusiastic of the constant tourists because it interrupts their workflow. So they moved to a new office just to get away from the noise. Guess they're workaholics. Even the founder Wang Xingxing has admitted that he was very surprised by the constant hype around their robots.
I enjoyed your piece overall. Keep up the good work!
curious re: "making money", you don't think it's a goodhart proxy issue??
(countries with less sophisticated capital markets cannot tolerate risk and deviation/variance to the extent of a Google X or spaceX, openAI etc... incubating things insulated from the market a bit and celebrating a la YC just "being a hacker" which leads to money downstream
presumably malthusian conditions culturally - and the USA experienced such in previous regimes - would force this "practicality over vision & ideas" too)
China is still an emerging market. It's unlikely to replicate models like SpaceX or Google X. Even in the US, Google X’s projects have been scaled back by Google, while SpaceX has long relied on NASA orders and government contracts. I think in the future, a lot of nations will adopt this hybrid model: private enterprises will develop new ideas fueled by initial venture capital investment, while the government provides support through contracts. However, within China’s state-led economic model, the role of VC is less prominent.
I think this is a real issue for companies outside the US including China. An early stage deep tech startup can get funding several order of magnitude in the US than in China. For e.g. robotics companies in China with greater technical capabilities are valued at a fraction of US counterparts. This results in Chinese companies almost always playing catch up in paradigm changing techs due to capital constraint incentivizing current market fit and incremental improvements over moonshot type innovation
well said, glad im not the only crazy one
Wonderful piece! Fascinating to hear how much these companies differ in perspective and outlook. Low PhD salaries are something I heard several PhD students complain about low salaries in China and it makes sense that AI companies fill the gap, one wonders if that will also lead to more practical research.
Wow, what a great post. Thank you so much.
Very interesting post. Thanks for sharing.
Great article! Really informative.
Great read.
Very interesting report, thanks!
Fascinating journalism, as a precious ByteDance employee, the posturing makes a lot of sense but I bet there is a fair amount of competition and rivalry that is just hidden rather than openly spoken about.
You literally sat at a dinner between Z.ai and Deepseek researchers - and I assume informative conversations still happened. I think that reinforces Nathan's POV that the labs don't have the type of rivalries that you see in the US.
The compute problem is on the cusp of being solved. As VCs, we have to project slopes out for a year, 5 years, or a decade. There's no doubt where the curve for Chinese labs and access to compute is pointing.
The sense I get is that if I sat between an OpenAI researcher and an Anthropic researcher at a dinner, it wouldn't be that weird. It can both be true that the rivalries seem a little bit more muted than in the US and that they still exist.
Do you have a source on the compute problem on the cusp of being solved? Curious for your reasoning behind this. (Also curious what you mean by solved, here. American companies are also desperate for compute, albeit at a higher absolutely level of supply).
Why are China’s companies so happy to pursue the things that Beijing declares are industrial priorities?
From AI to shares in mineral mines in Africa - often it’s smart, often it’s random, and always companies in my country would have smiled, nodded, and done nothing
I assume because in China, government favor matters more
The global market is fundamentally mispricing the physical commodity of silver. Silver can no longer be viewed merely as an investment vehicle, an inflation hedge, or "Gold's little brother." It has fundamentally transformed into a strategic resource vital to national defense and economic sovereignty.https://triggledger.substack.com/p/the-silver-key-the-irreplaceable?utm_source=share&utm_medium=android&r=8gc1qf
Fascinating. The last few months of immersion in intelligence research (working on scicomm for an upcoming issue of The Proceedings of The Royal Society A on World Models) have me convinced that AGI may actually *require* compute constraints (as in, human deliberation is an energy-saving adaptation) — and if those constraints are pushing China into robotics capable of using AI as it is right now, maybe this only strengthens their position to lean into embodiment and real-world problems under the necessary selection pressure while American AI continues scaling in pursuit of emergent behaviors that don't actually emerge from unlimited cash and compute flows. America will get extremely capable System 1 machines, but nothing resembling System 2 human intelligence unless we follow the evolutionary trajectory of social and resource-constrained cognition.