I think the big picture here is that the difference in quality is largely subjective at this point, while US companies are burning through orders of magnitude of cash which is obviously not sustainable. We shouldn’t underestimate the power of developing things in the open. Chinese open models benefit from the wisdom of an entire global research community while American engineers working on proprietary closed models are working in their own insular silos. It should be no surprise that the scientific community at large would pull ahead of these small teams. On top of that, doing research in the open amortizes the cost. Incidentally, this is exactly the same logic that led open source to dominate in recent years.
OpenAIs bet was that compute was the limiting factor. Sam Altman claimed that essentially no-one could obtain the level of compute necessary to develop something like GPT-4. Then Deepseek came out.
They’re still trying to make the original claim true because if they admit that they were wrong (or full of shit) it all comes crashing down, not just their companies but likely the economy as a whole.
It doesn’t hurt that American frontier models are phenomenally huge and cost an exorbitant amount per token in and out while the main source of real world value created by llms in the past year has been through agentic/harness/other words for massive context systems that show a real benefit from simply using more tokens.
I think the big picture here is that the difference in quality is largely subjective at this point, while US companies are burning through orders of magnitude of cash which is obviously not sustainable. We shouldn’t underestimate the power of developing things in the open. Chinese open models benefit from the wisdom of an entire global research community while American engineers working on proprietary closed models are working in their own insular silos. It should be no surprise that the scientific community at large would pull ahead of these small teams. On top of that, doing research in the open amortizes the cost. Incidentally, this is exactly the same logic that led open source to dominate in recent years.
OpenAIs bet was that compute was the limiting factor. Sam Altman claimed that essentially no-one could obtain the level of compute necessary to develop something like GPT-4. Then Deepseek came out.
They’re still trying to make the original claim true because if they admit that they were wrong (or full of shit) it all comes crashing down, not just their companies but likely the economy as a whole.
The irony is that the original OpenAI thesis was correct. Use a datacenter to generate a frontier model, then distribute that AI openly.
It’s funny to watch them take the whole US economy down on a bluff.
It doesn’t hurt that American frontier models are phenomenally huge and cost an exorbitant amount per token in and out while the main source of real world value created by llms in the past year has been through agentic/harness/other words for massive context systems that show a real benefit from simply using more tokens.