Whether these same observations can be extended to national security applications, and how, will be a big open question for the next couple of years. Balancing speed of adoption against growing risks from increasingly advanced and agentic AI will become a major challenge for all governments.
This framework shows that the true value in AI comes from turning raw intelligence into products embedded in workflows and context, not just from creating models
Hi Grace, thank you for sharing your insights on the Great AI Battle of the Century. IMHO, the winner is the entity who can swiftly calibrate a marginal net positive DNA (Direction, Needs, and Abilities) to create economically viable minimal products welcomed by the global masses. From the perspective of your analogy of pumping Crude Oil (intelligence commodity) to produce the Refined Gasoline (life-enhancing consumer products), I believe China does have a competitive edge due to (1) its wider and deeper value-driven AI supply chains, (2) its immensely rich data available for training LLMs, (3) its consistent and dedicated national support for AI data center power supply, and (4) its growing digitized socio-economic culture. I believe the Chinese AI ecosystem will prove to be the more integrated “Fertile Crescent” in which Agentic AI can thrive and quickly reap the highly sought ROI and profits, no matter how much the SPREADS expand or contract due to technological breakthroughs in AGI by the U.S. frontier models, which still lack a stable and fertile ground to grow and materialize. As observed nowadays, the many new ChatGPT or vibe coding tools using the best LLM model can merely give you the best recipe for dinner, but they are by no means the Michelin star winning restaurant that can truly deliver a delicious dish at an acceptable price. China has the right DNA to win the AI battle at this point.
> Real accumulated context: customer records, prior work product, compliance history, integration into adjacent systems, operational memory. The more useful the product becomes because it sits inside a growing body of context, the less interchangeable it becomes.
And how would you accomplish this technically? All of the current techniques (RAG, fine tuning, LoRA) have major drawbacks.
My goat, I’m a SWE trying to position myself in the AI market and your thoughts are so helpful
So many valuable insights here.
Whether these same observations can be extended to national security applications, and how, will be a big open question for the next couple of years. Balancing speed of adoption against growing risks from increasingly advanced and agentic AI will become a major challenge for all governments.
This framework shows that the true value in AI comes from turning raw intelligence into products embedded in workflows and context, not just from creating models
Hi Grace, thank you for sharing your insights on the Great AI Battle of the Century. IMHO, the winner is the entity who can swiftly calibrate a marginal net positive DNA (Direction, Needs, and Abilities) to create economically viable minimal products welcomed by the global masses. From the perspective of your analogy of pumping Crude Oil (intelligence commodity) to produce the Refined Gasoline (life-enhancing consumer products), I believe China does have a competitive edge due to (1) its wider and deeper value-driven AI supply chains, (2) its immensely rich data available for training LLMs, (3) its consistent and dedicated national support for AI data center power supply, and (4) its growing digitized socio-economic culture. I believe the Chinese AI ecosystem will prove to be the more integrated “Fertile Crescent” in which Agentic AI can thrive and quickly reap the highly sought ROI and profits, no matter how much the SPREADS expand or contract due to technological breakthroughs in AGI by the U.S. frontier models, which still lack a stable and fertile ground to grow and materialize. As observed nowadays, the many new ChatGPT or vibe coding tools using the best LLM model can merely give you the best recipe for dinner, but they are by no means the Michelin star winning restaurant that can truly deliver a delicious dish at an acceptable price. China has the right DNA to win the AI battle at this point.
> Real accumulated context: customer records, prior work product, compliance history, integration into adjacent systems, operational memory. The more useful the product becomes because it sits inside a growing body of context, the less interchangeable it becomes.
And how would you accomplish this technically? All of the current techniques (RAG, fine tuning, LoRA) have major drawbacks.