Source: The Reality Engine — Chinese AI strategy analysis
One of the more interesting contrasts in studying AI from a PolyU Hong Kong perspective is how differently the same transitions are framed in Chinese versus Western AI discourse. The Reality Engine document uses the term “activation” for the current period — not “development” or “progress” but activation, implying that a capability was always latent and is now being switched on.
The C-A-M-C framework — Cognition, Action, Memory, Communication — is their structural decomposition of what a general-purpose AI agent needs. This maps roughly onto the Western concept of agentic AI, but the emphasis is slightly different: Memory and Communication are treated as equal pillars to Cognition, rather than implementation details. I think that’s actually a more accurate framing of what makes current AI agents fail — not that they’re not smart enough, but that they can’t remember and can’t coordinate reliably.
The strategic implication is geopolitical as much as technical. Both the US and China are clearly in a race to define the architecture of the next generation of AI systems — not just the models, but the infrastructure standards, governance frameworks, and data ecosystems. DeepSeek’s open-source strategy, Huawei’s chip programme, the pace of Chinese model releases in 2026 — these are moves in a long game whose rules are still being written.
Understanding both framings — not to choose sides, but to understand the actual shape of the competition — seems increasingly important for anyone operating in the space professionally.