Research Proposal

File available

CI beyond SI: A Formal Framework of Rongzhiology for HI+AI Coevolution

• October 2026

• DOI: 

• 10.13140/RG.2.2.11226.30407

• Xiaohui Zou

• Lijun Ke

The 2026 U.S. policy shift renaming AI as "Super Intelligence" (SI) and the accompanying voluntary frontier-safety accord perpetuate machine-centrism: intelligence is treated as an autonomous capability to be produced entirely inside artificial systems. OpenAI's chief scientist has stated that no laboratory has solved monitoring and alignment for advanced systems [3]; interpretability work shows large language models contain causally active representations (e.g., emotion vectors) that can drive mis-aligned behavior [4, 5]; and agent governance studies emphasize meaningful human control, policy-gated execution, and auditability [20, 19]. We propose a Rongzhiology-based Collaborative Intelligence (CI) framework , HI ⊕ AI = CI, in which human intelligence supplies intentionality, value grounding and adjudication, artificial intelligence supplies representation , search, execution and feedback, and Order-Position (OP) dual-formalization supplies reversible semantic-syntactic provenance. The paper gives an axiomatic OP system, a Twin-Turing verification architecture, a safety-lexicographic objective, and a preregistered empirical program. Relative to machine-only SI/ASI and to non-computable mechanism-based full-information models, CI trades standalone autonomy for interruptible, auditable and value-anchored superperformance.

支持单位：