Existing classification standards for robots and intelligent machines mostly judge performance from external behaviors, automation degree and algorithm complexity, lacking a unified evolutionary framework integrating human operation cost and internal cognitive paradigm. This paper proposes an original six-level hierarchical evolution system from primitive manual tools to embodied general intelligence. The division criteria are two dimensions: human learning cost and essential paradigm leap of the machine system itself. Each level represents an irreversible structural mutation that cannot be achieved simply by increasing computing power. L1 is intuitive mechanical tools; L2 is rule-based tools corresponding to assembly language; L3 is general programmable machines based on high-level programming languages; L4 is natural language interaction systems built upon large language models, which is the upper limit of all commercial intelligent products today; L5 is imitation learning robots with demonstration replication ability, which only has preliminary prototypes confined to simulation environments; L6 is self-verifying autonomous modeling cognitive systems, which have not formed mature engineering paradigms and only have faint phenomena in virtual simulations. This hierarchical model can systematically evaluate the maturity of machinery, automation and artificial intelligence robots, and clarify the essential boundary between contemporary applied intelligence and general embodied intelligence.
Q Chen (Sun,) studied this question.