📄 中文摘要
该研究分析了通用人工智能（AGI）在设计系统时面临的挑战，特别是幻觉、可纠正性以及无法通过简单扩展解决的结构性差距。提出了具身基础的重要性，强调在构建安全的AGI时，必须考虑系统的可逆性和动态交互。通过深入探讨这些概念，研究指出，只有在具身的基础上，AGI才能有效地理解和应对复杂的环境，从而减少潜在的风险和错误。
反转错误：为何安全的通用人工智能需要具身基础和状态空间可逆性
📄 English Summary
The Inversion Error: Why Safe AGI Requires an Enactive Floor and State-Space Reversibility
The study analyzes the challenges faced by Artificial General Intelligence (AGI) in system design, particularly focusing on hallucination, corrigibility, and the structural gaps that scaling cannot address. It emphasizes the importance of an enactive floor, arguing that safety in AGI requires consideration of system reversibility and dynamic interaction. By exploring these concepts in depth, the research indicates that only with an enactive foundation can AGI effectively understand and respond to complex environments, thereby reducing potential risks and errors.