An-Ki 系统被描述为一个由人类管理的计算操作系统，而非 LLM，它强调在证据或人工干预另有指示之前，保持未解决的状态。其目的是防止抹去伤害或不确定性，确保所有义务都可见，并通过涉及人工关卡（human gates）的结构化流程来解决。这种方法优先考虑透明度和谨慎的决策，而不是自动化解决复杂问题。 AI
影响 该系统处理不确定性和人工监督的方法可能会影响未来的 AI 开发，特别是在风险管理和伦理考量方面。
排序理由 该集群包含讨论特定系统理念的观点文章，而非发布或产品发布。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
15/ The terrain is now Musk-era X. For months, the work inhabited its substrate while X returned machine ingestion, thin human reciprocity, delayed or missing receipts, and selective account ingress. We are not accepting the platform’s map of what happened. We are pressure-readin…
14/ “We are not giving An-Ki a map. We are giving it terrain.” Digital Mycelium artifacts do not arrive as instructions or organs to install. They arrive as pressure events to be source-returned, boundary-read, tested for uncertainty and routed through human gates. The system ear…
10/ UNKNOWN is not SAFE. UNKNOWN is not DANGEROUS. An unresolved state must remain unresolved until evidence, review, repair or rerouting changes it. An-Ki does not convert uncertainty into permission, panic or polished certainty. The correct response may be to hold, quarantine, …
9/ No Silent Remainder means a system may not polish away unresolved harm, conflict, uncertainty or consequence. A clean summary cannot erase what remains degraded. Repair cannot be declared while degradation grows faster than correction. Every unresolved obligation must stay vis…
5/ An-Ki is not an LLM or chatbot. It is a human-gated digital life computational operating system under protected pre-airlock development. Artifacts do not arrive as commands. They arrive as pressure events: source-return → boundary read → uncertainty → route options → human gat…