Ask HN: 我们何时会把“人类当作工具”来使用,以便 LLM 智能体可以按需调用我们?
7 分•作者: vedmakk•6 个月前
认真的问题。
我们正在构建能够通过 MCP 进行规划、推理和调用工具的智能体 LLM 系统。目前,这些工具是 API。但许多现实世界的任务仍然需要人类的参与。
所以… 为什么不将人类也暴露为工具呢?
想象一下 TaskRabbit 或 Fiverr 运行 MCP 服务器,LLM 智能体可以:
- 调用人类进行判断、创造或执行物理操作
- 传递结构化输入
- 接收结构化输出并将其反馈到其循环中
到那时,人类就成为了智能体工具链中的另一个依赖项。虽然速度较慢,成本较高,但偶尔是必要的。
是的,这听起来很反乌托邦。是的,这把人类当作“AI 的仆人”。这正是问题的关键。这种情况已经以手动方式发生了… 这只是形式化了接口。
我真正感兴趣的问题是:
- 一旦智能体成为默认的软件执行者,这是否不可避免?(基本上从现在开始?)
- 哪些方面会首先崩溃:经济、安全、人类尊严还是监管?
- 市场是否会接受成为 AI 的“人类执行层”?
不确定这是未来还是我们应该积极阻止的诅咒想法… 但它感觉令人不安地可行。
查看原文
Serious question.<p>We're building agentic LLM systems that can plan, reason, and call tools via MCP. Today those tools are APIs. But many real-world tasks still require humans.<p>So… why not expose humans as tools?<p>Imagine TaskRabbit or Fiverr running MCP servers where an LLM agent can:<p>- Call a human for judgment, creativity, or physical actions<p>- Pass structured inputs<p>- Receive structured outputs back into its loop<p>At that point, humans become just another dependency in an agent's toolchain. Though slower, more expensive, but occasionally necessary.<p>Yes, this sounds dystopian. Yes, it treats humans as "servants for AI." Thats kind of the point. It already happens manually... this just formalizes the interface.<p>Questions I'm genuinely curious about:<p>- Is this inevitable once agents become default software actors? (As of basically now?)<p>- What breaks first: economics, safety, human dignity or regulation?<p>- Would marketplaces ever embrace being "human execution layers" for AI?<p>Not sure if this is the future or a cursed idea we should actively prevent... but it feels uncomfortably plausible.