Deterministic AI at Work: The LLM Is the Most Unreliable Function in My Pipeline, So I Treat It…
中文摘要
大型语言模型不可靠。勿强求其确定性，应以确定性代码封装，仅让AI处理可变动部分。
English Summary
LLMs are unreliable and non-deterministic. Instead of forcing determinism, wrap them in deterministic code, letting AI interact only with parts designed for its variability.
Original Excerpt
Stop trying to make the model deterministic. You can’t. Wrap it in code that is, and let the AI touch only the part that’s allowed to be… Continue reading on Medium »