在生产环境中构建 LLM 应用程序

目录

Part I. Challenges of productionizing prompt engineering第 I 部分生产提示工程的挑战

The ambiguity of natural languages自然语言的歧义

Cost and latency 成本和延迟

Prompting vs. finetuning vs. alternatives提示与微调对比备择方案

Embeddings + vector databases 嵌入+矢量数据库

Backward and forward compatibility向后和向前兼容性

Part 2. Task composability 第 2 部分。任务可组合性

Applications that consist of multiple tasks包含多个任务的应用程序

Agents, tools, and control flows代理、工具和控制流

Part 3. Promising use cases第 3 部分。有前途的用例

AI assistant

Chatbot

Programming and gaming 编程和游戏

Learning

Talk-to-your-data

Search and recommendation 搜索和推荐

Sales

SEO

Conclusion

A question that I’ve been asked a lot recently is how large language models (LLMs) will change machine learning workflows. After working with several companies who are working with LLM applications and personally going down a rabbit hole building my applications, I realized two things:
最近我经常被问到的一个问题是大型语言模型 (LLM) 将如何改变机器学习工作流程。在与几家使用 LLM 应用程序的公司合作并亲自深入构建我的应用程序之后，我意识到两件事：

1. It’s easy to make something cool with LLMs, but very hard to make something production-ready with them.
   使用 LLM 很容易做出很酷的东西，但很难用它们做出适合生产的东西。
2. LLM limitations are exacerbate