HN 提问：你如何让 LLM 生成好的代码？
1 分•作者: bjourne•4 个月前
最近，我想看看是否能让机器人帮助我整理我的音乐收藏。具体来说，我希望它能编写一个 Python 函数，该函数接受一首歌曲的音频文件路径，并返回该歌曲的原始（即第一张）专辑名称。如果忽略重制版、现场录音、盗版等，这是一个简单的启发式满射映射：Smoke on the Water -> Machine Head，Come as You Are -> Nevermind，Fools Gold -> The Stone Roses。
我可以使用 MusicBrainz API 快速编写这个函数，我相信大多数其他 HN 用户也能做到。然而，许多小时过去了，无论我如何提示机器人，我都无法让它完成这项任务。以下是 ChatGPT 生成的内容——其他机器人也回复了类似的垃圾内容：
https://chatgpt.com/share/6a396ea9-44d4-83eb-bdfd-216dfcc87e99
它们抱怨问题含糊不清（事实并非如此），或者 MusicBrainz 没有数据（它有）。是我提示机器人不正确，还是没有使用足够强大的编码模型？结果非常令人失望。
查看原文
Lately, I wanted to see if I could get the bot to help me organize my music collection. Specifically, I want it to code a Python function that takes the path to an audio file of a song and returns the name of the canonical (i.e., first) album of that song. If one ignores remasters, live recordings, bootlegs, etc., this is an easy heuristic surjective mapping: Smoke on the Water -> Machine Head, Come as You Are -> Nevermind, Fools Gold -> The Stone Roses.<p>I can code this function quickly using the MusicBrainz API and I believe most other HN readers can too. Yet many hours later, no matter how I prompt the bot I can't get it to do it. This is what ChatGPT generates---other bots reply with similar garbage:<p>https://chatgpt.com/share/6a396ea9-44d4-83eb-bdfd-216dfcc87e99<p>They complain about how the problem is ambiguous (is not) or how MusicBrainz doesn't have the data (it does). Am I prompting the bot incorrectly or not using powerful enough coding models? The results are very underwhelming.