TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: Could LMs practically take text and directly generate IL/CPU opcodes?

2 点作者 wseqyrku3 个月前

3 条评论

petercooper3 个月前
You could train a model to do that. However, it&#x27;d probably be more practical to have an intermediate step like a text representation or some sort of script that then generates the final machine code (not necessarily assembly).<p>I just asked o3-mini <i>&quot;Can you write a Python script that outputs the bytecode to a file for a very basic JVM application that prints hello world?&quot;</i> and it actually did it, complete with bytecode, and it ran. (One benefit of this approach is it leaves comments next to each block of bytecode, which probably helps it contextually with generating the next ones.)
davydm3 个月前
Sure, why not? Code is code. Given the full instruction set, and trained on outputs from known code, it could produce the same buggy slop it usually does, but now, even harder to debug! Yay!
andrewfromx3 个月前
ha not even assembly? Right to opcodes?
评论 #43000331 未加载