Comment on AI agents wrong ~70% of time: Carnegie Mellon study

• MangoCats@feddit.it ⁨1⁩ ⁨month⁩ ago

  I have actually been doing this lately: iteratively prompting AI to write software and fix its errors until something useful comes out. It’s a lot like machine translation. I speak fluent C++, but I don’t speak Rust, but I can hammer away on the AI (with English language prompts) until it produces passable Rust for something I could write for myself in C++ in half the time and effort.

  I also don’t speak Finnish, but Google Translate can take what I say in English and put it into at least somewhat comprehensible Finnish without egregious translation errors most of the time.

  Is this useful? When C++ is getting banned for “security concerns” and Rust is the required language, it’s at least a little helpful.

  source

  • jsomae@lemmy.ml ⁨1⁩ ⁨month⁩ ago

    I’m impressed you can make strides with Rust with AI. I am in a similar boat, except I’ve found LLMs are terrible with Rust.

    source

    • MangoCats@feddit.it ⁨1⁩ ⁨month⁩ ago

      I was 0/6 on various trials of AI for Rust over the past 6 months, then I caught a success. Turns out, I was asking it to use a difficult library - I can’t make the thing I want work in that library either (library docs say it’s possible, but…) when I posed a more open ended request without specifying the library to use, it succeeded - after a fashion. It will give you code with cargo build errors, I copy-paste the error back to it like “address: <pasted error message>” and a bit more than half of the time it is able to respond with a working fix.

      source

      • jwmgregory@lemmy.dbzer0.com ⁨1⁩ ⁨month⁩ ago

        i find that rust’s architecture and design decisions give the LLM quite good guardrails and kind of keep it from doing anything too wonky. the issue arises in cases like these where the rust ecosystem is quite young and documentation/instruction can be poor, even for a human developer.

        i think rust actually is quite well suited to agentic development workflows, it just needs to mature more.

        source

        • -> View More Comments
• jsomae@lemmy.ml ⁨1⁩ ⁨month⁩ ago

  The problem is they are not i.i.d., so this doesn’t really work. It works a bit, which is in my opinion why chain-of-thought is effective (it gives the LLM a chance to posit a couple answers first). However, we’re already looking at “agents,” so they’re probably already doing chain-of-thought.

  source

  • Knock_Knock_Lemmy_In@lemmy.world ⁨1⁩ ⁨month⁩ ago

    Very fair comment. In my experience even increasing the temperature you get stuck in local minimums

    I was just trying to illustrate how 70% failure rates can still be useful.

    source
• davidagain@lemmy.world ⁨1⁩ ⁨month⁩ ago

  What’s 0.7^10?

  source

  • Knock_Knock_Lemmy_In@lemmy.world ⁨1⁩ ⁨month⁩ ago

    About 0.02

    source

    • davidagain@lemmy.world ⁨1⁩ ⁨month⁩ ago

      So the chances of it being right ten times in a row are 2%.

      source

      • Knock_Knock_Lemmy_In@lemmy.world ⁨1⁩ ⁨month⁩ ago

        No the chances of being wrong 10x in a row are 2%. So the chances of being at least right once are 98%.

        source

        • -> View More Comments