25+ yr Java/JS dev
Linux novice - running Ubuntu (no windows/mac)

  • 0 Posts
  • 13 Comments
Joined 1 year ago
cake
Cake day: October 14th, 2024

help-circle


  • I agree with you on a technical level. I still think LLMs are transformative of the original text and if

    when the number of sources that’s what ultimately created the volume of the N-dimensional probabilistic space they’re following is very low.

    then the solution is to feed it even more relevant data. But I appreciate your perspective. I still disagree, but I respect your point of view.

    I’ll give what you’ve written some more thought and maybe respond in greater depth later but I’m getting pulled away. Just wanted to say thanks for the detailed and thorough response.






  • Hey, so I started this comment to disagree with you and correct some common misunderstandings that I’ve been fighting against for years. Instead, as I was formulating my response, I realized you’re substantially right and I’ve been wrong — or at least my thinking was incomplete. I figured I’d mention because the common perception is arguing with strangers on the internet never accomplishes anything.

    LLMs are not fundamentally the plagiarism machines everyone claims they are. If a model reproduces any substantial text verbatim, it’s because the LLM is overtrained on too small of a data set and the solution is, somewhat paradoxically, to feed it more relevant text. That has been the crux of my argument for years.

    That being said, Anthropic and OpenAI aren’t just LLM models. They are backed by RAG pipelines which are verbatim text that gets inserted into the context when it is relevant to the task at hand. And that fact had been escaping my consideration until now. Thank you.




  • At work today we had a little presentation about Claude Cowork. And I learned someone used it to write a C (maybe C++?) compiler in Rust in two weeks at a cost of $20k and it passed 99% of whatever hell test suite they use for evaluating compilers. And I had a few thoughts.

    • 99% pass rate? Maybe that’s super impressive because it’s a stress test, but if 1% of my code fails to compile I think I’d be in deep shit.
    • 20k in two weeks is a heavy burn. Imagine if what it wrote was… garbage.
    • “Write a compiler” is a complete project plan in three words. Find a business project that is that simple and I’ll show you software that is cheaper to buy than build. We are currently working on an authentication broker service at work and we’ve been doing architecture and trying to get everyone to agree on a design for 2 months. There are thousands of words devoted to just the high level stuff, plus complex flow diagrams.
    • A compiler might be somewhat unique in the sense that there are literally thousands of test cases available - download a foss project and try to compile it. If it fails, figure out the bug and fix it. Repeat. The ERP that your boss wants you to stand up in a month has zero test coverage and is going to be chock full of bugs — if for no other reason than you haven’t thought through every single edge case and neither has the AI because lots of times those are business questions.
    • There is not a single person who knows the code base well enough to troubleshoot any weird bugs and transient errors.

    I think this is a cool thing in the abstract. But in reality, they cherry picked the best possible use case in the world and anyone expecting their custom project is going to go like this will be lighting huge piles of money on fire.