• 0 posts
  • 7 comments
Joined 3 years ago
Cake day: June 12th, 2023
  • Linux@lemmy.world•GitHub - NullMagic2/slate-ntfs: New Rust NTFS implementation with full NTFS and chkdsk support. Works on older kernels.bykeisatsu@infosec.pubEnglish
    3 days

    I never said manual code is bug-free. But all errors in manual code are human-made.

    I judge a project based on the apparent judgement of it’s owners/maintainers.

    A large part of software engineering is using your judgement, whether it comes to overarching design, language choice, algorithmic implementations, but also chosing dependencies, tool, collaborators and documentation. Humans make mistakes, but can learn from them and improve over time. So while a mistake can be embarrassing andor costly, in the long term they are part of the learning process.

    When you opt for using LLMs for any of that you are outsourcing your judgement to a box that’s incapable of reasoning or learning. You will still get bugs and mistakes will be made but noone except Scam Altman and Wario Amodei will be happier for it.

    Studies also indicate that this is actually negatively impacting our cognitive function in the long term.

    And then we can add the ethical and environmental concerns on top of that, and suddenly the collective judgement of a project accepting or condoning LLM use is looking pretty damn bad if you ask me. Even before reading a single line of code in the repository.

  • Linux@lemmy.world•GitHub - NullMagic2/slate-ntfs: New Rust NTFS implementation with full NTFS and chkdsk support. Works on older kernels.bykeisatsu@infosec.pubEnglish
    3 days

    1. I too have lots of experience in C and I fucking suck at it.

    2. You seem awfully confident given that you have seen exactly zero lines of my code?

    3. if you are implying that your computer draws as much power when you manually code review as an LLM does you are either full of shit or doing something amazingly wrong.

    4. I better fucking hope that there is noone on that team that is worse than me at kernel development, but that does not mean that some of them are not absolute dumbfucks. Nice strawman tho.

    5. Not really, most of my code is closed source and all of it is shit tier. It’s completely LLM-free however

    If you believe my comments constitute bullying you have lived a sheltered life my friend.

    “Make a company that solves these problems then!”

    … dude. Are you all right?

  • Linux@lemmy.world•GitHub - NullMagic2/slate-ntfs: New Rust NTFS implementation with full NTFS and chkdsk support. Works on older kernels.bykeisatsu@infosec.pubEnglish
    3 days

    Well first of all, if you look at some of my comments you will see that I usually look closer than that. But its true that my opinion of a project plummets as soon as I see a single LLM artifact. I don’t agree that there has to be a certain amount of low quality commits for the project to be slop. Its more like if someone pissed in the coffee. The problem isn’t the actual volume of urine, it’s that someone pissed in it in the first place.

    Second, I reject the notion that I have to prove that any code is worse for having LLM extrusions in it. The evidence is clear, you need to be overwhelmingly proficient in any field to be able to catch the multitude of errors that the stochastic parrots shits out and you need to be constantly in your guard, because the LLM is trained to give you output that looks good, not good output.

    And if one is even slightly impressed over the quality of what the LLMs output it’s a confession that the person don’t know enough what he/she just prompted to tell the difference.

    Third, even if the output was pristine, objectively correct, and for every given purpose relevant and optimized, just the glaring ethical and environmental concerns should be enough for any decent human to stay the fuck away from the models that are typically used for coding, review and translation. Or are you going to tell me that all the projects I have commented on use ethically sourced open weighted models that run on green energy? Yes this too determines if it’s slop, because it’s slopping society too.

    Also, I am sorry, but “even the Linux kernel devs” is a ridiculous justification. They have as a collective done some amazingly stupid shit over the years and Linus is included in that even before bringing in LLMs in the development.