394
Experienced software developers assumed AI would save them a chunk of time. But in one experiment, their tasks took 20% longer (fortune.com)
submitted 9 months ago by to c/programming@programming.dev
105 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 91 points 9 months ago* (43 children)

... That keeps making the same mistakes over and over again because it never actually learns from what you try to teach it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 43 child comments
    •
  • [–] 49 points 9 months ago (4 children)

    Yep, the junior is capable of learning.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [+] -10 points 9 months ago (12 children)

    This is not really true.

    The way you teach an LLM, outside of training your own, is with rules files and MCP tools. Record your architectural constraints, favored dependencies, and style guide information in your rule files and the output you get is going to be vastly improved. Give the agent access to more information with MCP tools and it will make more informed decisions. Update them whenever you run into issues and the vast majority of your repeated problems will be resolved.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 12 child comments
    •
  • [–] 18 points 9 months ago* (6 children)

    Well, that's what they say, but then it doesn't actually work, and even if it did it's not any easier or cheaper than teaching humans to do it.

    More to the point, that is exactly what the people in this study were doing.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 1 point 9 months ago

    If it's doesn't work for you, it's because you're a failure!

    Still not convinced these LLM bros aren't junior developers (at best) who someone gave a senior title to because everyone else left their shit hole company.

  • permalink
  • fedilink
  • source
  • parent
  • [–] -1 points 9 months ago* (last edited 9 months ago) (4 children)

    More to the point, that is exactly what the people in this study were doing.

    They don't really do into a lot of detail about what they were doing. But they have a table on limitations of the study that would indicate it is not.

    We do not provide evidence that: There are not ways of using existing AI systems more effectively to achieve positive speedup in our exact setting. Cursor does not sample many tokens from LLMs, it may not use optimal prompting/scaffolding, and domain/repository-specific training/finetuning/few-shot learning could yield positive speedup.

    Back to this:

    even if it did it’s not any easier or cheaper than teaching humans to do it.

    In my experience, the kinds of information that an AI needs to do its job effectively has a significant overlap with the info humans need when just starting on a project. The biggest problem for onboarding is typically poor or outdated internal documentation. Fix that for your humans and you have it for your LLMs at no extra cost. Use an LLM to convert your docs into rules files and to keep them up to date.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 11 points 9 months ago (3 children)

    Your argument depends entirely on the assumption that you know more about using AI to support coding than the experienced devs that participated in this study. You want to support that claim with more than a "trust me, bro"?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 0 points 9 months ago (2 children)

    Do you think that like nobody has access to AI or something? These guys are the ultimate authorities on AI usage? I won't claim to be but I am a 15 YOE dev working with AI right now and I've found the quality is a lot better with better rules and context.

    And, ultimately, I don't really care if you believe me or not. I'm not here to sell you anything. Don't use it the tools, doesn't matter to me. Anybody else who does use them, give my advice a try an see if it helps you.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 9 months ago (1 child)

    These guys all said the same thing before they participated in a study that proved that they were less efficient than their peers.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 0 points 9 months ago

    Again, read and understand the limitations of the study. Just the portion I quoted you alone is enough to show you that you're leaning way too heavily on conclusions that they don't even claim to provide evidence for.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 9 months ago (2 children)

    Codex literally lies about being connected to configured MCP servers.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 9 months ago

    In theory yes.

    In practice I find the more stuff like this you throw at it the more rope it has to hang itself with. And you spend so much time prompt adjusting so it doesn’t do the wrong things that you were better off just doing half of the tasks yourself.

  • permalink
  • fedilink
  • source
  • parent
  • [+] -29 points 9 months ago (24 children)

    This is why you use a downloaded llm and customize it, there’s ways to fix these issues.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 24 child comments
    •
  • [–] 25 points 9 months ago* (last edited 9 months ago) (22 children)

    Unless you are retraining the model locally at your 23 acre data center in your garage after every interaction, it's still not learning anything. You are just dumping more data in to its temporary context.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 22 child comments
    •
  • [+] -10 points 9 months ago* (15 children)

    What part of customize did you not understand?

    And lots fit on personal computers dude, do you even know what different llms there are…?

    One for programming doesn’t need all the fluff of books and art, so now it’s a manageable size. Llms are customizable to any degree, use your own data library for the context data even!

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 15 child comments
    •
  • [–] 11 points 9 months ago* (14 children)

    What part about how LLMs actually work do you not understand?

    "Customizing" is just dumping more data in to it's context. You can't actually change the root behavior of an LLM without rebuilding it's model.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 14 child comments
    •
  • [+] -6 points 9 months ago* (11 children)

    "Customizing" is just dumping more data in to it's context.

    Yes, which would fix the incorrect coding issues. It’s not an llm issue, it’s too much data. Or remove the context causing that issue. These require a little legwork and knowledge to make useful. Like anything else.

    You really don’t know how these work do you?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 11 child comments
    •
  • [–] 8 points 9 months ago* (last edited 9 months ago) (4 children)

    You do understand that the model weights and the context are not the same thing right? They operate completely differently and have different purposes.

    Trying to change the model's behavior using instructions in the context is going to fail. That's like trying to change how a word processor works by typing in to the document. Sure, you can kind of get the formatting you want if you manhandle the data, but you haven't changed how the application works.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] -2 points 9 months ago (3 children)

    Why are you so focused on just the training? The data is ALSO the issue.

    Of course if you ignore one fix, that works, of course you can only cry it’s not fixable.

    But it is.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 7 points 9 months ago* (2 children)

    Why are you so focused on just the training?

    Because I work with LLMs daily. I understand how they work. No matter how much I type at an LLM, its behavior will never fundamentally change without regenerating the model. It never learns anything from the content of the context.

    The model is the LLM. The context is the document of a word processor.

    A Jr developer will actually learn and grow in to a Sr developer and will retain that knowledge as they move from job to job. That is fundamentally different from how an LLM works.

    I'm not anti-AI. I'm not "crying" about their issues. I'm just discussing the from a practical standpoint.

    LLMs do not learn.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [+] -6 points 9 months ago* (1 child)

    Because I work with LLMs daily. I understand how they work.

    Clearly you don’t, because context data modifies how the training data extrapolates.

    You can use something, while not being educated on how to use it. And just using something does not mean you understand how they work. Your comments have made it QUITE clear that you have no idea.

    People who just whing about AI and pretend they know how they work are the worst kind of people right now.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 7 points 9 months ago*

    Your comments have made it QUITE clear that you have no idea.

    Odd, I can say the exact same thing about your comments on the subject.

    We are clearly at an impasse that won't be solved through this discussion.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 7 points 9 months ago (5 children)

    But

    All the fluff from books and art

    Is not inside the context, that comes from training. So you know how an llm works?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [+] -7 points 9 months ago* (4 children)

    Where do you think the errors are coming from? From data bleed over, the word “coding” shows up in books, so yes the context would incorrectly pull book data too.

    Or do you not realize coding books exist as well…? And would be in the dataset.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 3 points 9 months ago (3 children)

    Why would you put whole books into the context?!? Do you even know what an llm is?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [+] -6 points 9 months ago* (2 children)

    Because that’s how they work…? It’s not an actual physical book… you don’t seriously think this do you…? it’s the text data inside, like any other text file it would use for context.

    Where do you think it gets its data from…?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [+] -16 points 9 months ago (5 children)

    Sounds like you have no clue what an LLM/AI actually is or is capable of.

    https://medium.com/sciforce/step-by-step-guide-to-your-own-large-language-model-2b3fed6422d0

    It’s not hard to keep a data library updated for context, and some are under a TB in siz.

    Where are you getting your information from?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 11 points 9 months ago (4 children)

    It seems you are still confusing context with training? Did you read that text and understand it?

    Did you follow it yourself to build an llm?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] -4 points 9 months ago* (2 children)

    Why do you think it’s solely a training issue?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 9 points 9 months ago* (last edited 9 months ago)

    Yeah, but LLMs still consistently don't follow all rules they're given, they randomly will not follow one or more with no indication they did so, so you can't really fix these issues consistently, just most of the time.

    Edit: to put this a little more clearly after a bit more thought: It's not even necessarily a problem that it doesn't always follow rules, it's more so a problem that when it doesn't follow the rules, there's no indication it did so. If it had that, it would actually be fine!

  • permalink
  • fedilink
  • source
  • parent