Comment on The Fall of Stack Overflow

• li10@feddit.uk ⁨1⁩ ⁨year⁩ ago

  Because forum posts are always full of accurate and helpful information?

  In my experience it still makes good suggestions for most things, and is better than trying to phrase things in a way that Google likes, then trawling through irrelevant forum posts.

  It’s only there to make suggestions, so if someone is taking its output without understanding and treating it like gospel then they’re an idiot who’s inevitably going to end up in a world of trouble.

  If you take the suggestion, verify it with documentation, then make sure you actually understand it, chatGPT is a great tool.

  source

  • wren@sopuli.xyz ⁨1⁩ ⁨year⁩ ago

    Because forum posts are always full of accurate and helpful information?

    Not necessarily, but at least there’s much more opportunity for other people to jump in and correct false info or expand upon something. It’s by no means a flawless system, but it’s better than only have one source of information

    source

    • new_guy@lemmy.world ⁨1⁩ ⁨year⁩ ago

      When we use forums there’s also an opportunity to correct (or be corrected) on how we deal with problems.

      I’ve seen a few times people asking how to do X while they’re actually trying to do Y. ChatGPT would gladly direct them to the wrong path.

      source
  • agressivelyPassive@feddit.de ⁨1⁩ ⁨year⁩ ago

    If I’m honest, stackoverflow was always a shortcut for searching documentation to me.

    Simple stuff like how do I turn an InputStream to a String again? I can’t remember it, but I know exactly what to look for, I’m just to lazy.

    For that kind of stuff ChatGPT is almost perfect.

    source
  • SuperFola@programming.dev ⁨1⁩ ⁨year⁩ ago

    I didn’t say that people should go on the internet and pick the first forum post either ; that would be like trusting whatever chatgpt is handing you :p

    My point was more on the “people are lazy” side of things, but yeah you have to stay critical of both chatgpt and forum posts.

    source

    • li10@feddit.uk ⁨1⁩ ⁨year⁩ ago

      I agree, I just think that those lazy people will do what they do regardless of where they get their info.

      To butcher a saying; blame the craftsman, not the tools.

      source

      • Aimhere@midwest.social ⁨1⁩ ⁨year⁩ ago

        I half expect that, if enough programmers use ChatGPT-written code verbatim, someday it’s going to lead to Skynet. I mean, what’s to stop ChatGPT from inserting bits of extra code to be used for its own distributed processing botnet?

        source

        • -> View More Comments
  • nous@programming.dev ⁨1⁩ ⁨year⁩ ago

    Sadly there are so many people that take its output as gospel and don’t realise it can be wrong. So is a tool that commonly gets abused by people that don’t know how to use it.

    source
• gosling@lemmy.world ⁨1⁩ ⁨year⁩ ago

  You mean shitty code which you can just check and ask them to change in almost real time, over posting your question on SO and waiting for months for an answer?

  source
• ofak@lemmy.world ⁨1⁩ ⁨year⁩ ago

  Chatgpt is still a tool and it’s up to the user how to use it. If you google “bolognese recipe” you get one result; if you Google “traditional ragu from Bologna” you get another. Same for ChatGPT.

  source
• JesusTheCarpenter@feddit.uk ⁨1⁩ ⁨year⁩ ago

  At least ChatGPT will not flag the question as duplicate.

  source

  • Gork@lemm.ee ⁨1⁩ ⁨year⁩ ago

    “I’m sorry, as an AI language model this question has been asked too many times and there is insufficient computer resources to handle your request. You’ve been temporarily silenced for 15 minutes.”

    source
• EatMyDick@lemmy.world ⁨1⁩ ⁨year⁩ ago

  You are delusional and will be left behind if that is your view point. The code is usually largely accurate only needing a few tweaks. Easily one of the most powerful scaffolding and learning tool I’ve used in 25 years. Our developers embracing it are more efficient then ever and passing static analysis, owasp scans, coding standards just fine if not better than cranky old devs who think they couldn’t possibly be helped by a dumb machine.

  source
• Aidan@lemm.ee ⁨1⁩ ⁨year⁩ ago

  You can have it generate shitty code and then compare it against examples it finds online to iterate that code. Also, it was trained on the whole internet, including those good solutions, and can often reproduce them on its own. but you have to tell it to do all that, rather than just asking for the code.

  source