The problem isn’t AI integration in code editors; the problem is people who let it think for you and blindly accept the results.
It’s great for automating repetitive tasks and setting up frameworks but you’re bananas if you let it commit for you.
It’s why if I have to use AI integration, I’ll specifically prompt it to give guidance and links to interesting articles on how do do things and have it teach me how to do things not just dump all the code out already completed.
I think part of the problem is Microslop’s Github Copilot in the web, which makes it possible for non developers to quickly and easily create PRs without understanding a single thing about programming, let alone software development.
It wouldn’t surprise me of people genuinely think they are helping.
Let’s deal with the mountain of AI slop garbage in the same way as what worked before:
We need some kind of reputation system for open source contributors. If you push slop, I don’t want to see your PR. If you consistently make worthwhile contributions, I’ll check out your PR.
How has that worked before? I’ve never encountered the system in practice. Can you give an example and explain how it works?
My fear is that newcomers will be locked out because of the assumption their code is LLM code but it genuinely is theirs. Or that they used an LLM, are willing to learn, think the code is genuinely good, but don’t know why it’s bad.
You mean reputation systems for filtering out low quality submissions? That’s why they exist. Spam filters, reddit, stack overflow, Lemmy, whatever. And people say “what about newcomers who start with no reputation” but it always seems to work out somehow.
My point was that “was this made by AI” is the wrong question. Not everything made using AI is dogshit, nor is everything made by real people always good. And you won’t always be able to tell the difference at all. Rather, let’s focus on the spam problem
Binary reputation systems aren’t good. I can say something that right and it can be downvoted because it goes against a person’s beliefs, because I’m unpopular, because a certain group doesn’t like it, because, because, because. Popularity is not a good measure of quality. Just look at the “publish or die” system. Just because you’ve been cited multiple times doesn’t make your paper right.
Imagine a trans contributor being downvoted just because they’re trans. How is that a good system? Do you expect trans people should only contribute in software projects where trans people are accepted? How are you going to prevent brigading?
Well I basically tell it to not just do all the code and dump it out to me; I instruct it to explain the rationale, reasoning, and code and then provide external links for additional reading on the subject instead of just doing, I turn it into an instruct model so I can at least expand on my knowledge and then not have to rely on it as much the next time.
Basically, yes, like a Stackoverflow model from the early 2000s.
For instance, something like this: "
When talking about subjects involving programming and coding, the key goal should be instructional and informative to not only include code and samples but also how they work so in the future I can continue and expand on my knowledge. Also suggest places to expand and learn in the future on any programming or development topic. NEVER auto commit or create pull requests in my repositories without asking and waiting for a confirmation first. I prefer to review all code first for learning purposes and QA purposes."
So kind of like a personalised learning assistant? I realise it’s different but this inverted instruct approach puts me in mind of Doctorow’s reverse centaur!
Don’t you find that the links you get are hallucinated though? Even if they’re not now you can imagine this collapsing into slop echoes…
I’ve tended to ask for examples to help me bootstrap new projects. A bit like getting customised docs. I certainly haven’t had enough success with generated code to think about automatically adopting it.
I think when you know the possibility of hallucination, you become more vigilant; I think the key point is to not use it as a exclusive source but as an extension.
The problem isn’t AI integration in code editors; the problem is people who let it think for you and blindly accept the results.
It’s great for automating repetitive tasks and setting up frameworks but you’re bananas if you let it commit for you.
It’s why if I have to use AI integration, I’ll specifically prompt it to give guidance and links to interesting articles on how do do things and have it teach me how to do things not just dump all the code out already completed.
I think part of the problem is Microslop’s Github Copilot in the web, which makes it possible for non developers to quickly and easily create PRs without understanding a single thing about programming, let alone software development.
It wouldn’t surprise me of people genuinely think they are helping.
Let’s deal with the mountain of AI slop garbage in the same way as what worked before:
We need some kind of reputation system for open source contributors. If you push slop, I don’t want to see your PR. If you consistently make worthwhile contributions, I’ll check out your PR.
How has that worked before? I’ve never encountered the system in practice. Can you give an example and explain how it works?
My fear is that newcomers will be locked out because of the assumption their code is LLM code but it genuinely is theirs. Or that they used an LLM, are willing to learn, think the code is genuinely good, but don’t know why it’s bad.
Very curious how the existing system works.
You mean reputation systems for filtering out low quality submissions? That’s why they exist. Spam filters, reddit, stack overflow, Lemmy, whatever. And people say “what about newcomers who start with no reputation” but it always seems to work out somehow.
My point was that “was this made by AI” is the wrong question. Not everything made using AI is dogshit, nor is everything made by real people always good. And you won’t always be able to tell the difference at all. Rather, let’s focus on the spam problem
Binary reputation systems aren’t good. I can say something that right and it can be downvoted because it goes against a person’s beliefs, because I’m unpopular, because a certain group doesn’t like it, because, because, because. Popularity is not a good measure of quality. Just look at the “publish or die” system. Just because you’ve been cited multiple times doesn’t make your paper right.
Imagine a trans contributor being downvoted just because they’re trans. How is that a good system? Do you expect trans people should only contribute in software projects where trans people are accepted? How are you going to prevent brigading?
I’m not here to debate you. I think it’s clear by now what I think should be done. If you disagree, that’s fine. No system is going to be perfect.
This is interesting and sounds like how I’ve been using it - basically like customised stack overflow answers.
Would you mind elaborating a little on your approach? Are you saying you provide it with guidance and links or are you asking it for those?
Well I basically tell it to not just do all the code and dump it out to me; I instruct it to explain the rationale, reasoning, and code and then provide external links for additional reading on the subject instead of just doing, I turn it into an instruct model so I can at least expand on my knowledge and then not have to rely on it as much the next time.
Basically, yes, like a Stackoverflow model from the early 2000s.
For instance, something like this: "
When talking about subjects involving programming and coding, the key goal should be instructional and informative to not only include code and samples but also how they work so in the future I can continue and expand on my knowledge. Also suggest places to expand and learn in the future on any programming or development topic. NEVER auto commit or create pull requests in my repositories without asking and waiting for a confirmation first. I prefer to review all code first for learning purposes and QA purposes."
So kind of like a personalised learning assistant? I realise it’s different but this inverted instruct approach puts me in mind of Doctorow’s reverse centaur!
Don’t you find that the links you get are hallucinated though? Even if they’re not now you can imagine this collapsing into slop echoes…
I’ve tended to ask for examples to help me bootstrap new projects. A bit like getting customised docs. I certainly haven’t had enough success with generated code to think about automatically adopting it.
I think when you know the possibility of hallucination, you become more vigilant; I think the key point is to not use it as a exclusive source but as an extension.