Ask Lemmy
A Fediverse community for open-ended, thought provoking questions
Rules: (interactive)
1) Be nice and; have fun
Doxxing, trolling, sealioning, racism, toxicity and dog-whistling are not welcomed in AskLemmy. Remember what your mother said: if you can't say something nice, don't say anything at all. In addition, the site-wide Lemmy.world terms of service also apply here. Please familiarize yourself with them
2) All posts must end with a '?'
This is sort of like Jeopardy. Please phrase all post titles in the form of a proper question ending with ?
3) No spam
Please do not flood the community with nonsense. Actual suspected spammers will be banned on site. No astroturfing.
4) NSFW is okay, within reason
Just remember to tag posts with either a content warning or a [NSFW] tag. Overtly sexual posts are not allowed, please direct them to either !asklemmyafterdark@lemmy.world or !asklemmynsfw@lemmynsfw.com.
NSFW comments should be restricted to posts tagged [NSFW].
5) This is not a support community.
It is not a place for 'how do I?', type questions.
If you have any questions regarding the site itself or would like to report a community, please direct them to Lemmy.world Support or email info@lemmy.world. For other questions check our partnered communities list, or use the search function.
6) No US Politics.
Please don't post about current US Politics. If you need to do this, try !politicaldiscussion@lemmy.world or !askusa@discuss.online
Reminder: The terms of service apply here too.
Partnered Communities:
Logo design credit goes to: tubbadu
view the rest of the comments
I very much enjoy using AI for all the biloilerplate, test cases, suggestions, etc. It really makes me more productive, hard metrics behind it. Nobody is forcing me to, they just provide the license and let us use our judgment.
I honestly can't think of a project where 0% AI would be better. For 100% maybe a very trivial PoC, but even that would require at least a code revision.
So, as with many things, use in moderation is fine.
It's almost certainly also making your code worse.
It's not impossible to use AI effectively (although I would argue it's impossible to use large "frontier" models ethically, as the companies making them are burning the planet down to power the process), but you have to be extremely vigilant and thoughtful about what you're using it for, and you have to review every single line of code it produces, or you're going to miss bugs and you're going to lose skills.
A good way to test yourself is to see if you can still scaffold out an application by hand. Doesn't matter what... A to-do list, some buttons, whatever. Just test yourself to see if you can still do it.
If you can't, then you've lost the skills necessary to be certain that what you're producing with AI is actually good.
And if the idea of testing yourself like this makes you uncomfortable? Then AI isn't a tool you use, it's an addiction.
I mean, I do leet code semi-regularly, so I'm not too worried about getting rusty. Writing tests is boring as hell, the AI does a decent enough job for at least 90% of them.
Leet code is good for making sure you still have a good grasp of programming conceptually, but I don't think it's good for testing your own practical skills.
Seriously, just take an hour or two to scaffold out something new. Doesn't have to be complicated, just something to confirm for yourself that you can still do it. The only rule is to do it without AI.
When I did it myself, it was after months of my work requiring me to use AI, and there was a moment at the start where I was tempted to just fire up Copilot and tell it to do the work, which - of course - would have defeated the purpose. It was that moment where I realized I was addicted, and needed to go cold turkey.
Now I do the bare minimum with AI I'm required to at work, and focus on crafting my code carefully, by hand as much as possible. And it shows. My code quality has improved.
What do you mean by scaffolding something new? If it's writing all the boilerplate for the framework and dependencies, that's exactly what I don't care about. I use AI now and copy paste in the past.
I'd be curious to see what your hard metrics are based on.
https://mikelovesrobots.substack.com/p/wheres-the-shovelware-why-ai-coding
https://filebin.net/zf92vn7hm4zcabe9
Points per sprint, features shipped, test coverage. Defects remain unchanged.
Code quality? Maintainability down the line? Numbers for those aspects yet?
It's been more than 3 years since we started, and the metrics are stable, slight improvement even but that could be more experience or better models or anything. No apocalypse.
Happy that it's working out for at least a small margin of people. 👍
There's always the many ethical aspects as well, of course.
The ethics are debatable, but there's not turning back, there's plenty of open source models even that do a very decent job, so we will need to learn to deal with the reality. We never hired juniors anyway, but companies that did apparently have stopped, that can't be good.
In my opinion, the ethics are clear problems. There are way too many ethical issues for me to even consider using AI right now. I would never have a clear conscience using AI. Using AI as a service responsibly is impossible right now IMO.
I don't see it that bad. Specially open source models. Agree to disagree I guess.
Open source models aren't really used in AI as a service though, are they?