Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

40% of the code GitHub Copilot users check-in is AI generated and unmodified

microsoft.com

Post reply on HN
Print view View original

111–120 of 129 posts

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#111

Post by actsasbuffoon »

ramblerman wrote: post #78
sph wrote: post #69

Aside: am I the only one here that is still writing code by hand and not using AI to autogenerate it? Because it seems over the past couple years on HN everybody is making great use of Copilot and constantly talking about it, while honestly scaffolding boilerplate is the least of my problems, while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my t…

> while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my time. So don't use it there :) But I'm curious what kind of coding you do that you never have to do any scaffolding, refactoring, or come across any element of repetitiveness. That is where it shines imo. I find during those boring tasks it quite often just gets my intent, and I can say "yes…

I love it for writing tests. It’s good at picking up patterns, so if I need a bunch of unit tests then I can usually write one test and let it write the rest based on the RSpec descriptions.

I recently had a ticket that required refactoring some code that used net::http from Ruby’s stdlib over to using the HTTP gem. The rewrite was pretty easy to do by hand, but we had a bunch of test cases and rewriting them proved to be a bit painful because the mocks required were quite different.

So I spent half an hour rewriting the first test case with the new shape of mocks. Then I commented out the remaining tests and the AI recognized what I was doing and rewrote the remaining test cases with the new mock structure in seconds.

I treat AI like a junior programmer. I don’t trust it to do the hard stuff, but it’s great for scenarios where I have already established the pattern and I just need someone with basic coding proficiency to apply that pattern to a large chunk of code. It’s amazing at doing the grunt work.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#112

Post by figassis »

sph wrote: post #69

Aside: am I the only one here that is still writing code by hand and not using AI to autogenerate it? Because it seems over the past couple years on HN everybody is making great use of Copilot and constantly talking about it, while honestly scaffolding boilerplate is the least of my problems, while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my t…

I use it, for autocomplete. It’s great for creating the text for string constants, error messages, autocompleting if statements (which if variables are properly named, works great). Speeds up repetitive work but it’s not useful for actual problem solving.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#113

Post by TomSwirly »

KyeRussell wrote: post #98
sph wrote: post #69

Aside: am I the only one here that is still writing code by hand and not using AI to autogenerate it? Because it seems over the past couple years on HN everybody is making great use of Copilot and constantly talking about it, while honestly scaffolding boilerplate is the least of my problems, while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my t…

Weird flex or whatever dude. But all I’m hearing is that your anxiety has led you to a) refuse to look into a new piece of technology, b) hold incorrect views on it (like that it’s only useful for boilerplate code), c) look down on those that ARE using it, and d) pose that your incredible experience is why this wouldn’t be useful to you. If you can’t find your drive to look at new things enough to even see if they’re…

Being polite to others costs nothing and makes this a better place for everyone.

Personal insults are particularly destructive to civilized discourse.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#114

Post by ElCapitanMarkla »

Not sure if I’m using it wrong or the problems I’m coding for aren’t totally generic. The best help I get from it is a little string formatting or some basic error handling etc. I actually cancelled my subscription earlier today because I’m not getting enough benefit from it.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#116

Post by WastingMyTime89 »

JustSomeNobody wrote: post #110
powerapple wrote: post #107

Earlier quoted context omitted.

when designing, thinking, abstracting, we are actually not writing any code. I sometimes use a whiteboard for example, of course CoPilot cannot help. Once we have your designed, thought-through implementation in your head, while you are typing in an editor, CoPilot definitely can help you write code faster (or slower if the output is not exactly what you have in your mind). I see it is a productive positive product.

> CoPilot definitely can help you write code faster(or slower if the output is not exactly what you have in your mind). I see it is a productive positive product. The actual act of writing code "faster" baffles me. Without tools like intellisense, etc.[0], I am perfectly able to write code as fast as I can type. There's absolutely zero need for me to write faster. [0]Intellisense, etc. tend to get in my way, so I tur…

Intellisense is a godsend when you are working in a large code base using proprietary API with varying level of documentation. It's very good at filling in the full name of that pesky class or giving you a quick list of functions in that package for you to pick the one you want.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#117

Post by saurik »

sph wrote: post #80
ramblerman wrote: post #78

Earlier quoted context omitted.

> while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my time. So don't use it there :) But I'm curious what kind of coding you do that you never have to do any scaffolding, refactoring, or come across any element of repetitiveness. That is where it shines imo. I find during those boring tasks it quite often just gets my intent, and I can say "yes…

AI is not as foolproof as many want us to believe. In my mind it's like giving a less experienced and error-prone engineer the code I'm working on to refactor. Perhaps it's because I've always been a lone wolf engineer, but I'd rather do it myself than giving quite an important task to someone less experienced than me, whether I'm writing something new (where I need to be at my best) or refactoring (where I still nee…

> In my mind it's like giving a less experienced and error-prone engineer the code I'm working on to refactor.

You have almost 20 years of professional experience on your resume and likely were coding before that; it is difficult for me to believe sometimes, but most programmers have nowhere near that level of experience as the field is constantly increasing in size and scope: when I was in college, Computer Science was one small major people could enter that had only momentarily become popular due to the Internet... now it makes up a third of all students at major institutions that I would have assumed should be doing more interesting research work :(.

Regardless, you (and I) look at this AI as an engineer that is much less experienced than we are, and which is mostly able to code in repetitive ways with a ton of boilerplate that it sometimes gets wrong anyway; but, I think, for a remarkable number of people--even ones on this forum, where I often want to assume everyone should be an expert but in fact a lot of people just got into doing this work last week--that AI actually already has more experience than they do (and, now that they are using it so heavily to do their job, maybe ever will) and thereby makes fewer mistakes.

I don't know what that feels like, but it is certainly a very different experience than we are having. Maybe in another year or two, we'll have that experience also: if the AI gets good enough to become your peer, maybe we will find it interesting to use (or maybe we will become so demoralized, having not been in the position of having to work in a team of people better than us for so long--and these AIs wouldn't even be people!!--that we give up on the art entirely). I mean, imagine for a moment what it would be like if you found yourself in awe of the AI and were learning from it... I think it explains a lot of the confusing excitement about this technology.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#119

Post by commondream »

sph wrote: post #69

Aside: am I the only one here that is still writing code by hand and not using AI to autogenerate it? Because it seems over the past couple years on HN everybody is making great use of Copilot and constantly talking about it, while honestly scaffolding boilerplate is the least of my problems, while thinking, designing and abstracting novel ideas into code is the hard part where AI would fall short and just waste my t…

I tested it, the suggestions were fine but not so helpful to be worth the property theft risk, so I uninstalled it.

You are not alone.

Re: 40% of the code GitHub Copilot users check-in is AI generated and unmodified

#120

Post by markasoftware »

nathansherburn wrote: post #76
zarzavat wrote: post #73

Earlier quoted context omitted.

Copilot is way more than scaffolding boilerplate, it catches bugs as you are writing them. For example you expect: x[i] But copilot suggests: x[i+1] …and you think, and you think, and then one of two things happens. Either you satisfy yourself that the answer is indeed x[i], or you accept x[i+1] and sheepishly thank Copilot for catching a bug today. You could have copilot on in the background and never accept a sugge…

I found during my preview months ago that it would often write bugs that took me longer to correct than if I had written it myself. Has it gotten better? Also in the situation where it suggests x[i+1] and it's wrong, doesn't that annoy you? In both cases I'm sure this is just temporary while LLMs get better. But my experience was they were _just_ a little too annoying last time I used them.

Indeed, it introduces subtle bugs related to indexing pretty often. At one point I had it autocomplete maybe 15 lines of code, and I trusted it because the code was all pretty easy and I had written a descriptive comment, which tends to help out Copilot a lot.

However, it introduced a very subtle bug where it used a loop counter `i` where it should have been indexing into an array like `arr[i]`, but it still typechecked and happened not to cause insane errors. Took a lot of time to track down later, and it's a bug I never would have written in myself.

Post reply on HN