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

121–129 of 129 posts

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

#121

Post by markasoftware »

ojkelly wrote: post #19

So far I’ve found copilot is like a weather forecast. The shorter the prediction the higher probability of it being correct. The longer, and more complex the prediction the less likely it’s all correct. But, with a close review you can spot the errors and correct it. I’ve found most useful to get ideas on the page fast. It’s also absolutely insane to be writing a parser, and have it suggest unit tests and code where…

You bring up an important point in general:

Copilot is great at writing tests.

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

#122

Post by adlpz »

jaredcwhite wrote: post #75
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…

Indeed you are not. I've never used any generative AI tools when coding or otherwise, and continue to refuse to at this point in time.

I understand not doing it, just because you didn't get around to try it, but, refusing? why? Moral reasons? Copyright worries?

It's just another tool. So I guess the good old "will never stop using emacs" kind of attitude applies here too?

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

#123

Post by la64710 »

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…

It is great for starters but my experience it everything AI generates needs to be double and triple tested and almost always manually modified to do exactly what you want to do. The 40% number just seems to be a tall claim and unless the data is shared and verifiable I don’t really give a crap about these claims.

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

#124

Post by marssaxman »

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'm not worried. Automating our jobs away is part of what programmers do, and always has been, but somehow we never run out of work to do.

I have not yet grown curious enough about Copilot to try it out; it seems to solve a problem I don't really have, since my productivity is never limited by typing speed. As you say, thinking and designing is the hard part.

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

#125

Post by dandellion »

paro_nej wrote: post #29
dandellion wrote: post #21

Earlier quoted context omitted.

In my case about 50% of the test code is just boilerplate, so I usually type the test name and generate the rest. Most of the time I have to rewrite the actual test logic (although very trivial tests are sometimes correct), but I probably keep more than 40% if we're talking tests specifically.

> In my case about 50% of the test code is just boilerplate Why not replace it with a function call?

In the case I was referring to, mainly to keep the code consistent with what was already there and the PRs small, as the code base is primarily owned by a different team. It's also pretty innocuous short tests that read well as they are.

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

#126

Post by flykespice »

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 can't afford to use it but if I did, I would use it as moderate as possible.

I look forward on the future of AI assisted coding tools, it could make us more productive on generating part of codes that we don't care much about, while leaving us time to deal with code that we do, manually.

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

#128

Post by Zuiii »

adlpz wrote: post #122
jaredcwhite wrote: post #75

Earlier quoted context omitted.

Indeed you are not. I've never used any generative AI tools when coding or otherwise, and continue to refuse to at this point in time.

I understand not doing it, just because you didn't get around to try it, but, refusing ? why? Moral reasons? Copyright worries? It's just another tool. So I guess the good old "will never stop using emacs" kind of attitude applies here too?

I never relied on code-completion (not even copilot). I found that coworkers who do simply can't function without it. They simply find it very difficult or even impossible to remember more than a few API names, parameters, etc. I still remember the constant complaints in the early days of Rust when code completion was not available or would constantly break.

Relying on code-completion for coding or relying on google maps for navigation dims your ability to do these things. Given that copilot is a remote service that I do not control, I will not voluntarily tie my ability to write software to it.

Coilot is a conditional, unreliable, untrustworthy, ability-as-a-service. No thank you.

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

#129

Post by rjh29 »

Waterluvian wrote: post #24
rjh29 wrote: post #20

Earlier quoted context omitted.

Most of the time I know exactly what tests I want to write but it's annoying to write them. Copilot autocompletes, I check it and then I use it.

For me the worst part about this is that writing tests tediously makes me reach for layers of abstraction on my test code. And then suddenly my test code is complicated and needs its own test code, and changing a test can often be problematic to the abstractions I foolishly employed. Being able to churn out the boilerplate for tests, which can make DRY a non-concern, is great. I just hope that if I do this, I never g…

I agree! I try not to make test code too complicated, and prefer repetitive code over abstractions (else you have to test the tests!) so Copilot is useful there.
Post reply on HN