I have had some experiences with bogus bug reports and PRs, at BeamMP (a mod that adds multiplayer to a car game, ~750k registered players), we have very little external repo activity all things considered. This is due to not advertising it, and due to the shitty license (soon to be AGPL-3.0), but people find it and contribute every now and then.
We have had some AI generated PRs and issues. These are incredibly easy to spot for me, I'll get into how in a second, but its been a matter of taking them half-seriously just in case. When asked about their decisions, they either post an obvious AI answer, or garbage, abd you can reject the contribution.
There are two parts to this:
1. why is this a problem? cant you just take it at face value?
Well, the point at which I'll accept an AI generated PR is when I cannot tell. This has always been the goalpost for me -- if the contribution is good and was made by the person contributing, it gets approved. If not, then not. AI counts as "didnt write it yourself", because I want to be able to ask you later whx something related broke, when your chatgpt chat is already closed. You need to understand the code you post.
2. How do you spot them?
Usually what these AI tools, and people using them, do, is change unrelated code. Rewording a comment in the same or even different file, or simply reordering stuff for no reason. Another obvious sign is the solution itself. If it doesnt compile, or doesn't even look like the right language (e.g. C when the codebase is C++), thats a sign as well. Use of libraries that arent there, or even a PR body filled with text.
Real people are lazy. Real people tend to get the job done and gtfo, and not blabber on about whatever in their PR. Real people like me also wont merge anything that wastes my time, like AI garbage.
We have had some AI generated PRs and issues. These are incredibly easy to spot for me, I'll get into how in a second, but its been a matter of taking them half-seriously just in case. When asked about their decisions, they either post an obvious AI answer, or garbage, abd you can reject the contribution.
There are two parts to this:
1. why is this a problem? cant you just take it at face value? Well, the point at which I'll accept an AI generated PR is when I cannot tell. This has always been the goalpost for me -- if the contribution is good and was made by the person contributing, it gets approved. If not, then not. AI counts as "didnt write it yourself", because I want to be able to ask you later whx something related broke, when your chatgpt chat is already closed. You need to understand the code you post.
2. How do you spot them?
Usually what these AI tools, and people using them, do, is change unrelated code. Rewording a comment in the same or even different file, or simply reordering stuff for no reason. Another obvious sign is the solution itself. If it doesnt compile, or doesn't even look like the right language (e.g. C when the codebase is C++), thats a sign as well. Use of libraries that arent there, or even a PR body filled with text.
Real people are lazy. Real people tend to get the job done and gtfo, and not blabber on about whatever in their PR. Real people like me also wont merge anything that wastes my time, like AI garbage.