Godot will no longer accept AI-authored code contributions

As a software developer with 15 years of experience who’s been unemployed for over 18 months shit like this just really pisses me off.

That’s an easy PR decline where I work tbh.

It seems a lot of people still don’t realize that it can’t do analysis; it just predicts things that can look a lot like analysis and sometimes it even predicts the correct conclusion. But it’s training data will include many cases where a question is posed and responses jump right to the answer without showing any of the underlying work that got there.

So the training isn’t “user a asks question, user b considers factors x, y, and z, including how y and z interact complexly with each other, as well as how a subtle complexity of the question means the usual case for these doesn’t apply in the same way, and posts this response”, which is what an actual intelligent being would need to get from the question to the answer like an expert would.

The training is “user a asks question, user b posts response”. That training is more like how to train someone to pretend to be an expert because all the information is just the question/response pair (with no reasoning that wasn’t included in the response), and the tone or language used to convey the response.

That’s why they always sound so confident, because they are predicting what a confident response would look like, but the actual content of the response is based purely on correlations between the words used in the question and whatever words followed variations of those words used in other questions or stories or wherever the training data came from.

Though even if everyone wrote out their full line of thought for every response, it still wouldn’t result in actual intelligence because it’s still just correlating words, not applying logic and analysis to them. That’s all LLMs do. Hallucinations are fundamental to how they work.

No, so technically you really are gatekeeping slop rather than “AI” specifically.

So there can be the code itself, so this example from a recently claimed ‘user level anti-cheat’, we have a purported “kernel integrity checker": https://github.com/TuncorReUnion/TLAC-MODERN-LOCAL-ANTI-CHEAT-REUNIONED/blob/main/kernel/tlac_kernel.c

So a relatively intact structure for a kernel module, but just the dumbest concept imaginable (a kernel is ‘good’ so long as the words ‘rootkit’ nor ‘suspicious’ appear in /proc/modules, since rootkits obviously must name themselves ‘rootkit’ and other suspicious software must name itself ‘suspicious’. Basically look at almost any file in that repository and you’ll see similar stuff, in-tact, even suspiciously complete boilerplate with stupid logic. Alternatively, crazy verbose.

The there’s the git activity, the project above has a git commit history: https://github.com/TuncorReUnion/TLAC-MODERN-LOCAL-ANTI-CHEAT-REUNIONED/commits/main/

Activity comes in crazy number of commits, and you can kind of see for each one what kind of prompt would have driven that change, with the disconnect between syntax correction but the underlying concept is just stupid.

Then sometimes, it’s not the code or code activity itself, but the commentary. So this change looks plausible to most folks without much tell: https://github.com/rhinstaller/anaconda/pull/7074/changes

Then there’s the discussion, which mostly proceeded in a “no one has any idea what any of this is about, but everything sounds plausible and it’s niche enough that heck if I know if it really makes sense”. Eventually this comment: https://github.com/rhinstaller/anaconda/pull/7074#issuecomment-4509336034

It has that LLM smell “That’s a great question”, and the expanded description is enough for a broader audience to start seeing jargon that doesn’t make sense. But it’s still niche enough for most people to have no idea what to make of it.

No. But there are code smells, that might reveal it.

It is the same with any kind of output generative deep learning models produce.

Rather, now that it is officially a reason to reject submissions the sloppers cannot complain about unfairness.

You know those heavy genAI users ain’t smart enough to be able to not disclose it anymore. The long time users have probably atrophied their brains so much that they would probably just copy the whole genAI response into their contributions and then complain when they get rejected or get extremely defensive if asked to explain the code they put forth.

I ain’t banking on the intelligence of genAI users to be smart.

And it will be easy to tell because they won’t understand their own PR.

AI contributions that you can’t obviously tell are AI contributions are less a concern.

If you can tell it’s AI slop, that’s really the concern.

If it’s an AI assisted quick fix for something, then you might not tell (except the description in the pull request can bury the explanation in a sea of LLM crap). You can review it, maybe a bit more guarded knowing that “looking right” is no longer as strong an indicator as it used to be, and make a decent decision.

But when people unleash the firehose and it’s too much volume, then a nice policy to just ignore. If you aren’t sure, you can look at the developer’s general activity and see hundreds of commits a day across dozens of projects… yeah that’s a vibe coder.

Slopper