An open discussion on the AI activity on the Godot repo + forums

Over time, I’ve had less and less reasons to want to engage with issues on the Github repo, as well as any forum, due to the amount of generated AI content on them. Sadly, this kind of thing is the fault of the users though, it’s everywhere and really hard to curtail

It’s completely bloating conversations, im sure it’s increasing work for the maintainers, and making it difficult to read through any comment using markdown or exclamation points without questioning reality

I mean just LOOK at this: iOS 4.6.2 export template ships broken arm64 simulator xcframework slice on Apple Silicon · Issue #118161 · godotengine/godot · GitHub

My main concern is that everyone has access to LLMs already, why in any world would we want to hear the output from your specific prompt? It’s a waste of everyone’s time

Beyond this, the promotion of AI generated game tooling and one-shotting is not helpful to post on forums such as this, especially when it’s blatantly promoting of a service

I will be up front: even though I am forced to use these things at work, they are horrible and do not do anything but remove the desire to learn, which diminishes everything in a human (not only their games…)

I really wish there was a good solution for this kind of thing, so im not asking the Godot Foundation (who already work too hard for us) to do anything. I know from my job that open source needs a better platform, or, support from existing ones to improve the AI slop. And Github isn’t exactly the company to do anything about that… But just wanted to vent my frustrations some I guess and maybe read some human-typed words :upside_down_face:

21 Likes

Hi!

As a fellow human being I want to send you some pleasant “I concur wholeheartedly”'s

It sucks that they’re shoving it through your throat at work too.

Best and in hopes of some more kind an human responses.

15 Likes

I completely agree with your perspective here.

It’s honestly becoming exhausting to open a thread and spend the first few seconds trying to determine whether a comment was written by a person or generated by an LLM. The content itself almost feels secondary at that point.

The part that resonates with me most is that everyone already has access to these tools. Community spaces are one of the few places where people still go specifically to hear other people’s experiences, opinions, and insights.

In any case, thank you for putting this into words. It’s reassuring to know I’m not the only one feeling this way.

Wishing you a pleasant day and many future human-authored discussions.

9 Likes

WAZZZZAAAAAAP!!!

Anyways, I had a whole discussion about this and the downgrade in forum Help topics on the Monkanics Discord.

Our theory is when new devs run out of Claude tokens, they come here, expecting people like @gertkeno and @dragonforge-dev to spit out answers like they can read their mind.

Really disheartening, by-the-by.

10 Likes

I think they get that people can just give answers is the fact that they’ve seen niche cases work in the past and that they’ll answer EVERY topic, granted it has effort or not. We’re here to help, not make your games, people.

thanks all for the replies, its great to read through them :slight_smile:

1 Like

My response is 100% USDA NON-GMO fresh homemade artisan quality assured lifetime warrenty no artificial colors or dyes satisfied or your money back organic human brain meat made.

5 Likes

Something really disheartening is that I can see the decay of “asking questions” in my work. Questions should be something like this in an engineering team;

“I’m working on X, but Y is happening. I’ve tried ABCD and still ended up with Y. Any ideas?”

Now the questions are more along the lines of;

“How do you do Z?” - Notice how I didn’t mention Z earlier? That’s because they went down a rabbit hole with an LLM and are facing a completely different problem than what the underlying issue is

Beyond this, I also receive AI outputs as replies there as well.. And at work I cant be as open about these thoughts since our bonuses rely on it :sweat_smile: .. Very unfortunate, but I wholeheartedly believe it will get better

6 Likes

I’m waiting for the eventual and inevitable scenario that this post, just like many other posts that have the slightest mention of AI, becomes a warzone that is only laid to rest with intervention from the mods.

4 Likes

:rofl: im hoping it stays civil, genuinely I want to hear all sides, just not stark “its the future, get used to it.” brick-wall arguments which are born of emotion

3 Likes

Nobody really wants this:

3 Likes

ChatGPT users when they ask if the sky is blue

wyp07fmsr91g1

5 Likes

The other day I was really stuck on something (I forget exactly what now). Basically, I was getting errors that I couldn’t trace back to any specific code. So after Googling, I decided that before I spent the time to make a post, I would try out Claude and see if it could help me debug it. Because admittedly, it’d be faster. Also, Claude now gives your references and links to threads from where it’s pulling answers.

You know what it did?

It regurgitated a post I had made a year before on the Godot forums and tried to use it to help me debug my problem. None of it was relevant, and it was quoting me to me!

I ended up realizing I was going to have to figure it out myself. I finally did solve the problem. I think by looking at the Godot source code, and then commenting out swaths of code until I figured out where the problem lay.

LLMs are spicy autocomplete machines.

14 Likes

Best image I’ve ever seen on r/wallstreetbets.

8 Likes

Agree with @Demetrius_Dixon
I have seen these 4 kinds of help topics recently :
1.About tilemap,navigation and grid
2.ARRAY,
3.Complex BUT useless thing,
4.Resource

Now the OP’s use AI duh

1 Like

Commenting here because I got tagged from your original edit, lol

Imho the world is changing - We’re in a Eternal September phase of the internet, where the new inbalance is that the effort required to create content is now less than the effort required to consume it.

I think we’ll continue to make changes to how things operate. Github PRs should probably require a certain level of trust to make (i.e they have to be small at first), and forums should have rules for no AI generated posts/comments. Perhaps the Github issues can go through a pruning process where some volunteers confirm the issue before the maintainers waste time looking at it.

I think we’ll continue to make changes to how things operate. Github PRs should probably require a certain level of trust to make (i.e they have to be small at first), and forums should have rules for no AI generated posts/comments. Perhaps the Github issues can go through a pruning process where some volunteers confirm the issue before the maintainers waste time looking at it.

This is assuming the humans that are supposed to do this:

  • are with enough
  • have enough time
  • are willing to do it all for free

Who will organize the logistics for this? A couple of agents?

We’re in a Eternal September phase of the internet

Nothing lasts forever

Seems more like a couple of failed harvests to me. Where everyone’s looking at the barren soil, wondering whatever was wrong with the latest pesticide we sprayed over the crops.

1 Like

You’re addressing exactly this issue, right? Coincidentally, my favorite vlogger dealt with the subject today..

3 Likes

There might be hope, as the token costs for coding shops using Claude (and the other fuckery thingies) are having to face the fact that replacing Junior coders with LLMs costs more than their salary + benefits (+ mythical bonuses)

Source (in French) :

Cheers ! Have a good week one and all

5 Likes