Every time an LLM-generated PR gets merged to any repo, my trust for it goes down a bit. Glad Godot is safe for now. Hope it stays that way.

Show Post

That's not even considering the emotional hit of getting a perfectly reasonable pull request automatically rejected by an AI because the dice landed the wrong way. Delegating a technical decision upon a probabilistic model is a stupid move, and no reasonable project should opt into doing this.

Show Post

LLMs can't even "agree" with themselves upon repeated querying. You could try this and get a different judgment each time, regardless of what the worth of a contribution might be. It's not going to save anyone any stress if maintainers need to double check the decision the machine made.

Show Post

It's going to be a burden maintainers either way. What makes you think an LLM can make a good, automated judgement call on the worth of a contribution regardless of whether it's made by a human or another LLM? LLMs are not a tool for decision making and it's naive to think that it is.

Show Post

and AI judging human contributions is especially not a good thing AI shouldn't be making any judgement at all

Show Post