Yes, that’s exactly right—and as far as the policy is concerned, this would be no different from a human deriving their contribution from existing open-source work: rewriting from principles is okay, copy-paste is not.
That’s fair, and you’re right that adopting a policy that takes a stand on a social issue could attract developers who align with that cause. That said, I do think this could cut both ways; the surveys linked above show that a majority of developers have adopted the use of LLM tools in some form, and an oppositional policy could alienate those developers. Which is why I would generally favour a neutral policy, though (as in this case) I recognise that even a seemingly neutral choice is a choice which could be considered disagreeable.
I’m truly sorry to hear that. I respect your choice; know that I appreciate the thoughtful manner in which you brought up the topic, and of course sincerely hope you do remain a part of this community.
My view of AI is more negative than positive and so I don’t use it. I am worried about the cognitive offloading the most. And seeing rdb suddenly start using em dash in what should be a genuine discussion about the use of AI is worrying to me.
I have used em-dashes for as long as I can remember; I have for a while avoided them because they have come to be a telltale sign of AI-generated prose, and more recently decided I would use them again because I simply refuse to let AI take a useful punctuation mark away from me. And note that @Thaumaturge used them in his posts as well.
I strongly take issue with your implied accusation that I am using LLMs to generate my posts. I have a strong distaste for LLM-generated prose and its instantly recognisable style, on top of the disrespect it would signal for the discussion at hand. I have never in my life copy-pasted LLM-generated prose and passed it off as my own.
For what it’s worth, in my (limited) experience, LLMs tend to surround the em-dash with two spaces, my personal style is not to do so.
At end of the day this is about trust, as we can see from your implicit accusation. If Thaumaturge doesn’t trust rdb he can simply stay on version up to which he did trust rdb’s work. It’s a fair thing. For e.g. If rdb started working for Epic games perhaps we’d question future work on panda as well, who knows. I know rdb (and his fantastic track record!) enough that if I wasn’t LLM user yet this would be the point where I’d go and take a look at this “fable model”…
I am suspicious by nature. Maybe it is just the agent language rubbing off on you.
I know that Thaumaturge likes writing verbose comments so it wouldn’t surprise me that much but it is a weak argument if it is supposed to be one.
From my limited experience text generators don’t usually use spaces (example deepai):
Also in this article you can see chatgpt and google ai use no spaces.
I am not claiming that you use(d) it. I am not expecting you to defend yourself, since we can’t really prove anything.
And to your benefit, I found a rare, pre-2021 instance where you used em dash. (with no spaces)
LLMs are trained on human text, or did you think that the em-dash was invented by AI companies? Surely their use implies that the em-dash is a common enough feature in the English language? It would be more apt to say that human use of the em-dash has “rubbed off on” LLMs.
What would be fair to notice is that I’m probably measuring my words more carefully than normally on the forums, due to the divisive nature of the topic, which would explain why I’m putting more effort into crafting my posts and using correct interpunction than I might for casual question-answering.
I will take your word for it; I haven’t used these systems often enough to have noticed. As for DeepAI, I don’t know what that is.
I would strongly advise you, then, to carefully consider the effects that a false accusation (or the suggestion of one) might have on someone personally, before flinging an unfounded assertion into the public discourse.
It is quite interesting to see how experienced programmers seriously talk about the contribution of AI to open source. You probably just forgot that writing working code requires serious brain activity and extraordinary solutions, as well as confident knowledge of the engine.
If you have an example of AI code that adds pathfinding, it’s similar to PandAI.
The use of shortcodes in violation of copyright has been common before. No one demonstratively refused to use the Panda3D engine under such circumstances.
At the moment, the artificial intelligence system is unable to make a significant contribution, the problem is far-fetched, all it can do is generate ridiculous code in a few lines.
@Thaumaturge has taken the initiative not to use the genAI code, but @Thaumaturge has not figured out how to recognize such a code…
I can raise my concern without some hard evidence. If you don’t use em-dash for months and months and then decide to use it twice in a discussion about AI, that is suspicious. Like you are asking for it or playing some prank.
Well, you don’t need to take my word for it, you can reflect on your own experience using it.
As I said, I reject this chain of reasoning (em-dash, therefore LLM) entirely, and I’m further concerned about the notion that we should dumb down our language out of fear of being seen as an LLM—so I will quite happily keep using the em-dash, thank you. I would encourage others to do the same.
Ha, I’d forgotten all about that; I concede that LLMs can evidently use unspaced em-dashes as well. (That post was a joke, as should hopefully be obvious by the context, and I assume you noted my disclosure at the top of the post.)
As for the topic at hand, I will think a bit more about the points made in this thread and see if I can come up with some concrete suggestions for improving the contribution policy. Other ideas are still welcome.
