introducing grpm

[zen_desu]

Not enough momentum and things don't move, but if every change is "5 steps in some direction" I think that is a recipe for issues, and makes collaboration extremely difficult.

Imagine you have someone who wants to help, they look at some bit of the code and think "oh i can improve this" and next week that entire section of code is changed. Do you think they will say "oh time to review all of these changes and discard my work" or "well I'm not part of the AI junior dev team so why bother"

Would you accept a 1k line commit from someone who said it was AI generated and worked on their machine? Would you ask AI to review the code they shared? would you take the time to read every line and be sure it's good for the project?

During the rapid development phase, there's no point in contributing; testing and creating issues are sufficient. It's no longer a problem to write code as before; the problem is then polishing it and making sure the software runs like clockwork. Perhaps if it's a completely new feature that wasn't planned, then it makes sense, but unless it's something unique, there's no point in creating a PR at this stage.
Although there was a recent precedent in the repository https://github.com/coregx/gxpdf, where someone wrote about 10 high-quality small PRs, which I accepted after his revisions. Whether he used AI or not, I didn't question it, but his contribution was valuable. So I think there's no point clinging to the old development style; the world has already changed, and there's nothing we can do about it...

I don't think approaching this with a sense of helplessness really helps anything

[b11n]

What will the haters say now?

Would you rather this be an AI discussion thread or about your software?

Exactly. Make up your mind, mistbow. You're under no obligation to engage with skeptics, you don't owe me anything, and I don't care if you call me a hater, but if you're going to have a sook about people expanding on the topic and then link some guy's tweet just to rage bait, it's not a good look.

I wonder what they'll say to the author of that tweet... I don't mean just this topic, but in a broader sense, all skeptics.

Nothing. I don't actually hunt down AI-coders across the internet looking for fights to get into. If you want to know how AI skeptics are responding to that tweet, go and look on Twitter. Don't try to drag more drama in here while ignoring your own.

[Banana]

Moderator hat on

I love discussion, even if they are sometimes slightly heated. In this topic I see a trend, we as moderators, do not want here.
There are some good replies here, nothing about them.

Please be civil, on topic and useful. Beeing bad or angry does not help here.

So, a hint:
If you want to get some feedback about something, good, but don't force it. Right now it is drifting into a bashing thread without any usefulness.

[dmpogo]

I did not extrapolate anything onto the case of LLM, noir built any argument for how to use it. I stated on what principle copyrighted works can still be used if the author is unknown and showed that the blanket statement "I do not know the author of prior work thus I cannot use it" does not follow from the copyright law.

As you brought it up in this thread I think I was reasonable in assuming you considered it germane to the topic at hand.

You said the rule is:

So basic rule their - if the author is unknown, reasonable steps should be taken to establish the author,

I don't know the law (less still the law where you are, although I imagine it descends from ours), so please correct me as needed: but I believe the "reasonable steps" are taken to defuse liability. If an author (or copyright inheritor) of one of these historical letters made themselves known and asserted their copyright, the journal would not be liable to any remedy against having published it in the first place, but it would be required to withdraw from further publication (or agree terms with the author).
[/code]

Basically yes, copyright is inalienable from the author (except by his explicit decision), so the "unknown author all steps are taken" protects that use, but not further uses, if the author becomes known/contactable. In my area, there is no such thing as "further publication" (in a sense, what is published is published, it is not that I cannot reprint it) though if it is a second edition of, say, a published book, one, I believe, needs to revisit copyright issues. Nor it absolves from due diligince further authors using the same material, though they have an option not to use it directly, but to refer to the previous publication.

Code: Select all

Let me try to clumsily juxtapose the LLM context onto yours. If the institution from which your wife sourced a letter told her: "Yes, we knew who the author was when we took custody of it, and we knew we should have obtained a licence from them if we intended to share it with anyone, but we decided not to bother and then we burned the record of their identity, so we can't tell you." She knows that she has obtained it in possible violation of the author's copyright; does that satisfy the journal's lawyers?[/quote]

This is beyond my legal expertise,  however I believe (guessing) it largely does, though the situation when the source will explicitly say that they share this letter in contravention to copyright  of the author that they knew about before has never been encountered in her practice.   Usual answer is just "we don't know who is the author".   Moreover,  the author is often known (letter signed) but if reasonable attempts  to contact him/her/heirs failed,  it is usually enough.