Long message here. Sorry in advance! 

;TLDR; The AI policy page needs edits which I feel unqualified to make, but I 
think a discussion by the group as a whole is warranted. Delete if you're not 
interested. 

In another thread, the GnuCash AI policy (wiki.gnucash.org/wiki/AI_Policy) was 
cited as a concern for script consideration. I was unaware of this page and 
consulted it. 

I appreciate the effort that John Ralls has put into this page, and the amount 
of welcome detail included therein.

However, I see a number of areas that concern me about the page, and I think 
that the larger group should examine the page and weigh in.

First off: there is, actually, no policy stated on the page. The Opening 
Section raises a number of legitimate concerns regarding AI use and ethics, 
which strongly suggest that the GnuCash team does not want AI code included. 
But then the Other Projects section presents information from numerous other 
MAJOR open source projects that cover a very wide range of AI acceptance. This 
section in essence undermines whatever position may have been the team's 
intention.

In other words, the two sections make it difficult for a reader to articulate 
what GnuCash's policy is or might be. That needs to be fixed, most likely with 
a GnuCash Policy section that unambiguously states what the team will or will 
not accept. See below. 

Another concern I have is that the opening section includes a range of concerns 
that are fundamentally different in their focus. I think some consideration 
should be made as to how these are presented or argued. Perhaps, for example, 
quality and skills could be grouped and addressed together, with a focus on 
results for GnuCash code, while the legal issues could focus on GnuCash 
liability. 

I won't quibble about the section about quality, except to note that bad 
quality code can come from any source--including and especially from human 
contributors. One reason I don't contribute code anywhere is because I SUCK at 
coding, and my code is terrible. I'll also note that the tools have advanced at 
a staggering pace, to the point where Linus himself is embracing their use in 
the Linux kernel. 
(<https://lore.kernel.org/linux-media/CAHk-=wi4zc+ze8e+p3tmv8ttg_80kzsz1syl9anbtmeh5z4...@mail.gmail.com/>)

I have no qualms about the provenance piece. That's clearly an area that John 
has a lot of expertise! 

I personally don't have the legal expertise to judge the conclusions of the 
copyrightability of AI code, and while I respect John's greater knowledge and 
experience in this area, I question whether anyone can tell where these legal 
boundaries lie, and where they will end up.

While the assertion that AI generated code cannot be copyrighted is accurate at 
this time, I suspect the boundaries on this will move substantially when the 
legal world gets involved. For instance, I guarantee that someone will file a 
case claiming that their code deserves protection based on some percentage of 
their own oversight and contribution, and the courts will issue some kind of 
decision that will piss someone else off, who will then appeal, etc. And 
eventually, the superannuated fuddy-duddies in Washington will call for 
hearings (some one of them will angrily insist that they need to see this 
Claude guy in the witness box Im-me-di-ate-ly!) and eventually create a badly 
crafted law to establish how copyright should apply in these situations. But 
I'm definitely not in a position to know where this ends up. 

Moving on, the question of skill development is fair, but one that is much 
larger than the GnuCash project, and I question whether it's relevant for 
GnuCash's AI policy page. To put this in perspective, college students around 
the world are questioning whether they should stick with Comp Sci as a major, 
given the market trends. And while it's in GnuCash's interest to have skilled 
coders contributing to the project, see my note above about quality code. AI 
isn't necessarily the problem; lazy people who think they can get complex code 
handed to them without doing their own work are. 

I personally agree with the issues regarding LLMs as netizens. 

Because I am Unqualified But Opinionated (UBO), I'll offer here a suggested 
addition to the wiki page. It could go at the top or the bottom, as appropriate:

---------
GnuCash AI Policy
Given the concerns and issues raised here, as well as the overall uncertainty 
about how AI-generated code fits in computing universe, the development team 
for GnuCash has determined that code contributed to the project should contain 
only code that the human contributor affirms is as accurate, error-free, and 
efficient as they can determine. 

Ideally, this code will not include AI-generated code, but if it does, the 
submission MUST:
1) State clearly the use of AI-generated code.
2) Name the AI tools and specific prompts used.
3) Identify the extent of AI-generated code.
4) State workflows used to evaluate and process AI-generated code. 

Finally, the use of AI-generated code will likely result in additional scrutiny 
of such code to verify its quality, accuracy, and legitimacy and may result in 
outright rejection of such submissions. 
----------
David T. 

_______________________________________________
gnucash-user mailing list
[email protected]
To update your subscription preferences or to unsubscribe:
https://lists.gnucash.org/mailman/listinfo/gnucash-user
-----
Please remember to CC this list on all your replies.
You can do this by using Reply-To-List or Reply-All.

Reply via email to