Re: [Hamlib-developer] The use of LLM generated code in Hamlib (long)

[Greg T. <gd...@le...>]

"Mikael Nousiainen" <mik...@fa...> writes:

> On Fri, Feb 27, 2026, at 16:16, Greg Troxel wrote:
>> "Mikael Nousiainen" <mik...@fa...> writes:
>>
>>> Did we ever land to any decision regarding LLM usage in Hamlib?
>>
>> This is an interesting opinion, and I think it has considerable merit.
>>
>> https://lists.debian.org/debian-vote/2026/02/msg00060.html
>
> Thanks for noting this. However, I read this particular opinion
> saying: "there is no way genAI can be done right and we should
> therefore abandon it" - which is a rather extreme viewpoint that I
> don't share. The fact that many large AI companies are not doing a
> particularly good job ethically (training their models) should _not_
> mean automatically that "all genAI is bad and will be like that
> forever". Not sure if I understood something incorrectly?

I think you're technically correct but not correct in practice.  I agree
that it is theoretically possible that genAI could be done properly.

(Separating ethics from unsettled questions of copyright law.)

Can you point to a single example of an LLM that has been trained solely
with content where the author/copyright-holder has truly consented to
such use?  I don't mean technically clicked through something they don't
understand, but true consent.

> Exactly. How can we trust any contributor, since "faking it" (=
> generating the code) is easier than ever and requires practically no
> effort? Also, I'd argue not all LLM-generated code is automatically
> proprietary - just like not all StackOverflow or "whatever Google
> gives you" is proprietary. It's a rather complex issue.

Sorry, I didn't mean to say LLM code was proprietary here.  I just meant
if there are rules, we currently trust people to follow them.

> My intent was more like: How can we keep Hamlib alive and thriving
> when we know masses (including many of Hamlib's current and potential
> future contributors) are going for genAI? And: can we find
> responsible/ethical ways of using genAI?

The first part I think is the key question.

I guess it's a real question how much is lost by saying no.

The larger question is if it's a generational thing and that in 20 years
will everyone who participates think it's fine to use LLM code.  I think
it's to early to presume a societal outcome.