Date: Fri, 31 Jan 2025 12:54:25 -0800 From: paige@paige.bio To: David Chisnall <theraven@FreeBSD.org> Cc: Sulev-Madis Silber <freebsd-hackers-freebsd-org952@ketas.si.pri.ee>, freebsd-hackers@freebsd.org Subject: Re: Provisions to the contribution guidelines for using LLM generated code Message-ID: <EB9FCB17-3197-499F-9961-AF4237498EF6@paige.bio> In-Reply-To: <1E478400-5DFD-4C45-B466-F29EFD76A29E@paige.bio> References: <49B92974-E37A-4786-A456-E258D5A1D35E@paige.bio> <4922BB4E-1361-4AE9-A40D-D75E4875033D@freebsd.org> <EA849D16-0FA6-4808-BEE1-D0D4AAFB898C@ketas.si.pri.ee> <7F5CCEEE-A8A9-459A-A2C1-9ADC31BC91C6@FreeBSD.org> <1E478400-5DFD-4C45-B466-F29EFD76A29E@paige.bio>
next in thread | previous in thread | raw e-mail | index | archive | help
Also I want to retract my previous statement about a renewing a patent, =
I was absolutely talking about something I apparently know nothing about =
there;=20
No, you cannot renew a patent in the United States. However, you can =
maintain your patent to extend its validity period. Explanation
=E2=80=A2=20
The term of a US patent is fixed, and it cannot be extended past the =
period for which it was issued.=20
=E2=80=A2 The term of a US utility patent is 20 years, and the term =
of a US design patent is 15 years.=20
=E2=80=A2 When a patent expires, its rights are discharged into the =
public domain.=20
=E2=80=A2 The term of a patent can be extended in rare cases, but =
these cases must be validated by a special act of Congress.=20
> On Jan 31, 2025, at 12:43=E2=80=AFPM, paige@paige.bio wrote:
>=20
>> In the second case, I have deliberately used a plagiarism machine
>=20
> I get what you're saying that it makes easy work of an otherwise =
difficult task but I don=E2=80=99t think that inherently is what makes =
it a plagiarism machine. I think people who have lives and kids to raise =
generally like to contribute anything that adds quality to their own =
life and given the circumstances will want to take the path of least =
resistance. It=E2=80=99s entirely possible for somebody with good =
intentions to use something like an LLM and for things like =
Microsoft=E2=80=99s obscure hash table patent to be completely lost on =
people who are responsible to say whether or not something gets merged. =
There are of course people who will blatantly break the rules with the =
intent to deceive and put things in places that they don=E2=80=99t =
belong but that is a different problem than the one I have in mind and =
my point is that even though the two are mutually exclusive they are not =
always handled in their own unique way like they should be and that=E2=80=99=
s unfortunate for people who have good intentions and the overall =
reputation of LLMs. =20
>=20
>> Microsoft has issued an explicit patent grant of the exFAT patents =
*for Linux*. The =E2=80=98Open=E2=80=99 Innovation Network
>=20
> Sorry to mix threads here, but you=E2=80=99re right and this is also =
what I mean; a lot of people might see something has a GPL =
implementation and won't immediately arrive at the conclusion that =
it=E2=80=99s only because they have permission to implement that idea =
and make it GPL. The only reason that I know any better is because =
I=E2=80=99ve watched Paragon Software for more than 20 years try to make =
NTFS-3G a thing for Linux users. If I=E2=80=99m being honest with you, =
Microsoft doesn=E2=80=99t just have an idea they have a monopoly on how =
you can exchange data between computers that effectively makes it =
impossible (still to this day) to use anything that they=E2=80=99re not =
vetting.=20
>=20
>> If a committer deliberately violates copyright, the code will be =
removed and the committer will, most likely, lose commit access.
>=20
> Honestly I know it doesn=E2=80=99t do a whole lot of good to speculate =
about what could become of LLMs at the moment, but I feel like if they =
keep improving this that pretty soon somebody will be able to generate =
their own driver for virtually anything they want and they won=E2=80=99t =
need to share it because anybody else will be able to do the same. For a =
few hours of work I already have:=20
>=20
> - a KEXT for ExFAT (compiles)
> - fsck_exfat (compiles)
> - newfs_exfat (compiles)
> - mount_exfat (compiles)
>=20
> And granted none of them produce the correct filesystem (it=E2=80=99s =
trying to) or handle a filesystem created by any other means (it also =
makes a concerted effort to do this)--it=E2=80=99s really close. I think =
we might actually see something that is powerful enough to create a =
solution like this given a prompt in the next couple of years and =
realistically contributions won=E2=80=99t mean much because people will =
be able to make whatever they want or need for themselves and they =
won=E2=80=99t have to distribute it.=20
>=20
> I guess what I=E2=80=99m wondering is how will FreeBSD stay relevant =
when this becomes a reality? I understand reality is much different from =
this as it stands but I also gather the intention is to improve LLMs to =
bring this reality into fruition. I think there=E2=80=99s an opportunity =
to embrace the technology that is coming, but that there should be rules =
and a vision behind it. I think it=E2=80=99s coming faster than a lot of =
people can even keep up and there might not be any time as good as the =
present to start thinking about it.
>=20
>> Next year, I believe, all patents on the original version of exFAT =
will have expired
>=20
> I mean... they could renew their patents, but one has to wonder =
towards what end at this point? As far as I know the only benefit to =
patenting something is just so that you can=E2=80=99t reproduce =
somebody=E2=80=99s idea and reap the benefits of redistribution. Really =
makes you think=E2=80=A6=20
>=20
> -Paige
>=20
>=20
>=20
>> On Jan 31, 2025, at 3:23=E2=80=AFAM, David Chisnall =
<theraven@FreeBSD.org> wrote:
>>=20
>> On 30 Jan 2025, at 12:03, Sulev-Madis Silber =
<freebsd-hackers-freebsd-org952@ketas.si.pri.ee> wrote:
>>>=20
>>> what happens if you take the word llm out and put a human in there?
>>>=20
>>> there are ton of fbsd contributors and i often wonder if some of =
them bring something in. apparently it's no "code-id" where we can put =
code for checks. esp i worry about all those linuxkpi things. where's =
the voluntary no consequences drug test that proves you didn't smoke any =
gpl before you opened code editor
>>>=20
>>> it's like llm is right out but humans are all ok?
>>=20
>>=20
>> No, as I said, the following two are equivalent:
>>=20
>> - I copy some GPL=E2=80=99d code (or code with a license that =
requires an attribution) and contribute it in such a way that violates =
the license.
>> - I use an LLM to copy some GPL=E2=80=99d code (or code with a =
license that requires an attribution) and contribute it in such a way =
that violates the license.
>>=20
>> The difference is that, in the first case, I *know* that I am doing =
so. In the second case, I have deliberately used a plagiarism machine =
but don=E2=80=99t know whether this specific output is copyright =
violation or not.
>>=20
>> If a committer deliberately violates copyright, the code will be =
removed and the committer will, most likely, lose commit access. =
Committers are responsible for the code that they commit, but if they =
are using a plagiarism machine then the chances of them committing =
accidental copyright infringement are much higher and that=E2=80=99s a =
risk to the project.
>>=20
>> David
>>=20
>>=20
>=20
>=20
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?EB9FCB17-3197-499F-9961-AF4237498EF6>