Re: Re: Proposal -- Interpretation of DFSG on Artificial Intelligence (AI) Models

> Is the change technical or legal/philosophical? You could call this
> a Turing test for copyright.
This is not a new issue at all. I remember that back in the day in
order to legally reverse engineer a computer program, companies had to
set up two separate teams of developers.
One team reads the code and writes documentation. The second team reads
the documentation and writes the new code. It was crucial that no
member of the second team sees the original code in order to rule out
any copyright issues.

> Processing of experiences into expert opinion is IMHO not directly
> comparable with compilation of source to a binary.
DSFG does not only apply to programming languages and program binaries.
For all data blobs in Debian packages, it is preferred to include the
scripts that generate it, for images it is preferred to have the SVG
code over the generated pixel graphics, etc.

For a reason, the relevant licenses do not define “source code” by
being in a programming language readable by humans. They define it like
this (example from GPLv3):

> The “source code” for a work means the preferred form of the work
> for making modifications to it.

In that definition, training data is quite obviously relevant. No one
tweaks neural network model weights manually.

Compare this to the previously mentioned example of S-boxes in
cryptography. They are small and usually created manually.

Regards
Stephan

-----BEGIN PGP SIGNATURE-----

iHUEABYKAB0WIQRB1rjSpCJd8a7h6mNgNUJZCjx8YgUCaA+i0wAKCRBgNUJZCjx8
YhAlAP0S6RiRSBO5HWAgrz3qxtAy7rZtKMQ6SD+iUwfrv97RHQEAsVty3X0BDav8
PGf0grE7hfxUse5WKVB9aa/MPpMlJwM=
=wMch
-----END PGP SIGNATURE-----