[License-discuss] Curious about derived works and AI...

[Stefano Maffulli]

On Tue, May 28, 2024 at 10:14 AM Bruce Perens via License-discuss
<license-discuss at lists.opensource.org
<mailto:license-discuss at lists.opensource.org>> wrote:

    It is unfortunate that Japan made the choice that it did, because it
    makes Open Source software fair game for those who would profit from
    the work of our developers without attribution or remuneration and
    entirely outside of their license terms.


You should realize that you're implicitly arguing for an extension of
copyright at the expense of freedom of research, open knowledge and open
science, before Open Source.

I'm no fan of megacorporations and I'm not interested in giving them
more powers: But you need to realize that  the mega corps already have
swallowed your code and writings and pictures and use patters, graph of
friends, etc, through perfectly legal means, like the GitHub terms of
service. And they'll continue to do so with bilateral deals (Reddit,
Stackoverflow, etc.)

Your argument is going to deprive open research groups like Eleuther AI,
LLM360, LAION and future ones of an opportunity to compete with the mega
corps, imposing restrictions to text and data mining that will limit
only the **real** Open Source AI.

I  think that Japan is on a better path, although incomplete: Maybe a
better balance with open science and society would be that text and data
mining be made unrestricted **provided that** the results of training
(the model parameters, weights and biases) are also made publicly
available. That'd probably be a better tradeoff.

In any case, this is a very complex debate. Creative Commons started
exploring it
https://creativecommons.org/2023/10/07/making-ai-work-for-creators-and-the-commons/
and Open Future published a very interesting primer on the issue that I
highly recommend reading
https://openfuture.eu/publication/towards-a-books-data-commons-for-ai-training/,
it's a good introduction to the trade offs.

/stef