World Intellectual Property Day: A crunch point for journalism, culture and democracy
This yearʼs World Intellectual Property Day – celebrated on 26 April 2023 – is marked by critical technological developments that will change the relationship between citizens and creative works – including, crucially, the creativity that goes into producing news reporting that is factual, ethical and engaging.
The International Federation of Journalists (IFJ) represents more than 600,000 journalists in 140 countries worldwide, and the European Federation of Journalists (EFJ) over 320.000 journalists in 45 countries.
We are committed to quality ethical journalism, recognising that this depends on journalists making a living.
Machine-learning systems are “trained” on the work of human authors, among whom are journalists, photographers and illustrators. They respond to questions by predicting what words or images these imply. The text systems made public now – such as ChatGPT – have rapidly become notorious for “confabulating” – that is, inventing fake stories, including fake source references, and doing so with a fluency that implies confidence.
For two centuries democracies have depended on ethical journalism to give citizens truthful information that informs their voting decisions.
Anti-democratic forces have, though, long sought to undermine the very idea of truthfulness, to promote authoritarian populist appeals to fear and anger over truth.
Attribution is essential to trust
If words, images and other works generated by machine-learning systems circulate without being declared as such, the world faces a bleak future in which authoritarian populists can convince their prey that “nothing is true”. It is most obviously destructive of human societies if individuals lose trust in news reporting or academic research. Artistic creative works must also be clearly distinguished from “AI” pastiches.
We believe there is an urgent need for legislation to provide that when AI outputs are published or made available to the public they must be labelled as such. The corollary of this is that the need to give attribution to human authors is increased, to give the consumer a “guarantee” from a fellow-human. The advent of the internet showed how essential are the “moral rights” of authors to be credited and to defend the integrity of our work: AI makes them foundational. Machine-learning outputs should reference the works from which they are derived.
Advancing knowledge, science, arts and culture
Authors’ Rights exist to advance knowledge, science, arts and culture. Authors have provided the foundations for AI technologies – as our works have been “scraped” to build training sets, without permission or remuneration. Some believe that machine-generated works may drive authors out of the market.
Established copyright laws must not be weakened based on a mistaken belief that doing so is necessary to incentivise innovation.
Licensing remains the proper way to grant access to authors’ works. Collective management is particularly beneficial when there is a need to use large numbers of materials from numerous rightsholders, as is the case when using works to train AI systems. Mechanisms exist to extend collective licensing remuneration to authors who are not, yes, members of collecting societies.
Authorship is a human activity. Human authors must get recompense when our work is used to train “AI” and machine-learning systems.
For an example see www.londonfreelance.org/fl/2303ai.html