Opinion. Creators of all countries, it’s time to unite against AI

Back then, in 2023, the entire world was transfixed by the rise of OpenAI’s dazzling chatbot, GPT Chatthat metastasized at the speed of a fungal infection, accumulating tens of millions of additional users every month. This has resulted in multibillion-dollar partnerships and an influx of investments. Image generators using artificial intelligence (AI), such as Middle of the journeythen took flight.

But, a year later, things are no longer good: the sector is reaping few profits, with IT costs skyrocketing. And one issue threatens to bring this still incipient field of activity back to harsh reality: that of copyright (the equivalent of copyright in English and American legal regimes).

“A legal obligation”

At the end of December, The New York Times took resounding action against Microsoft and OpenAI, arguing that “Millions of its articles have been used to train chatbots that now compete with the newspaper as a trusted source of news.” The lawsuits join a multitude of others: illustrators, the image bank Getty Images, George RR Martin and the Author’s Guild (the largest professional organization of writers in the United States), or simple Anonymous users of social networks have notably initiated actions collectives.

They all protest against the use of the work of authors, journalists and artists, among others, by companies likely to profit from generative AI, without having requested the consent of the people involved or granted them any compensation, in defiance of their copyright.

The (American) Congress, researchers and AI experts are also increasing the pressure. So last week, at a congressional hearing, senators and media representatives agreed that AI companies should pay licensing fees on materials used to train their models. “It’s not just a question of morality, said Connecticut Senator Richard Blumenthal, who chairs the subcommittee that initiated this hearing, It is a legal obligation.”

The power of demonstration

Additionally, an uncompromising study, co-written by cognitive scientist and AI expert Gary Marcus and film industry expert Reid Southern, was recently published in IEEE Spectrum (an American technology magazine). According to this research, Midjourney and Slabtwo of the leading AI image generators, were reportedly trained on copyrighted materials.

“Companies have not been very transparent about what they are using. Therefore, it was important to demonstrate that they were using copyrighted materials,” explains Gary Marcus. Especially since it only takes a little for works that infringe copyright to be generated.

“There is no point in sending the AI ​​the written command ‘make a C-3P0’, you can only ask it to ‘draw a golden android’. In the same way, the simple expression ‘Italian plumber’ allows us to obtain a drawing of Mario.”

Leave a Comment