The pattern is clear. Before creating Facebook, Mark Zuckerberg did an experiment: he came up with Facemash.com from hundreds of photos of Harvard classmates, and was accused by the Department of Information Services of violating the rules of the university. For the creation of the Image.net image bank, designed for deep computer vision learning, the researchers enlisted the services of Amazon Mechanical Turk, which pays paltry wages for mechanical work, leading to millions of bad photos. labeled (and a piece of art that went viral: ImageNet Roulette, by Trevor Paglen and Kate Crawford, which denounced the aberrations of the scientific project). Cambridge Analytica went further and stole thousands of personal data on Facebook to manipulate voting orientation. The history of the internet is rife with theft that fuels exercises in behavioral engineering and surveillance capitalism.
There are currently several million-dollar lawsuits against OpenAI, which appears to have followed this illegal path to train its GPT language models, and against Stability AI, which may have misappropriated photos from the Getty Images online archive. . In order to train the artificial intelligences that generate text and image, information from all available sources has been accumulated. The statistical generation has learned from our digital libraries, with our permission and without it.
“Every act of creation is first of all an act of destruction”, said Pablo Picasso. He was referring to wild artists like him, but human societies have also followed that slogan. The history of culture is a succession of great expropriations, material and symbolic. When the Sumerians conquered a city, they requisitioned both the tablets of their adversaries and their best scribes. As Lionel Casson explains in Las b, the woman from Alexandria accumulated her monumental heritage because she seized the scrolls that were in all the ships that docked in her port, to copy them. The great European museums show the fruits of a looting of centuries. Jeff Bezos built the Amazon empire thanks to a double initial appropriation: that of the prestige of the book and that of the name of the great ecological reserve of humanity.
The difference in the case of OpenAI has to do with speed. There has never been a massive appropriation in such a short time, nor have such remarkable results been achieved thanks to it. But everything points to the end of the permissiveness that has characterized our relationship with the big technology companies: the reactions have also been very fast. The theft of the century is giving rise to the debate of the century. In social networks, journalism, classrooms and courts.
