SAccording to author Frank Wynne, in his book I was Vermeer, the Dutch forger Han van Meegeren painted as Johannes Vermeer, more than two centuries after the artist's death. According to the novel, Meegeren was born in the wrong era, and possessing skills that were highly valued during the Renaissance but were of no relevance in the era of the Cubists and Surrealists, he produced works in the Vermeer style. To do so, he used the best technique, skills and chemical knowledge, motivated especially by opposition to a curator of the time. He became rich from the works, but denounced himself after the Second World War, so as not to be considered a collaborator of Nazism, proving that he had deceived the Reich when he sold them works by Vermeer. The fake works are so perfect in style and chemical elements that, to this day, experts remain suspicious about the true authorship of some works in the collections of major museums.
This story perfectly illustrates what is happening today with the production of works by artificial intelligence. The difference is that it does not just copy an author, nor does it require an artist with super specialized training. Generative AI learns from all the works available on the network and, based on specific commands, creates new works as if determined by the original author. It is important to understand that it is not an identical or nearly identical copy, but a new work, made from all the references, style, technique and mannerisms of the copied author. Thus, it is possible to produce pseudo-new works by Picasso, Matisse or Renoir without any participation from their respective authors.
In this age of information avalanche, it does not seem unlikely, therefore, that a work created by AI could be “discovered” as if it were by its original author, follow the path of Vermeer’s fake works and end up on the walls of a museum. The point is that what seemed romantic and under control – after all, to paint like Vermeer, the plagiarist needed to be exceptional – can now be scaled and given credibility on the networks.
Furthermore, AI has been creating new works by combining ad infinitum the works available on the networks. In this case, these are no longer works presented as those of a specific author, but of new authorship (who programmed the AI?) made from all the works created by the most diverse authors in Brazil and around the world. AI learns, free of charge, from these human geniuses, and monetizes the resulting works. This AI learning is not to be confused with human learning, whether due to the scale of storage, classification and synthesis, or the fact that it does not contribute to the purchase of books or the payment of royalties, courses and schools. It is a parasitic activity, and quite profitable. These are productions that do not have an author, as we know it until now.
The current result, which many argue will be surpassed, is still very superficial. However, the production of these works by AI will become increasingly perfect and sophisticated, and has the clear potential to impact the artistic production market, among other things: circulating fake works as genuine; creating works that pay the owner of the technology and not the author and artist; evading payment of copyright and image rights for works used by AI; scrambling the authorship and the figure of the author after an infinite series of reconstructions.
Thus, the works generated by AI will be the current, technological version of Vermeer's works, which still lack certainty as to their authorship. However, with the added problem that they will be produced on an uncontrollable scale, with the use and exploitation of third-party creations and without any tracking of networks. The solution for authors and artists will be to endorse what is human production as a work. made in human. But will we have time to prefer works made by humans?
*Cris Olivieri is a lawyer, director of Olivieri & Associados Advogados, specializing in culture, art and entertainment law.
