Cette séance du séminaire du CIS s’est tenue en ligne le mercredi 7 octobre 2020 de 14h à 15h30.
Artificial Intelligence and Copyright: The Way Forward
EN. We are living through a revolution in the production of artistic, literary and musical works that have been generated in some shape or form by artificial intelligence (AI). Take the following paragraph:
“One can have many opinions and still have strong values and morals. The phrase “all opinions are equal” is only the correct one when we understand that no opinion, no matter how stupid or contrary to reality, will ever be given any special place in our discussion. No one has the right to censor you because of your opinions…”
These is a readable text, but it doesn’t make much sense, you may expect it to come from a non-native speaker, or perhaps to be the ramblings of a bored Internet users. However, the text was generated by a language model called GPT-2, developed by Open AI.1 This is just one of the many tools that are being developed that can produce uncanny works that mimic human creativity. Take two other Open AI tools, Musenet and Jukebox. Musenet can produce musical compositions for ten instruments having been trained in classical and popular music; while Jukebox can produce music, lyrics and singing in the style of several artists such as Elvis Presley and Katy Perry.
Other AI-generated works have been widely reported in the press in recent years. For example, the AI painting called Edmond de Belamy made the news when it was sold in an auction for $432,500 USD. The painting was made by the French art collective Obvious using a machine learning algorithm designed specifically to generate images, known as a Generative Adversarial Network (GAN). The artists fed the AI over 15,000 portraits from various epochs, and produced a set of portraits of the fictional Belamy family.
There has been considerable discussion in recent years with regards to the copyright implications of the above increase in the creative capacity of artificial intelligence, and there is no consensus on the legal protection given (or not) to AI works. This presentation will discuss some of the questions, but most importantly will try to look beyond the hype to ask real questions of the viability and reach of AI creative tools, and whether we need to make changes to the law.
Dr Andres Guadamuz is a Senior Lecturer in Intellectual Property Law at the University of Sussex and the Editor in Chief of the Journal of World Intellectual Property. His main research areas are on artificial intelligence and copyright, open licensing, cryptocurrencies, and smart contracts. Andres has published two books, the most recent one of which is “Networks, Complexity and Internet Regulation”, and he regularly blogs at Technollama.co.uk. He has acted as an international consultant for the World Intellectual Property Organization, and has done activist work with Creative Commons.
1 ↑ The text was created by entering the text “but this individualism is a sham” as prompt in “Talk To Transformer”, a GPT-2 model implementation, found at: https://talktotransformer.com/.