Scientist Margaret Mitchell proposes model to compensate ordinary users for content published online (if it is used to train Artificial Intelligence).
A generation of future artificial intelligence (AI) systems could financially compensate ordinary people, artists and authors for the content they publish online — as long as it is used to train artificial generative models.
The hypothesis was presented by Margaret Mitchellscientific director of ethics at US open source AI company Hugging Face, at the event , which took place in Cairo, Egypt, this month.
The researcher maintains that artificial intelligence companies should develop mechanisms capable of tracking the results produced by generative models back to the original sources that helped train them. If they did, it would be possible to identify who contributed to certain AI-generated content and create proportional forms of remuneration to this contribution.
m-mitchell.com
Margaret Mitchell, founder of AI company Hugging Face
The problem, explains the computer scientist, is not limited to professional creators, such as writers, musicians or visual artists: ordinary Internet users also end up feeding these systems with texts, photographs, comments and other publications left over the years on digital platforms.
We would be living in a world where, deep down, we are all content creators. Therefore, a possible compensation model should cover anyone whose content has been incorporated into training bases.
Mitchell also guarantees that there are already technological approaches that could be adapted to make this scenario viable. Among them are clustering algorithmsmachine learning techniques that allow you to find similarities between data sets and help in attributing authorship. These methods could even be applied to large linguistic models already developed, explains the scientist.
The main obstacle to the idea would be the current business model of AI companies, which does not favor investment in tracking and compensation infrastructures, but some companies have already started to explore the idea. In 2021, internal Anthropic documents later made public by court order revealed that executive director Dario Amodei considered a payment distribution system similar to the Patreon platform model.
To protect privacythe proposal would depend on user consent. Each person could choose to associate their identity with their data, becoming eligible for allocation and eventual payment.
“There is a lot of content to be appropriated from creators – including artists and authors, but also ordinary people who write online – who are not being compensated for this work”, said the scientist at the event, cited by , in a statement that recalled the growing use of creative works by AI systems without authorization or payment – the case of Japanese director and animator Hayao Miyazaki and the reproduction of the visual style associated with the visual style immediately catches the eye.
“I imagine a future where we are able to attribute and identify who was present in the entry space that enabled [a um modelo] create your exits, and compensate them”, points out Mitchell.
