[08:08 Tue,28.May 2024 by Rudi Schmidts] |
According to Bloomberg, Meta and Alphabet/Google have recently offered various Hollywood studios millions of dollars for licence agreements to use feature films as training data for their own generative AI models. According to the report, however, neither Netflix nor Disney are in principle "willing to licence their content", but have at least "expressed interest in other types of collaboration". Diametrically opposed to this, Warner Brothers Discovery Generally speaking, courts around the world do not yet consider pure training to be an infringement of copyright. Conversely, unsolicited data can of course turn out to be real time bombs, as individual data can no longer be "extracted" from an AI model at a later date. If a plaintiff were to win a lawsuit here, this could, in the worst case, lead to a complete ban on the use of a model or to astronomical additional payments. This is why more and more AI companies are currently concluding licence agreements with media companies. Just earlier this week, OpenAI and NewsCorp signed a multi-year agreement to provide news content for ChatGPT. And according to Business Insider, Meta is also considering paying some publishers for training access to "news, photo and video content". Although unsolicited use is usually impossible to prove in concrete terms, clear details can emerge by chance in AI output that can undoubtedly be attributed to a source. As was recently the case with the example of an unusual pizza recipe covered in glue. more infos at bei www.engadget.com deutsche Version dieser Seite: Alphabet und Meta wollen Hollywood Spielfilme als KI-Trainingsdaten nutzen |