In a recent interview with The Wall Street Journal, Mira Murati, the CEO of OpenAI, discussed AI’s creation of the video Sora. When asked where Sora trained the video data, Murati mentioned using public videos from various sources and licensed videos. However, when questioned if YouTube videos were used, Murati refused to respond, claiming ignorance.
This response sparked further interest when Neal Mohan, the CEO of YouTube, gave a special interview to Bloomberg. Mohan was asked if he knew whether OpenAI used YouTube videos to train AI Sora. Mohan did not confirm if Sora utilized YouTube videos for training, stating that doing so would violate the platform’s Terms of Service (TOS). He also commented that content creators who invest heavily in producing and uploading their work on YouTube expect the platform to adhere to guidelines, viewing downloading transcripts and videos as a breach.
Mohan indirectly implied that if OpenAI did indeed use YouTube videos to train Sora, it would be a misstep. When asked if YouTube allowed Google access to use videos on the platform to train AI, Mohan confirmed that, in general, training AI would breach terms but agreements could be made with consent from creators.
TLDR: CEO responses regarding the use of YouTube videos to train AI Sora by OpenAI spark controversy over the infringement of platform terms and content creator expectations, requiring a clarification of guidelines and agreements.
Leave a Comment