YouTube warns OpenAI: using its videos to train Sora is against the rules

litter This is by far the most efficient text-to-video conversion model available today. Artificial intelligence OpenAI It is not yet available to the general public, but there is already considerable debate about what materials it was trained on. Speculation surrounding the possible use of video YouTube It did not take long to arrive, although this has not been proven. And the leader of the platform did not hesitate to issue a warning to Sam Altman and company: If you use his video, you will be breaking the rules..

In an interview with Bloomberg, Neil Mohan, the CEO of YouTube, warned that extracting videos published on the platform to train AI models is a violation of its terms of use. However, he admitted that he had no “first-hand knowledge” that OpenAI was using a Google-owned service to train and improve Sora.

The executive branch also indicated that Even Google, its parent company, does not have unlimited access to YouTube for Gemini education., its next-generation AI. Mohan explained that not all creators publish content under the same licensing terms and that a minority of videos may be used by those living in Mountain View. But it all depends on the type of agreement signed by users.

As for OpenAI and Sora, Susan Wojcicki’s successor, she explained that the terms of use not only prohibit the use of full videos posted on YouTube. Also not allowed scrap fragments of clips or audio transcriptions as training material for language models.

Sora cannot be trained through YouTube videos.

Even though he doesn’t know that OpenAI is using YouTube videos to train Sora, it is clear that Neil Mohan does not stay away from rumors. “From an author’s perspective, when you upload your work to our platform, you have certain expectations. One of these expectations is that the terms of service will be followed. They do not allow the uploading of things such as transcripts or video clips that constitute a “Clear violation of these rules. These are the rules of the game in terms of content on our platform,” the executive noted.

Suspicions that OpenAI may be using YouTube to train Sora intensified after the interview Mira Murati With Wall Street Magazine. When asked if they used this platform to improve their new AI, he said he doesn’t know. Something quite curious, considering that she is the company’s chief technology officer.

Sam Altman’s team has avoided providing details about what the datasets used with Sora include. They only indicated that they use “publicly available information” and licensed content. Regarding the latter, the only thing that is known is that OpenAI has entered into an alliance with Shutterstockone of the world’s leading providers of stock photos and videos.

The mistrust of OpenAI in Sora’s development is not unfounded. Earlier this year the company said that it was impossible train new AI models without using copyrighted materials.

“Limiting training data to public domain books and drawings created more than a century ago might be an interesting experiment, but it would not allow for the creation of artificial intelligence systems that meet the needs of today’s citizens.”

OpenAI.

The Californian company has big plans for Sora. Recently OpenAI started doing lobby in Hollywood for use in films and TV series. Its use has even been revealed to some top level directors and artists so that they can try it out and see its potential.