Google keeps Open AI 'Sora' in check… “You wish permission to make use of YouTube for learning”

-

(Photo = Sora)

Google is betting on OpenAI's video creation artificial intelligence (AI) model 'Sora'. Amid continued criticism that the source of Sora's learning data was unclear, it was checked that using YouTube videos for learning was a transparent violation of service.

Bloomberg reported on the 4th (local time) that YouTube CEO Neil Mohan emphasized in an interview that “using YouTube videos to coach AI models without permission is a violation of the Terms of Service.”

In accordance with this, CEO Mohan stated, “It’s unclear whether videos posted on YouTube were used during Sora's learning process, but in the event that they were, it might be a violation of YouTube's terms of service.”

He also said, “Creators have expectations when uploading their hard-earned work to YouTube,” and added, “Considered one of them is that the Terms of Service can be complied with.” He added, “Downloading videos without permission or using video and script data is prohibited in keeping with the terms and conditions.”

Google also stated that it complies with individual contracts with YouTube and creators before training Gemini. Google has the multi-modal model Gemini, in addition to image creation AI 'Imagen' and 'Lumière'.

Meanwhile, what data OpenAI uses to coach Sora is a matter of debate.

OpenAI CTO Mira Murati said in an interview with the Wall Street Journal in March, “It is just not clear whether Sora learned from user-generated videos on YouTube, Facebook, and Instagram.” She later modified her statement to say that she “learned only from affiliated images.”

The Information previously reported that OpenAI used YouTube data to coach 'Whisper', a model that converts voice to text. It’s reported that a few of Whisper's training data was also utilized in 'GPT-4'.

In this manner, YouTube content has change into a very powerful source for learning multimodal models. Amongst these, Google is actually sending a warning to Open AI and other corporations.

Open AI didn’t make an official statement regarding this.

Meanwhile, OpenAI is currently involved in about 10 copyright lawsuits. A representative example is the lawsuit against the Latest York Times.

We’re also discussing partnerships with major Hollywood studios to secure learning data for Sora.

Reporter Park Chan cpark@aitimes.com

ASK ANA

What are your thoughts on this topic?
Let us know in the comments below.

0 0 votes
Article Rating
guest
0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments

Share this article

Recent posts

0
Would love your thoughts, please comment.x
()
x