In this era, artificial intelligence is the sole focus of all tech giants. Even though AI is clearly displacing creators’ jobs, everyone is eager to develop their own large language model to compete with each other.
From generating text-based responses to creating entirely new graphics from simple text prompts, AI plays a key role. And now, AI is ready to create complete cinematic videos from just a few text-based prompts. So, the question arises: how do they train their AI models, and where do they get their data?
Recently, Neal Mohan, CEO of YouTube, publicly stated, or we can say directly warned OpenAI, a leader in artificial intelligence. He said that although there is no direct evidence that OpenAI uses YouTube videos to train Sora – a text-based AI video-generating model, he warned that this behavior violates the terms of service of the YouTube platform.
He believes that from a creator’s perspective when creators upload their original work to YouTube, they have certain expectations. One of those is to comply with YouTube’s terms of service, which does not allow for the downloading of content such as transcripts or video clips, a clear violation of YouTube’s terms of service. These are the rules by which content on the YouTube platform is governed.
Mira Murati, OpenAI’s chief technology officer, was unable to specify the source of Sora’s training data. In contrast, Neal Mohan clarified that Google used some content from YouTube when training its Gemini models, but they had authorization from the creators before using the content and followed the individual contracts between YouTube and the creators.
