YouTube Raises Concerns Over OpenAI's Use of Videos for Training Sora

In a recent statement, YouTube CEO Neal Mohan expressed concerns about OpenAI potentially using YouTube videos to train its text-to-video generator, Sora. Mohan stated that such usage would be a violation of YouTube's terms of service.

While Mohan admitted to not having direct knowledge of OpenAI's practices, he emphasised that if the company had indeed utilised YouTube videos to refine Sora, it would be a clear violation of the platform's terms of use. He highlighted the importance of creators' expectations and the need to abide by the rules set forth in the terms of service.

YouTube's terms of service explicitly prohibit the downloading of transcripts or video clips, making it incompatible with the use of YouTube content for training AI models. Mohan emphasised that these rules are in place to ensure the integrity of the platform and protect creators' work.

The debate surrounding OpenAI's training methods has been a topic of public discussion. OpenAI's generative AI tools, including Sora, rely on gathering data from various sources on the internet to generate new content, such as videos, photos, and narrative text. As companies like OpenAI and Google strive to enhance their AI capabilities, they seek to gather as much content as possible to train their models effectively.

OpenAI, supported by Microsoft, has not yet responded to YouTube's concerns. In a previous interview with the Wall Street Journal, OpenAI's Chief Technology Officer, Mira Murati, expressed uncertainty about whether Sora was trained using user-generated videos from platforms like YouTube, Facebook, and Instagram. However, recent reports suggest that OpenAI has discussed training its next-generation language model, GPT-5, using transcriptions of public YouTube videos.

Mohan emphasised that Google and YouTube prioritise individual contracts with creators when considering the use of videos from the platform for training their own AI model, Gemini. He acknowledged that creators have different licensing agreements for their content on YouTube. While some portion of the YouTube corpus may be used for training models like Gemini, Google and YouTube ensure that this usage aligns with the terms of service and contracts signed by creators.

As the debate continues, it remains to be seen how OpenAI will address YouTube's concerns and whether any changes will be made to the training methods of Sora and other AI tools.

Source: YAHOO