YouTube creators get new option to allow third-party AI firms to train models on their videos


YouTube announced a new update on Monday that will allow content creators on the platform more control over third-party artificial intelligence (AI) training. The move comes after the video-streaming giant introduced new tools to protect creators from deepfakes that mimic their likeness, including their face and voice. The new option will let content creators decide whether they want third-party AI firms to have access to their videos to train large language models (LLMs). They can also allow specific AI companies, while blocking others from using their videos.

YouTube lets creators decide which AI firms can train models using their videos

Companies are now racing for new data sources to train AI models while developing LLMs. Now that publicly available data has been exhausted by these AI firms, they are looking at new ways to find large deposits of high-quality data to train models and make them more efficient.

While some AI companies have taken the content-sharing route, it is generally considered expensive to source such data. Another option is synthetic data, created by other generative AI models. However, there is a risk that such data may be low-quality, which could negatively impact the development of new models.

As such, companies are trying to collaborate with content creators to find new high-quality data to train AI models. For example, Grok is currently trained on public posts on X (formerly known as Twitter), and Meta AI is trained on public posts on Facebook and Instagram.

Given the large amount of human-generated data, YouTube has also become a platform of interest for AI firms. As video generation models advance, this data becomes even more valuable. However, till now the video-streaming giant has not allowed companies to crawl and scrape videos in an unauthorized manner to protect the creators.

In a support document, the company announced a new option that will allow content creators on the platform to choose whether they want to let an AI firm access their videos to train LLMs. In the next few days, YouTube plans to release an update that will add a new option to Studio settings under the “Third-party training” section.

There, creators can choose to allow specific AI companies to scrape their videos. The list of companies currently includes AI21 Labs, Adobe, Amazon, Anthropic, Apple, ByteDance, Cohere, IBM, Meta, Microsoft, Nvidia, OpenAI, Perplexity, Pika Labs, Runway, Stability AI, and xAI. Notably, creators can make their videos accessible to all AI companies by choosing the respective options.

YouTube highlights that only those videos will be eligible for AI training that have been given permission by the creators as well as the applicable rights holders. Additionally, the company’s terms of service still apply, meaning AI companies cannot illegally remove videos from the platform.

This new option does not include any mention of compensation from AI firms to creators for using their videos. However, YouTube highlighted that it will continue to facilitate new forms of collaboration between creators and third-party companies.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *