Revolutionizing YouTube: Startup Offers Creators a New Way to Profit from AI Training Data

To go back to this article, head to My Profile and then click on View saved stories.

A New Company Aims to Ensure YouTube Creators Are Compensated for AI Training Material Use

Until now, AI firms have utilized the rich resources found in YouTube's videos, captions, and additional materials for training purposes without obtaining consent. A startup specializing in AI content licensing, named Calliope Networks, is setting out to alter this practice through its innovative "License to Scrape" initiative, which targets prominent YouTube personalities.

"AI firms are clearly interested in harvesting content from YouTube, evident from their activities. Our goal is to facilitate this process legally and easily," states Dave Davis, CEO of Calliope Networks. In contrast with major social media networks such as Reddit, YouTube has not entered into agreements with leading AI companies for content scraping. The charm of the License to Scrape lies in its approach to bypass the need for the company to offer a bulk of YouTube content directly, instead assembling a collective of creators to secure a comprehensive license.

Davis, who previously worked in conventional media licensing at the Motion Picture Licensing Corporation, transitioned to start Calliope, convinced that the AI sector would shift from unauthorised data gathering to embracing licensing standards. This conviction isn't unique to him; there's a surge in startups focused on AI data licensing. As an initial member of the Datasets Providers Alliance, Calliope Networks advocates for a model where all content creators and rights owners must agree to data collection.

Davis envisions the following process: Content creators on YouTube interested in licensing their data would engage in an agreement with Calliope. Calliope would then grant sub-licenses to their content for the purpose of training base models of generative AI. To make this proposition appealing to companies working on AI, a substantial volume of content is necessary. Therefore, the initiative requires the participation of numerous YouTubers to kick off effectively. Calliope would earn a portion of the licensing revenue from the AI firms.

In the realm of artificial intelligence, nothing currently mirrors this concept, but Davis drew inspiration for the scraping license structure from elsewhere within the entertainment sector. This includes organizations such as Broadcast Music Inc. (BMI) and the American Society of Composers, Authors, and Publishers (ASCAP), which both employ comprehensive licenses for music.

Davis mentions that the recruitment phase is just beginning. He predicts that Calliope must present at least 25,000 to 50,000 hours of YouTube content to gain credibility within the AI sector. The requirement for such a significant amount of material to secure comprehensive licenses illustrates the importance of collaboration amongst creators to profit from AI training. In this industry, the quantity of data is crucial, as video generators rely heavily on extensive datasets.

No high-profile endorsements have been secured for the license so far, however, Calliope has engaged several influencer marketing firms such as Viral Nation to attract clientele. "The response from content creators has been very positive," Bianca Serafini, who leads content licensing at Viral Nation, commented. She believes that a significant portion of their nearly 900 YouTube clients will join. "This is a unique proposition that hasn’t been offered to us previously."

Regarding this matter, what is YouTube's stance? Although Davis hasn't collaborated with the platform specifically for this initiative, he is of the opinion that it aligns with the desires of the video giant. "From my perspective, YouTube is interested in empowering creators with greater autonomy," Davis remarks.

YouTube refrains from discussing individual licensing entities, yet it clearly backs its creators in forging personal deals. "In broad terms, creators are free to negotiate agreements with external organizations concerning their content on our platform," states Jack Malon, a representative for YouTube. Malon highlights a recent blog post by the company that underscores its commitment to granting YouTubers "enhanced autonomy" in the era of AI technology. For YouTube, the key issue is obtaining proper authorization: "Accessing creator content without permission contravenes YouTube’s Terms of Service, and we are dedicated to implementing safeguards to ensure that third parties adhere to these rules."

The success of the License to Scrape initiative hinges on more than simply attracting prominent YouTubers. It necessitates a significant alteration in the strategy AI firms use for basic training. Given that over 30 copyright disputes related to unauthorized data scraping are currently proceeding through US courts, such a change might become a legal requirement. Nonetheless, since tools that generate text-to-video content typically require vast quantities of superior data to function effectively, the search for additional data sources might demand a new strategy.

Up to this point, it remains uncertain whether the leading figures in artificial intelligence intend to halt their practice of harvesting what they refer to as "publicly accessible" data from platforms like YouTube. (When agreements are made that encompass the training of base models, such as the arrangement between the video-centric AI startup Runway and film company Lionsgate, the data in question usually isn't "publicly accessible.") The majority of their agreements with various platforms and publishers are aimed at acquiring content for AI search tools like SearchGPT, rather than for the training of foundational models. OpenAI, after being threatened with legal action by the well-known UK parenting forum Mumsnet, informed WIRED that its main interest lies in securing licensing for large datasets that are not publicly accessible.

Meanwhile, advocates for the initiative argue that it's crucial to take proactive steps now instead of waiting for AI firms to show their willingness to participate. “We need to be one step ahead,” Serafini emphasizes.

Discover More…

Direct to your email: A selection of WIRED’s most fascinating and unusual tales

Elon Musk poses a threat to national security

Interview: Meredith Whittaker Aims to Disprove Capitalism

What's the solution for a challenge such as Polestar?

Occasion: Be part of The Significant Conversation happening on December 3 in San Francisco

Additional Content from WIRED

Evaluations and Tutorials

© 2024 Condé Nast. All rights reserved. Purchases made through our website may result in WIRED receiving a commission as a result of our affiliate agreements with retailers. The content on this site is protected and may not be copied, shared, transmitted, or used in any form without explicit written consent from Condé Nast. Ad Choices

Choose a global website