Don’t Sleep On Your B-Roll, AI Giants Are Buyers (Guest Column)

by

What sets apart a high-quality generative AI product? How does one such AI solution stand out from its competitors in an increasingly crowded market? As more AI solutions emerge and competition intensifies, understanding what makes a good product is crucial. The quality of the content employed to train the AI models behind these generative AI products plays a significant role in answering this question.

To ensure an AI model delivers expected results, it needs to be initially educated on extensive, trustworthy, and precise data sets. Often, these models are schooled using web-accessible content that might not have been verified. Such a practice can pose issues during product creation. For instance, if numerous AI systems learn from the same easily obtainable resources, their algorithms may start making similar predictions or producing alike results, reducing the uniqueness and value of both the AI models and the outputs they create for users.

Furthermore, employing content without any kind of assessment, such as verifying its origin or authenticity, can introduce risks from multiple perspectives. For example, might using specific content in AI training infringe upon the intellectual property rights of others? These intricate intellectual property matters are currently under debate in courts. What about potential biases or inaccuracies within the content? How is the risk associated with these issues managed?

This development paves the way for a burgeoning licensing market in AI, allowing developers to acquire licensed material for their model’s education from reliable resources. In the bustling AI product landscape, developers are not only focusing on minimizing risks during training but also on setting their AI models apart from competitors. By educating their AI models with unique content and data, they strive to develop algorithms that generate distinctive and thus more valuable results.

Simultaneously, content developers and rights holders are exploring methods to generate income from their content within the AI marketplace, while also preventing unauthorized use of their intellectual property for training AI models, unless they explicitly give consent.

High-quality content is often in short supply, and the realm of AI training isn’t an exception. An underutilized source of quality content that can benefit both AI creators and content providers or owners is B-roll footage and archived content. This kind of footage offers a new avenue for monetization for content owners who previously lacked such opportunities, while for AI developers, it presents a vast collection of potentially unique training material with reduced risk.

Creators who share videos on platforms such as YouTube, Instagram, and TikTok are already tapping into a profitable market by licensing their unused video content to artificial intelligence (AI) companies. As reported by Bloomberg News, these transactions can fetch thousands of dollars per deal, with AI developers typically paying between $1 and $4 for every minute of footage. The worth of this footage is influenced by several elements, including the video’s quality, its uniqueness, and how scarcely it’s accessible to other AI developers.

Under this licensing setup, both the content creator and the AI developer stand to gain advantages. On one side, content creators can now generate income from previously unmonetized or underused assets while preserving ownership and control over their content via strategic licensing. This means they can establish contractual terms that restrict the application of their content to certain scenarios and safeguard against misuse or inappropriate reproduction of premium content. For instance, creators might aim to limit the employment of their content within AI systems used for purposes potentially harmful to their own industry or benefiting competitors.

From another perspective, this licensed content presents four significant benefits for AI creators. Initially, it offers distinctiveness from rivals due to exclusive, confidential training material that isn’t accessible to the general public. This can grant an edge in a market where numerous AI models have been trained on identical data. Consequently, AI developers striving to develop the next generation AI model may find this advantageous.

As a gamer, I prioritize ensuring high-quality gaming experiences by using authentic player-created content. This helps me steer clear of the increasing issue of artificial intelligence-generated content pollution. It’s all about preventing AI models from learning and replicating content that was initially generated by AI, which can degrade the quality and creativity in the results.

The third strategy involves minimizing potential security threats by using data with verifiable origins, which includes a comprehensive history of its creation, modifications, and the people who handled it during that process. This is crucial for avoiding contaminated data from corrupting their AI systems during training.

The advantage of the fourth point is that it offers more assurance about intellectual property rights and potential risks from external sources when it comes to content used for training. However, this perk of exclusive content does not come for free; it results in higher expenses during development and training phases.

Lately, DeepSeek – a Chinese AI model akin to ChatGPT – has been gaining attention as one of the most downloaded apps globally. While it resembles other AI models available, DeepSeek stands out due to its reduced development and training expenses. There’s talk that DeepSeek might have been trained using distillation, a common method where data is drawn from larger, more capable AI models, enabling cost-effective training. However, the use of distillation underscores the importance of seeking unique content and data for training purposes, allowing an AI model to offer distinct outputs that set it apart from other market options by generating results based on exclusive data.

The emerging market for B-roll licensing presents a special platform where AI developers and content providers can forge mutually beneficial partnerships. To capitalize on this chance, creators and owners should devise methods for assessing and selecting fitting resources, set up licensing terms for safeguarding high-value content, obtain necessary clearances for rights, and construct pricing models considering aspects such as exclusivity, quality, and type of content.

Simultaneously, AI creators ought to prioritize setting up reliable processes for verifying and evaluating content, addressing potential copy issues in outputs, and designing flexible data systems to manage numerous licensing agreements and protect content. If both parties implement these strategies, it could foster a prosperous environment where content creators and AI developers can flourish together.

Adrian Perry and Robyn Polashuk are teammates at the law firm, Covington & Burling, where they both serve as leaders for their Entertainment and Media department.

Read More

2025-02-13 19:25