VideoDB Documentation

VideoDB Documentation

Building World's First Video Database

Misalignment of Today's Web

Ashutosh Trivedi

The internet as we know it today might not be around tomorrow. What we call “The Internet” today is primarily a vast repository of websites or URLs containing various forms of information, including

📄 Text based content: Blogs, Stories, News, Articles,

🎑 Images: People, Places and Things and

⏯️ Video Content (over 80% of all internet traffic)

Educational videos (e.g., tutorials, lectures, documentaries)

Entertainment videos (e.g., movies, TV shows, music videos)

User-generated videos (e.g., vlogs, product reviews, gameplay footage)

News and event coverage (e.g., live streams, breaking news reports)

Instructional videos (e.g., cooking demonstrations, DIY projects)

Professional videos (e.g., corporate videos, advertisements, promotional content)

Video is the most data-intensive medium. People spend several hours watching videos online compared to reading text or viewing images.

🚨 The Pitfalls of Attention Economies:

Platforms like Facebook, Instagram, and YouTube have evolved into mature ecosystems that govern how content is created, shared, consumed, and monetized. The majority of these platforms operate on attention economies, where the primary objective is to offer users an abundance of choices, encouraging them to spend more time on the platform. This approach aims to drive consumer behavior, enabling monetization through an advertising-driven economy and elevating the influence of content creators, often referred to as "influencers."

However, the strategy of presenting users with an overwhelming number of choices is a hack that can trap individuals into dopamine loops, leading them down rabbit holes and reinforcing biases. Platforms like YouTube and Facebook have been criticized for forcing choices upon users, inadvertently perpetuating these detrimental patterns. There’s now an increased awareness about these unfair practices.

Furthermore, these platforms have consistently demonstrated an inability or unwillingness to effectively tackle several critical issues:

Misinformation and Fake News: The spread of false or misleading information on these platforms can have severe societal consequences, yet effective content moderation remains a significant challenge.

Algorithmic Bias: The algorithms employed by these platforms to curate and recommend content can perpetuate existing biases and prejudices, creating echo chambers and reinforcing harmful narratives.

Fair Creator Monetization: Ensuring fair compensation for content creators has been an ongoing struggle, as platforms grapple with maintaining sustainable business models while adequately rewarding those who contribute valuable content.

Centralization and Market Dominance: The concentration of power and influence among a few major platforms raises concerns about monopolistic practices, stifling competition and limiting consumer choice.

These platforms, while offering convenience and connectivity, have faced criticism for prioritizing engagement and profit over ethical considerations and user well-being. As the digital landscape continues to evolve, addressing these issues will be crucial in fostering a healthier and more responsible online ecosystem.

📈 Indexing and Ranking

A critical aspect of the internet's functionality is indexing, which involves understanding and modeling information to facilitate retrieval. Companies like Google have monetized it through ranking, shaping how information is accessed and consumed.

However, last year we have witnessed a significant shift in user behavior, with people spending less time on traditional search engines and gravitating more towards conversational agents or AI assistants like ChatGPT, Perplexity, Claude, and Gemini. These agents, powered by foundation models (large language models trained on vast amounts of data), offer a more interactive and personalized way of accessing and processing information.

They also have power of personalization and power to rank content according to user’s needs and giving it back to the end consumer instead of manipulated feeds and choices.

🔥 The Disruption of the Internet Landscape

As we move forward, the internet as we know it is set for a substantial disruption driven by several forces.

New Behaviour: User-Facing Agents

The emerging paradigm revolves around user-facing agents that can consume, process and take actions without the need for users to visit individual URLs. A Post URL internet has arrived with agents. These agents can leverage their own ranking models and personalization algorithms to deliver tailored experiences to users. By directly accessing and streaming content, agents have the potential to disrupt the traditional ranking economy model that has long underpinned the internet.

GenAI Content: Misalignment of Incentives

Furthermore, the rise of generative AI (GenAI) foundation models has sparked concerns among creators regarding the potential misuse of their content. Recent lawsuits against companies like OpenAI (by the New York Times) and Stability (by Getty Images) highlight the growing tensions between creators and GenAI models trained on publicly available data. Creators fear that sharing their content publicly could empower these models to generate similar content, rendering the original creators irrelevant and undermining their livelihoods.

User Fatigue: Transition from algorithmically curated choices

One of the most significant factors in user behaviour is the transition from an abundance of choices to a focus on quality. The current landscape, where platforms like YouTube and Facebook employ algorithms to present users with an endless stream of content is set to create a negative value of confirming biases and misinformation.

When these forces play together, they create a dangerous cocktail that has potential to destroy what we call today “The internet” but also create another valuable ecosystem.

🤗 VideoDB’s Vision for Tomorrow: Bridging the Gap

In this evolving landscape, VideoDB aims to position itself as the internet's infrastructure for the GenAI era, specifically for video content. The platform offers a revolutionary approach by enabling creators to upload and index their video content, making it accessible to agents like ChatGPT, Perplexity, or Arc and many more to come.

Unlike traditional platforms, VideoDB does not rely on the concept of files; instead, it employs indexes to find and stream content directly to end consumers using advanced retrieval and generation (RAG) techniques. This approach eliminates the need for users to visit individual URLs, as agents can access and rank and personalize the content seamlessly.

A preview of this model can be seen in our open-source repository,

StreamRAG⁠

, where it is possible to interact with your videos through ChatGPT and publish a GPT of your video content, either publicly or privately.

We want to align of incentives between creators, agents, and end-users. Agents and users can leverage their own ranking models to consume and personalize content directly from VideoDB's streams. The value generated by this process is then divided among the agent, VideoDB, and the creator, ensuring fair compensation for all parties involved.

|Consumer| < — > | ChatGPT| < — > |VideoDB|< — > |Creator|

Creators can be compensated based on the data streams consumed, regardless of the video's duration. This approach incentivizes creators to contribute high-quality content, fostering an ecosystem that promotes creativity and value-added experiences.

Providing a secure and equitable platform for creators, agents, and end-users has the potential to shape the future of the internet, fostering innovation, creativity, and a more harmonious ecosystem for all stakeholders.

We invite product teams like Browser Company, ChatGPT and Perplexity to have a healthy conversation on this model 🙏🏼

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.