Training the AI of the Future - Where B Roll Becomes Data: Teaching Old Footage New Tricks
We Got That B Roll!
In the early days of video, specialty TV, and budget advertising “B Roll” became a business. The idea was to sort and catalog generic shots that could be used for multiple purposes. Then it was simply a matter of clearing the rights to the footage and licencing various uses. Broadcast, non-broadcast, TV, documentary, feature film. You’ve seen more B Roll than you would imagine. The market for B Roll became very professional over the last 10 years and very high quality.
And now there’s a new customer in town. All the AI engines you hear about. They’ve devoured all the words and most of the pictures (legally or not). But if AI is to really fulfill its promise, it needs to make really good moving images. Photo real. Just from simple prompts.
What’s the goal?
The end goal of building AI and training it with visual images is much bigger than just putting sound stages, actors, Hollywood, and all those B Roll companies out of business — though that’s certainly the first thing that’s going to happen… I give it about 24 months.
The deeper goal is to teach machines how to "see," "interpret," and "create" the visual world the way humans do — but faster, cheaper, and at unimaginable scale.
Here's the broader picture:
Entertainment and media: Yes, AI could (and will) generate entire movies, TV shows, commercials, and social content without traditional sets, actors, or even crews. Think personalized storytelling: a Netflix movie starring you, dynamically created on demand. But that's the tip of the iceberg.
Training and simulation: Military, aviation, healthcare — industries are already pouring billions into AI-generated visual environments for ultra-realistic simulations. A battlefield, a surgery, a city evacuation — all simulated, trained, and rehearsed in photorealistic worlds that cost a fraction of live training.
Education and research: Imagine a textbook that visually adapts to a student's pace. Or historical documentaries where AI reconstructs extinct cities or ancient battles with breathtaking, evolving fidelity.
Retail and advertising: Hyper-targeted ads where products appear in your own living room through AI-augmented visuals. Virtual try-ons, virtual showrooms, virtual experiences that feel more real than the real thing.
Urban planning and architecture: Designing cities, roads, and buildings in AI-created simulations that model traffic, sunlight, population flow, or even social behavior — before a single shovel hits the ground.
Robotics and automation: For machines to navigate the real world — think delivery robots, autonomous cars, drones — they need visual training data that mimics the chaos of real life. Not the polished “final cut” version — the messy, boring, everyday raw footage. The more real-life nuance they see, the smarter (and safer) they get.
Surveillance and security: (The darker side.) AI is being trained to watch and recognize human behavior — identifying threats, anomalies, or specific individuals in real time. The ethics here are complicated, to say the least.
It's all about the quality of your data.
In this debut episode of The Search Party podcast, I had the chance to sit down with Chris Keevil, CEO of Versos, to talk about something new and quietly revolutionary: how old TV footage—raw, unpolished, story-rich visual content—is suddenly one of the most valuable commodities in the AI age.
We explore how my company, Arcadia Entertainment, became among the first professional TV studios in the world to successfully license our media archive for AI training. We dive into what AI companies are really looking for (hint: it's not polished shows, but authentic, real-world footage), why story-driven content is now premium data, and what this means for the future of creativity and ownership.
If you’re a regular reader of THE BEE and only imagined from my meandering writing how scattered I would be to talk to in real life, you can picture it here as poor Chris spends an hour trying to get to the point of this creative economy cavalcade.
Inside the episode, we cover:
How story—the real, messy kind—became treasure in the AI economy
Why unfinished footage is often more valuable than the final product
The technical and legal wild west of licensing footage for AI training
How AI, far from killing creativity, will unleash a new renaissance
Why optimism, imagination, and valuing your own IP are the real keys to success
The conversation ranges from shipwrecks and treasure hunting (of all sort) to carpentry and beauty as the ultimate and maybe only real long-term wealth creator and store in history. It's a deep and (I hope) inspiring look at how creators can survive and thrive in the new AI-powered world.