I’ve recently been giving several talks on AI to various business groups. I’m not an expert on AI. They don’t exist as the space is moving too quickly. What I do is help business apply AI to their problems and I’m seeing major problems.

I keep returning to Yann LeCun’s post on AI and its implications for the pace of AI development. For those of us testing AI, every day is a Cambrian explosion of capabilities, but for my students and most business clients it seems a still distant technology with low day to day business adoption.

Some of that adoption friction is about fear, relevant security/trust concerns, user habits, and IT departments that are reluctant to spread this unknown power within the permitter of the organization.

What if your text documents are relatively useless in this new AI age? Let assume that per Yann we need Video and as they New York Times reports lots of it. Most business I interact with don’t have any reasonable scale of video to train an LLM. Maybe there are a few internal training videos and a random collection of post pandemic Zoom and Teams meetings. Nothing that is going to give them a significant leg up over YouTube/Google, Meta, LinkedIn, and others who all have vast troves of our uploaded video. Indexed, organized, and ready to go.

The blocker that I encounter is often IT and the simple storage and editing of video at scale. It needs to be as easy to use as YouTube and as useful as Descript. IT wants us to use Sharepoint and other tools that are poor substitutes for YouTube and a video editor as easy to use as a word processor.

Assume video is the new data currency in an AI era. The majority of my students and talk attendees have their camera off during a call or class unless prodded. Does that leave Mr Beast and YouTube extroverts to shape future LLM’s for your business? How do we begin to take back that AI data gold for our organizations and/or personal use?

I don’t have an answer, yet. But my students and clients need one.