I Tried to Explain AI. I Got It Wrong. So I Learned How It Actually Works.

What actually happens inside AI?

After asking a podcast guest to explain AI—and then realizing my own explanation wasn't quite right—I went back to the basics.

In this short episode, I break down AI in plain English: training data, data preparation, model weights, prediction, error, and how a trained model generates an answer from a prompt.

I also look at where concepts like ontologies, relationships, probabilistic outputs, and modern AI search fit—and where they don't.

No Stanford degree required. I do, however, own the shirt.

• Why saying “AI is data” doesn't tell the whole story

• How training data is cleaned and prepared

• What model weights actually are

• Prediction → error → weight adjustment → repeat

• How models learn statistical patterns at scale

• Training versus inference

• What an ontology actually describes

• Why LLMs are probabilistic

• How AI search differs from traditional search

• Why modern systems can understand much longer, messier questions

Jason T Wade is the founder of BackTier and NinjaAI and host of the AI Visibility Podcast. His work focuses on AI Visibility—how AI systems discover, understand, cite, include, and recommend entities.

BackTier — AI Visibility strategy and systems
NinjaAI — AI SEO, GEO, and AEO
OpenAI — AI research and models
Stanford HAI — Stanford Institute for Human-Centered Artificial Intelligence

In this episodeAbout Jason T WadeRelevant Links