Generative AI Concepts — Visual Explainer

If you’ve ever wondered what people actually mean when they say “training”, “fine-tuning”, or “inference” — this interactive visualizer walks through each concept with real-time demos.

VISUAL EXPLAINER

Generative AI

⬡ Training◈ Latent Space⚄ Sampling⟁ Fine-tuning▷ Inference

⬡

LEARNING FROM EVERYTHING

Training

The model reads billions of text samples and adjusts billions of internal numbers until it gets good at predicting what comes next.

HOW IT WORKS

①Feed text into the model

②Model predicts next token

③Compare to actual next token

④Compute error (loss)

⑤Adjust weights via backprop

⑥Repeat — trillions of times

LOSS CURVEloss = 2.800

forward

loss

backward

update

QUICK FACTS

Training data

~1 trillion tokens

Parameters

billions → trillions

Training time

weeks on 1000s of GPUs

Core task

predict next token

ANALOGY

"Like a student reading every book ever written — then taking a billion practice tests."

← prev

next →

---

What You’ll Find

This isn’t a wall of jargon. Each of the 5 concepts comes with:

• A one-line explanation — what it is in plain English
• An analogy — connecting it to something you already understand
• Quick facts — the numbers that matter
• A live visualization — animated loss curves, vector spaces, temperature sliders, pipeline flows

The 5 Concepts

Concept	The Gist
Training	Feeding the model trillions of tokens until it learns to predict
Latent Space	A coordinate system where every concept has a location
Sampling	Adding controlled randomness so outputs aren’t boring
Fine-tuning	Taking a general model and specializing it
Inference	What happens when you actually ask the model something