What is generative AI, really?

Generative AI is software that learns patterns from large collections of examples, then produces new text, images, audio, or code that resembles those patterns.

A useful mental model

Imagine an extremely advanced autocomplete system. It receives your input, breaks it into smaller units, and predicts what should come next. The prediction is informed by patterns learned during training—not by human-like understanding or lived experience.

Four terms worth knowing

• Model: the learned system that makes predictions.
• Training: the process of adjusting the model using many examples.
• Prompt: the instruction or input you give the model.
• Token: a chunk of text the model reads and generates.

Why answers can sound certain and still be wrong

The model is optimized to produce plausible continuations. It may invent a source, mix up facts, or miss context. This is often called a hallucination. Fluency is not proof of accuracy.

Beginner rule: Use AI to draft, explore, and transform. Verify it before using it to decide, publish, pay, diagnose, or advise.

Your first safe experiment

Choose a non-sensitive paragraph you wrote. Ask the model to produce three clearer versions for different audiences. Compare the changes and keep only what genuinely improves your meaning.