1. Foundations6 lessons
  2. Inside the Model5 lessons
  3. How Learning Works5 lessons

About the project

Why KI einfach verstehen exists

You type a question, and a few seconds later there is an answer that sounds as if someone had thought it through carefully. A recipe, a cover letter, a piece of code. Handy. And a little bit eerie when you have no idea what is actually going on.

That is exactly how this project started: with a pretty simple question. How does all of this really work?

Using is not understanding

Like many others, we started out by simply using AI. For writing, for programming, for research, to automate things. A lot of it works surprisingly well, even without knowing what runs in the background. At some point, that was no longer enough for us.

What happens inside a language model between question and answer? What are tokens, what are parameters? How does a model learn anything at all? And why can the same AI write a decent essay but miscount the letters in a single word?

Too shallow or too deep

So we started looking. There is more than enough information about AI out there. A good way in was still surprisingly hard to find.

Many explanations are like a paddling pool: you are standing in it, but you hardly get wet. “The AI learned from data”, done. Others are like jumping off the ten-metre board: two paragraphs in, the formulas arrive, and anyone who could not already swim goes under. What we missed was the pool in between. Shallow enough to get in, and deep enough that in the end you understand whysomething works the way it does.

Three pools side by side: on the left a paddling pool with a bored child standing in it, on the right a diving tower above deep water with a hesitant child on top, in the middle a large pool with wide steps that a child happily walks down into the water.
Too shallow, too deep, or step by step: here, you take the stairs in the middle.

That is how KI einfach verstehen came about: the thing we had been missing ourselves.

What you will find here

You do not need a computer science degree. It starts with the basics and goes deeper step by step, from a simple spam filter all the way into a language model.

The project is never finished. Every answer raises new questions, and those are added bit by bit. You can help: the website is developed in the open as an open-source project. If you spot a mistake, an explanation does not click or a topic is missing, you can report it there or contribute a correction or addition yourself:github.com/silvio-l/ki-einfach-verstehen-website.

The first step is smaller than you think:How does a spam filter know what junk mail is?

Everything free — how you can help the project

Every lesson is freely accessible, without a paywall and without signing up, and it should stay that way. If the project is worth something to you, there are three small ways to move it forward:

  • Merch & Support — a share of every purchase goes straight into research, new lessons and graphics.
  • A star on GitHub — costs nothing and helps more people find the project.
  • Sharing — with someone who also wants to finally understand AI.