I see this conspiracy topic come up whenever AI is discussed and I feel the need to address something here. No, the page scraping that you are describing (Google), is not how models are trained. Data is obtained from many sources, including but not limited to books, transcripts, physical scans, etc. Hence, the legal battles with IP content being represented in data output.

As someone who has been researching psychology and building "AI" tools long before OpenAI, Grok, or any of the others showed up, I can tell you with confidence that what most people call “artificial intelligence” isn’t all that mystical. Under the hood, it’s just math, layers of statistics math in the realm of multiple regression, coefficient juggling, and weight tuning dressed up with clever marketing. Gonna go off on a huge ADHD tangent, but bear with me here.

When we talk about “intelligence,” what’s actually happening is a system crunching absurd amounts of numbers to find patterns, kind of like guessing which note comes next in a melody after hearing millions of songs. It’s probability, not consciousness. The magic is in how the math layers stack, each layer refines the noise from the one before it until something coherent starts to form.

Underneath, as stated before, it’s a storm of stats maths, linear algebra and calculus. Matrices, vectors, and derivatives are constantly interacting, gradients show the model which direction to “move” to get better answers, and loss functions tell it how wrong it currently is. Every neuron in a neural net has a weight, and those weights adjust through a process called backpropagation, which is basically feedback on steroids.

Training a model is just repetition. The model runs through billions of examples, makes predictions, gets corrected, and shifts its internal math ever so slightly each time. Do that enough times across enough data, and suddenly it starts writing poetry, generating code, or pretending it understands your feelings, but really, it’s still just math.

Once you understand the basics, everything clicks. Just like if you learn a programming language such as C, you fundamentally understand all of the other languages inherently from a foundational perspective. Or if you learn basic algebra and actually retain the concepts, the maths that follow become more understandable.

At the end of the day, the idea of AI isn’t new, it actually traces all the way back to ancient Greece. What’s changed isn’t the concept, but the capability. As technology caught up with imagination, what was once myth and theory turned into a practical tool: Sharper, faster, and finally able to keep up with the ideas people had centuries ago.