Making the First Guess

Starting from Scratch

Imagine you're taking an exam in a language you've never seen before. With no dictionary or guide, your very first answer would be a total, blind guess. Every AI model begins its journey in this exact state—a completely blank slate.

The Internal Dials

Inside the model, information flows through a network of nodes. Each connection has an adjustable dial called a 'weight.' Before training begins, engineers set every one of these dials to a purely random number. This gives the model a neutral starting point to find patterns.

Your First Prediction

Let's look at a model designed to predict house prices. This house is 2,000 square feet and is actually worth $500,000. Spin the dial to see what the model's random starting weight produces as its first guess. Because the dial was set randomly, the model calculates a guess of just $10,000. It's wildly incorrect, but this 'Guess' is the essential starting gun for all learning.

The Starting Gun

It's easy to assume AI starts with common sense, like knowing a house can't cost ten thousand dollars. In reality, it has no intuition. A wildly inaccurate first guess isn't a failure— it's the baseline the model needs to measure its error and begin to improve.

Lesson Recap

Let's review. AI doesn't start smart; it starts random. These random weights create the first guess, which serves as the foundation for the entire training cycle.