Intro to Embeddings

The Grocery Store Map

Imagine a grocery store.

• Apples are right next to Oranges.
• Milk is far away from Apples.
• Shampoo is even further away.

You know this because you understand categories. Computers don't. To a computer, the word "Apple" is just 5 letters: A-P-P-L-E. It doesn't know it's a fruit.

Enter Embeddings

Embeddings are a way to turn words into numbers (coordinates on a map).
Instead of "Apple", the computer sees: [0.9, 0.1, 0.5]
Instead of "Orange", it sees: [0.8, 0.1, 0.5]

Notice how the numbers are similar? That tells the computer, "Hey, these two things are related!"

The Magic of Math

Because words are now numbers, we can do math with them.
The most famous example is:
King - Man + Woman = Queen

If you take the numbers for "King", subtract the numbers for "Man", and add the numbers for "Woman", the result is almost exactly the numbers for "Queen".
This proves the AI understands the concept of gender and royalty, not just the words.

Why It Matters

Embeddings are the secret sauce behind:

• Search: Finding "running shoes" when you search for "sneakers".
• Recommendations: "You liked this movie, so you'll like that one."
• Chatbots: Understanding what you mean, even if you use slang.