Day 25

Today, I understood about Embeddings.

At first, I thought they were just another technical AI word.

But here is how I understand them now:

An embedding is a numerical representation of text.

It helps an AI system compare pieces of text based on how closely related they are not only whether they use the exact same words.

For example:

How do I get back into my account?

and

Steps to reset your password

use different words, but they are about a similar problem.

Embedding-based retrieval can help an AI system find that connection.

That is one reason RAG can retrieve relevant information before an LLM responds.

My takeaway:

- Tokens are the pieces of text an LLM processes.
- Embeddings help AI systems represent and compare information by meaning.

Day 25 of 1% better.