Embeddings (and how to create them) are, perhaps, the most interesting idea behind Large Language Models. I built a simple model to help you understand embeddings from scratch. Here is a step-by-step video explanation:

Here is the same video, but now on YouTube: And here is the code I used in the video:

This is the biggest productivity cheat code right now. Kiss reading documents goodbye. You can get an instant summary of any document with this tool.

@PLynchado

Impressive work Santiago. Thanks for the detailed tutorial!

How do embeddings improve AI’s reasoning? Can this method work beyond text, like for images/videos?

Yes, this works for images/videos. In fact, in this example, I create embeddings for images.

Do embeddings matter as much anymore? We’ve got smolagents, better instruct models, and bigger contexts that allow for better hierarchical search and broader contextual understanding? Embedding search is cheaper, sure, but if the docs change over time don’t the embeddings “fail”

How do you see embeddings improving the efficiency of your model? They can really help in capturing semantic relationships, which is crucial for better understanding context in AI.

@memdotai meme this

Great educational content on embeddings! These fundamental concepts are crucial for understanding how modern AI systems work. Looking forward to the video explanation.