The Three Layers of Generative AI

Published by


When most people think about generative AI, they think of ChatGPT…

After all, this is what put artificial intelligence and machine learning in the spotlight back in January.

David Ryder | Bloomberg | Getty Images

According to Amazon CEO Andy Jassy though, there is a better way of thinking about it…

He believes that there are multiple different layers when it comes to AI that should be considered:

  1. Bottom Layer
  2. Middle Layer
  3. Top Layer

The bottom layer is the “technical” layer for AI. It is the raw computational power that allows for everything else to take place above it.

This is the layer in which a Nvidia operates in…

They provide the tech that other companies can use to build everything else on top of it.

This brings us to the middle layer…

This is where the Large Language Models are!

We are talking about GPT-4 from Open AI or PaLM-2 from Google.

Let’s use computers as an analogy to help make better sense of this…

A Dell desktop computer would be akin to the bottom layer – the physical hardware upon which the rest is built.

The middle layer would be the operating system (Windows).

Without the physical computer, Windows would have nothing to run on…

Likewise, if we didn’t have Windows we wouldn’t be able to run Google Chrome or any other application on a computer. 

This brings us to the top layer!

These are the specific applications that are built using the Large Language Models.

The most famous example of this is Chat-GPT…

Thomas Trutschel/Photothek via Getty Images

It uses the GPT language model to allow for a Q/A style sophisticated chatbot functionality. 

Of course, as generative AI development continues to progress, we will eventually have a nearly endless amount of such applications to choose from!

That’s all for now…

Until Next Time,


%d bloggers like this: