It is clear that ChatGPT, specifically the LLM (Large Language Model) technology underneath it, will be a transformative force in the future.
Just like the internet and the mobile phone, LLM technology will change how we live, how we work and how we relate to each other. Thus it is important for everyone to understand how it works at the appropriate level of detail.
Unfortunately currently there is very little knowledge available about how ChatGPT and LLMs work.
Most of the content online is about (unsubstantiated) claims about what it “could” do. Or detailed research papers for experts.
There is nothing I could find that explains how ChatGPT works in a way that everyone can understand whether they are a data scientist, a technical person or a non-technical person.
Levels
My goal with this series is to fill this need by providing explanations of the key parts of ChatGPT in three levels:
- For the average non-technical person
- For the technical person
- For the data scientist
You can choose your level and click the appropriate links below. Of course, feel free to start with simpler levels or try out more advanced levels.
Parts
I’m going to explain ChatGPT and LLMs in the following parts:
- All ChatGPT does is complete the sentence
- Many tasks are simpler than we thought
- Creating a map of words
- The concept of distance in a map
- Embeddings and the coordinate system of the map
- Transformers give LLMs memory
- How LLMs construct the map of words
- Feedback makes LLMs smarter
- Putting it all together
At the end of this journey, you should be able to understand, at the appropriate level, how LLM technology works so you can smarter decisions about how to leverage it in your life, business and relationships.