The GenAI Guide#

GenAI capabilities used to be limited to researchers or folks in a few organizations. Not anymore. These capabilities are widely available, whether it be popular chat style interfaces, through apis, or embedded in other folks. This is a guide for those looking to navigate this rapidly changing field.

You may ask Isn’t there already a lot of material out there? You’re right, and that’s precisely why this is a guide book. So much is being shared across papers Youtube, blog posts, tweets, GitHub etc, keep tracking of it all without getting overwhelmed is its own challenge.

Here’s what’s inside

  1. A guided tour of the fundamentals

  2. The minimal resources needed to get a great understanding

  3. Deep Dives into particular libraries, topics and papers

Who’s this for?#

  • Applied Practitioners - Developers looking to utilize these models in some way and are want to understand how to shape them and include them in their use cases.

  • Curious Users - Folks being exposed to GenAI (which is everyone these days) who want to learn more

  • Thought Leaders - Those figuring out where we are so to figure out where to go next

See also

If you enjoy this you may also like my GenAI Book Club where every couple of weeks we spend an hour talking through a different topic.

The contents#

Here’s what to expect as you click into the various chapters

  • End to End Span - From the mathematical fundamentals to how GenAI fits in the real world

  • Intuitive Explanations - A distillation of each topic into a couple of paragraphs

  • Code Tutorials - Learn through hands on code tutorials

  • Production Grade Code - Become proficient with the same tools the professionals use

  • Curated References - A distillation of the best external resources for each topic

Read Non Linearly!#

If you know what you to read just skip to it.

For those unsure here’s the learning approach I suggest

  1. Start with fundamentals - Knowing the basic theory and how to read the code will make understanding everything else easier.

  2. Tune a model on your own - Tuning a model will give you a fantastic sense of how these work, don’t work. This chapter is not written yet but soon will be

  3. Read in whatever order you want - There are many branching off points here, from reading into particular training methods, to red teaming. Pick whatever suits your interests.

What this guide is not#

  • Self Contained Resource - This is a guide to navigate you to the best content. It is not meant to be encyclopedia

  • Academic article or textbook - We’re not presupposing that you have detailed level of knowledge of every topic or writing in the style of traditional academia.

  • A static body of work - The field is moving fast, this text will change accordingly.

Who am I?#

I’m an applied data person and I think I’ve worked at some cool places (if I may say so myself). For those who want my professional summary can look at my LinkedIn or Website.

However, here’s the summary. I’m fascinated with applied math. Math is awesome. Math lets us do so many amazing things, build structures, launch rockets into space, and generally understand the world. But comparatively these new models It can “read” vast amounts of human knowledge, compress it into a small space, and then “talk” to you generating words, images, and videos that change your perception of the world.

This could be new facts

When was the first transistor invented?

idea generation

Give me 5 recipes for eggs

or story and entertainment

Tell me a story about a boy who loved math and wrote a book

This stuff is insane, and this is coming from who launched stuff into space. However it’s not magic. It’s just math, and it’s better if more folks understand it.

Text organization#

This guide is organized into topics. You can dive straight into the one that interests you. The primary motivation came from my study club series and namely my inability to find a single resource that captured the relevant knowledge. Large Language Models, and specifically text sequence to sequence models will be the first topic. Over time I plan to write more about other models, tasks and datasets, such as images, sound, and other generative mediums.

Getting the latest updates the soonest#

Every now and then I send updates over email. If you enjoy that here’s where to sign up.