Some Of The Best Educational AI Resources Are Free

…and some of the worst are Amazon books and paywall-blocked Medium posts

Source: https://jalammar.github.io/illustrated-transformer/. The Illustrated Transformer resource is licensed under Creative Commons, meaning it is free to reproduce and cite so long as it is not used on a commercial and/or paywall-blocked Medium post. It should be noted that the encoder/decoder architecture is not necessarily used in all transformers (GPT is based on the decoder-only architecture)

I recently read The ChatGPT Millionaire, which I do not recommend — it is an Amazon book that is primarily generated by ChatGPT, yet it has 4.5 stars and thousands of sales. The author name seems to be a pen name. The author bio just says “computer science major.” What you get if you read this book is a lot of misinformation overestimating what ChatGPT can do, followed by a lot of prompts and copied-and-pasted ChatGPT content.

Something similar is happening on social media:

…but good educational content is out there, and a lot of it is free. This, for example, is a YouTube video by Computerphile about transformers (the “T” in GPT)

This is the paper they are referring to in their talk, though it should be noted that GPT does not follow this architecture exactly (decoder-only)

Attention Is All You Need

This is a GitHub page by Jay Alammar, who includes customized animations in his blog posts and licenses his work under a Creative Commons Noncommercial license.

The Illustrated Transformer

Free Resources For AI Overviews

This is an introductory video about how machine learning works, explained in very simple terms. Smack had issues with it because he thought it was too simplified, and gave the impression that training is done randomly — but I think it’s a good starting point.

This is a blog post by Stephen Wolfram about how ChatGPT works.

What Is ChatGPT Doing ... and Why Does It Work?

The blog post above is readable, but not nearly as light as the CPGGrey YouTube video.

This is a video overview of machine learning.

For completeness, this is the website for CS230N, a Stanford course on deep learning. Notice that the slides are open for the world to see, and that they even have midterms and midterm solutions up without requiring a Stanford login.

Syllabus

“Flavors” Of AI

This is a YouTube channel called Serrano Academy, and below is their video on Recurrent Neural Networks. Recurrent Neural Networks can be thought of, to some extent, as the predecessor to transformers.

For the hell of it, I am going to link Smack’s thesis at Sacramento State, “Cross-Domain Adaptations For RF Fingerprinting Using Prototypical Networks.”

Cross-Domain Adaptation for RF Fingerprinting Using Prototypical Networks | Proceedings of the 20th…

I have already provided a few links about transformers above.

This is an illustrated guide to attention (that’s not what it’s called, that’s just what I call it). The author is Jay Alammar.

Visualizing A Neural Machine Translation Model (Mechanics of Seq2seq Models With Attention)

The blog post references this paper about sequence to sequence learning with neural networks

And this paper about using an RNN Encoder-Decoder

Finally, here is the famous, 155-page paper about GPT-4 and its “sparks of artificial general intelligence.”

You could read that and/or watch the YouTube overview.

The video above is not affiliated with the researchers, per se, but it does have the little stamp of approval from Sebastien Bubeck.

Closing Thoughts

If you do not know anything about ChatGPT, please take the time to investigate some of the free educational resources before buying an Amazon book with a title like “The ChatGPT Millionaire.” Notice I did not provide a link to that book.

There are dozens with that exact same title.