Happy to help and thanks! Do let me know how the material could be better. Always looking to improve it.
To answer your question:
EECS Major in college; 20 years of engineering and product management experience. I have given a few technical talks at conferences and I do enjoy the process of explaining things though it takes a surprising amount of work.
When I went to school ironically neural nets were the one thing they didn't cover in the intro to AI courses. I've basically learned modern AI from just filling my own curiosity over the years through online resources on nights and weekends. Learned a lot from Jeremy Howard's Fast.ai and Andrej Karpathy's stuff just like everyone else. I really wanted to know how every step of GPT worked, kind of like how you learn Computer Architecture in college: you learn how CPUs work in principle starting with circuits. Then I got a crazy idea the whole model could fit in a spreadsheet because well I just really like spreadsheets. Went down a 2-3 month rabbit hole in my non-existent sparetime to make it work.
The passion and curiosity you have for the topic are palpable in the videos. Thank you again for also sharing with us your exploration into this learning endeavor.
This work is another classic in the "neural nets meet spreadsheets" genre [0]. Really helps visualize what is going on in (at least some) latent spaces.
This is so awesome!! I'm going to have to show this in the embeddings video I'm working on when I discuss non-text embeddings and CLIP.
While I created spreadsheet-are-all-you-need.ai as teaching tool, as I've been playing with it I've been having a growing suspicion the spreadsheet interface for AI might be useful beyond teaching, either as a power user control interface or for interpretability. For example, making simple changes to the architecture of GPT and observing how it changes the model behavior can be as simple as cloning a tab and a few spreadsheet functions. Of course, you can do the same in python as well so it remains to be seen.
It's actually available in beta. It was announced while I working on this project but I kept going with pure Excel functions because I wanted to illustrate the transformer without abstractions getting in the way. It would make many aspects easier but also make it easier to hide a lot.
That being said, Python+Excel makes a ton of sense in general. And in this project, it would help in the tutorials. For example, in the embeddings tutorial I'm working on I wanted use PCA plots and SVD to illustrate the workings of embeddings but neither are natively supported in Excel without paid plug-ins. But both are easy in Python.
Not 100% sure this will work, but there is a Mailchimp signup, and Mailchimp newsletters have RSS feeds. Based on the newsletter URL the RSS feed should be https://us21.campaign-archive.com/feed?u=96d33fd949860389358.... There's nothing in there yet but I would expect posts to show up if they start sending out emails.
To be notified of updates. Why in preference to email? Web feeds are typically presented as information streams while email is presented as interpersonal dialogue. Organizing your information streams into an aggregator is clarifying.
Cannot answer for parent post, but can share personal perspective. I love Atom / RSS because it allows me to aggregate all the information that I care about in one place. It greatly improves the signal to noise ratio and spares my communication channels for direct human interactions.
Next video will be on embeddings and hopefully done soon-ish.