I said it many times before: Artificial Intelligence is not Machine Learning and you do not need to know ML in order to use AI. With the advent of pre-trained models you can leverage the pre-existing knowledge of models by just calling an API. But...

...in order to truly understand how AI works you really want to study ML. That still doesn't mean you will have to train models from scratch, but knowing how models are trained and then use for inference gives you an edge.

Understanding Machine Learning gives you the most critical edge over any other software engineer that just knows how to stitch engineered prompts with langchain and RAG.

So, when in doubt, learn ML and start from scratch. And that is what I do every now and then when I get nostalgia about all the core, basic ML knowledge I learned in school many years ago. And also because it is fun to go back to basics, and program cool stuff in Python!

So here are the top 3 resources to truly learn AI, that we all revisit every now and then

📌 #1 - Neural Networks Zero to Hero by Andrej Karpathy

This is the most valuable ML resource available: 10 videos in between 2hs and 4hs. Starting from neural networks, all the way to deep learning and implementing an LLM from scratch, it covers A LOT. You really want to dedicate time to this and follow all the tutorials. Going over a 2h video will take you more than 2h if you are truly studying, following the tutorials and understanding every detail vs just passively watching like a Netflix show.

This is probably everything you need, but if you want to keep going down the rabbit hole...

📌 #2 - Learn PyTorch by Daniel Bourke

PyTorch is the core ML / math library behind all modern AI and LLM ML models, so you want to become extremely familiar with its API and concepts. Ultimately all these models are just data translated into formulas that are expressed via the PyTorch API.

Studying this will give you so much more than you will ever need. If you still want to keep going for more ....

📌 #3 - Stanford CS25 - Transformers United by Stanford University

Beyond understanding neural networks and deep learning, you want to understand how the most used deep learning architecture works. This course will give you that! Also it's fun to see how it kept evolving from v1 to v4.

Let me know if you enjoy these resources, have other resources that should be on this list, and if you also enjoy learning for fun!