I have no background — how do I get started with AI?

One of the most common questions I get from my friends is: how do I get into generative AI? In this post I'll share my take, assuming you have no prior experience in AI, ML, or deep learning.

1. Learn Python. Python is the language of deep learning and generative AI. The first step is to get an intuitive feel for it — and the best way is to learn it alongside the material in step 2.
2. Do the Deep Learning Specialization by Andrew Ng. Before generative AI, you need a foundation in deep learning. Watch the lectures on Coursera and write the code along with them. This is also where you'll sharpen your Python. Link →
3. Optionally: Machine Learning Specialization by Andrew Ng. Covers supervised/unsupervised learning, Random Forest, regression — useful context for understanding the broader ML landscape. Link →
4. Read "Attention Is All You Need." Generative AI in language was sparked by this one paper from Google Research. You don't need to understand every equation — just look at the Transformer architecture and get the feel of the attention operation. Link →
5. Watch Andrej Karpathy's talks on LLMs. The two streams of generative AI are large language models (language) and diffusion models (images and video). For LLMs, Karpathy's introductory talk is the best primer I know. Intro talk →   Training deep-dive →
6. Write your own inference code with Hugging Face. Hugging Face aggregates all open-source AI models and has a transformers library to work with them. Try running Meta's Llama 2 locally — the code is on their site. Link →
7. Use Google Colab for experiments. Colab gives you free GPU access. Start there — load a model, ask it a question, break something, fix it. Link →

From here, you can go deeper — fine-tuning with LoRA, building RAG pipelines, exploring diffusion models. I'll cover those in future posts.