If I had to learn AI PM again, I’d start here --- And what's not worth learning: (unless you're in a research lab or infra team) 1. Reinforcement Learning Cool theory, almost no applications in real product work. Most “RL” for PMs = fine-tuning. 2. Python Being able to read scripts is more than enough. Visual orchestration > coding. 3. Old Deep Learning Theory CNNs, LSTMs, attention math, optimizers, gradients. No impact on your daily work. 4. Academic AI Ethics Learn practical governance & security. Skip the philosophy debates. 5. Architecture Research MoE internals, RoPE math, distributed training, scaling laws. You only need high-level intuition. 6. ANN Algorithms for RAG FAISS internals, IVF/HNSW math, vector quantization. Focus on how retrieval behaves, not how it’s implemented. 7. ML Math Linear algebra, divergence measures. You need ML intuition, not calculations. Hope that helps!