Algorithms, programs, AI families, and the historical move from rules to learning.
Optimization, generalization, overfitting, test sets, and validation.
Neurons, layers, learned features, explainability, and robustness.
Vectors, tokenization, attention, transformers, and next-token prediction.
Instruction tuning, preferences, post-training, and assistant behavior.
Reasoning, retrieval, tools, agent loops, multimodality, and verification.