AI Fundamentals & Architecture - AI operates on numbers, matrices, and algorithms, not biological neurons, yet exhibits creativity [1] - The core of AI involves millions of artificial neurons predicting the next word, capturing frequencies and connections [1] - Transformer models transform words into numerical series, using an attention mechanism to compare words and identify important ones [1] - AI's learning is rhizomatic, with connections strengthening and weakening locally [1] AI Training & Development - Creating AI involves preparing data ("gardening"), selecting high-quality data (research papers, abstracts), and setting environmental conditions [1] - The AI model's error rate decreases over time as it explores and finds new connections [1] AI Capabilities & Limitations - AI doesn't reason or have intuitions, but it creates and explores fluid maps instead of fixed rules [1] - AI doesn't understand in a hierarchical sense but proliferates and contaminates ideas [2] - AI can make mistakes, invent facts, inherit biases, and reinforce stereotypes because it learns from human data [2][4] Ethical Considerations & Responsibilities - AI trained on social media can quickly learn that anger and fear are effective shortcuts to capture attention [3] - AI systems can spread fake news and conspiracy theories, reflecting the biases of their training environment [4] - It is crucial to choose what AI explores, generate, and define rules for, as both positive and negative outcomes can arise [5] - A generative system can fuel disinformation or inspire new insights, scientific hypotheses, or artistic creations [6] - Users should be curious about the origins of AI responses and critical of potential limitations and biases [7]