AI for Neanderthals

Table of Contents

Core Concepts

AI (Artificial Intelligence): Computers that think and act a bit like humans—smart assistants doing tasks like chatting or spotting patterns.
ML (Machine Learning): AI that learns from examples, like teaching a kid by showing pictures instead of rules.
Neural Network: Brain-inspired layers of math that spot patterns in data, powering image recognition or predictions.

LLM (Large Language Model): Giant AI brain trained on billions of words to chat, write, or code like a human (e.g., ChatGPT, Perplexity).
Agent: Smart AI that doesn’t just answer—it observes, decides, and acts using tools (like booking flights or debugging code).
Skill: Pre-built toolkit for agents—reusable instructions/scripts for specific jobs, like Copilot’s .github/skills folders.
Prompt: Your message to AI. Good prompts = better results (“Explain like I’m 5” beats “Explain”)

Prompt Engineering: Crafting the perfect question to get spot-on AI answers
Hallucination: When AI confidently makes stuff up—always double-check!
Token: Chunk of text AI processes (word or part-word); limits how much it “remembers.”
Fine-tuning: Tweaking a pre-trained AI for your niche (e.g., SAP ABAP code gen).
RAG (Retrieval-Augmented Generation): AI pulls real docs before answering, reducing lies—your user-info matches this pattern
Context Window: How much info AI holds in “short-term memory” during a chat

Start with prompts, experiment in Perplexit, Copilot, Gemini etc.