Introduction to AI Tools and Fundamentals

Speaker: Software Engineer, 15 years of experience

Session Overview

The speaker introduces participants to AI tools — Gemini, Claude, NotebookLM, and a text-to-video generator — before diving into the fundamentals of artificial intelligence.

Fundamentals of AI

Machine learning is a subset of AI where systems learn patterns from data rather than relying on fixed rules. It has three main approaches:

Supervised learning – the model trains on labelled data with known input-output pairs.
Unsupervised learning – the model finds patterns without labels, used in clustering and dimensionality reduction.
Reinforcement learning – the model learns through trial and error using a reward system (e.g. AlphaGo).

AI Architectures

A Convolutional Neural Network (CNN) is commonly used for image generation and visual tasks.

The three pillars of modern AI are: predict, create, and execute tasks.

How Language Models Generate Text

Words are broken into tokens, and the model predicts the next token using a probability function — it doesn't truly "understand" language, it calculates it.

Transformer Architecture

Earlier models like RNNs read one word at a time, making them struggle with long context. Transformers solve this with an attention mechanism that looks at the entire sentence and connects relevant words simultaneously.

Example: In "The cat sat on the mat because it was tired," the model must determine what "it" refers to — this requires understanding the full sentence context.

Model Creation and Improvement

Base model – trained on a large corpus of internet text to build general language understanding.
Fine-tuning – instruction-tuned using human-created datasets to teach the model to answer questions and be helpful.
Reinforcement learning – the model is given a question and answer pair and iteratively improved through feedback.

Key Takeaway

LLMs operate on probability, not understanding. What appears to be reasoning is an illusion — it is fundamentally mathematics.