WEEK 8

AI in Your

Day-to-Day Life

From machine learning fundamentals to generative AI

BinaryTree

What Is AI? Two Types We'll Cover

BinaryTree

Agentic AI

• AI that takes actions in the world on your behalf
• Customer service chatbots and virtual assistants
• Smart glasses with real-time scene understanding
• Workflow management and task automation
• AI that perceives, decides, and acts

Generative AI

• AI that creates new content from a prompt
• OpenAI's GPT models (ChatGPT)
• Anthropic's Claude
• Sartify's PawaAI
• Generates text, images, code, audio, and more

AI Relies on Data

BinaryTree

Every AI system — from the simplest regression to the most advanced LLM — runs on one thing: data.

Regressions

Simple ML algorithms that find relationships between variables. Predict a house price from square footage.

Neural Networks

Systems of interconnected nodes inspired by the human brain. Each layer extracts more complex patterns.

Layers of Complexity

Deep learning adds many layers to neural networks, enabling them to recognize images, speech, and language.

Data

All of the above algorithms depend on large quantities of high-quality training data to work accurately.

What Is Data?

BinaryTree

Data is structured or unstructured information recorded by a machine.

Numbers & Statistics

Text & Documents

Images & Video

Audio & Speech

User Behavior (clicks, time on page)

Sensor & IoT Readings

Data Isn't Always Clean

BinaryTree

Real-world data comes with problems — and the field that handles this is called Data Science.

Garbage Values

Corrupted or nonsensical data points that slip into datasets

Incorrect Labels

Misclassified training examples that teach the model wrong patterns

Varying Formats

Images, audio, text, and numbers must be standardized before use

Missing Data

Incomplete records that need to be filled in or removed carefully

Deep Learning in Action

BinaryTree

Example: AI in Self-Driving Cars

BinaryTree

Self-Driving AI CAN...

• Recognize and classify other vehicles on the road
• Detect and avoid pedestrians in crosswalks
• Read and respond to traffic signs and signals
• Adjust speed and following distance automatically
• Navigate complex highway merges and lane changes

Self-Driving AI CANNOT (Yet)...

• Reliably recognize hitchhikers by the road
• Interpret hand signals from cyclists or pedestrians
• Handle highly unusual or edge-case scenarios
• Navigate fully unpredictable human behavior
• Make ethical decisions in unavoidable accident scenarios

Generative AI

BinaryTree

How Generative AI Works: Pre-Training

BinaryTree

• Generative AI models are trained on massive datasets of text from the internet, books, and code
• During pre-training, the model learns to predict the next token (word piece) given prior context
• A tokenizer breaks text into smaller units — try it: https://platform.openai.com/tokenizer
• The model sees billions of examples and learns patterns, facts, and reasoning
• Visualization of the internals: https://bbycroft.net/llm

How Generative AI Works: Prompt Engineering

BinaryTree

Anyone can be a prompt engineer — and a skilled one can create almost anything.

Be Specific

Vague prompts get vague answers. Include details, format, tone, and length.

Chain Prompts

Break complex tasks into steps. Build on prior responses iteratively.

Few-Shot Examples

Show the model examples of what you want before asking it to do it.

Include Context

The more relevant context you provide, the better the output.

What You Can Do With Generative AI

BinaryTree

• Write best-in-class social media posts and marketing copy
• Generate ideas for expanding your business or side project
• Expedite document analysis, summarization, and synthesis
• Learn almost any new skill through interactive AI tutoring
• Write, debug, and explain code across any language
• Draft professional emails, cover letters, and proposals

Thank You!!!

A special thanks to:

• Andrew Ng's DeepLearning.AI Coursework
• 3Blue1Brown's Machine Learning Video Series

BinaryTree