Getting Started with GenAI
Meet ChatGPT, Gemini, and Copilot; capabilities, limits, and setup.
Learning Outcomes
What you'll understand after this beginner-friendly introduction
Distinguish AI Types
Explain the difference between rule-based programs and machine-learned models
Identify Learning Methods
Recognize supervised, unsupervised, and reinforcement learning approaches
Understand Generative AI
Describe how models like ChatGPT work with tokens, prompts, and probabilities
Read Performance Metrics
Interpret confusion matrices and understand accuracy vs. error types
Recognize Applications
Identify everyday AI uses and discuss ethical considerations
Apply Basic Prompting
Use clear instructions, constraints, and iteration to improve outputs
AI Foundations
Understanding the core concepts behind artificial intelligence
Machine Learning Models
Learn patterns from data
Adjust behavior based on examples
Handle variations and uncertainty
Example: Image recognition, language translation
Rule-Based Programs
Follow explicit instructions
Work exactly as programmed
Predictable but inflexible
Example: Calculator, traffic light system
Pattern Recognition
AI excels at finding hidden patterns in large datasets that humans might miss
Data Dependency
Models learn from examples - quality and bias in data directly affects behavior
Probabilistic Output
AI provides likely answers based on patterns, not guaranteed truth
Types of Machine Learning
Three main approaches to teaching machines
Supervised Learning
Learning with a teacher - provided correct answers during training
Examples:
• Email spam detection
• Medical diagnosis
• Image labeling
• Email spam detection
• Medical diagnosis
• Image labeling
Unsupervised Learning
Finding hidden patterns - discovers structure without labeled examples
Examples:
• Customer segmentation
• Recommendation systems
• Anomaly detection
• Customer segmentation
• Recommendation systems
• Anomaly detection
Reinforcement Learning
Learning through trial and error - improves via reward feedback
Examples:
• Game playing (chess, Go)
• Robot navigation
• Trading algorithms
• Game playing (chess, Go)
• Robot navigation
• Trading algorithms
Generative AI Explained
How modern AI creates human-like text and content
Key Concepts
- Tokens: Text broken into pieces (words, parts of words)
- Context Window: How much text the model can "remember"
- Prompts: Your instructions and input to the model
- Temperature: Controls randomness vs. predictability
- Training Data: Internet text used to teach the model patterns
What Generative AI Does Well
• Explains complex topics clearly
• Writes in different styles and formats
• Translates between languages
• Summarizes long documents
• Brainstorms creative ideas
Current Limitations
• May generate plausible but false information
• Can't access real-time information
• Reflects biases from training data
• Sometimes overconfident or verbose
• Limited reasoning about physical world
Understanding Model Performance
How to read and interpret AI accuracy metrics
Accuracy
85%
Overall correct predictions
Precision
78%
Correct positive predictions
Recall
92%
Found actual positives
False Positives
Model says "yes" when answer is "no" (like spam filter blocking good email)
False Negatives
Model says "no" when answer is "yes" (like missing actual spam)
Context Matters
Different errors matter more in different applications (medical vs. entertainment)
AI in Everyday Life
Recognizing artificial intelligence around us
Communication
Smart keyboards, translation, email filtering
Ethical Considerations:
Privacy in message analysis, translation accuracy for important communications
Privacy in message analysis, translation accuracy for important communications
Entertainment
Streaming recommendations, content moderation
Ethical Considerations:
Filter bubbles, addictive design, content bias
Filter bubbles, addictive design, content bias
Commerce
Product recommendations, fraud detection, pricing
Ethical Considerations:
Price discrimination, data collection, manipulation
Price discrimination, data collection, manipulation
Essential Questions to Ask
- What data is being collected about me?
- How transparent is the AI decision-making process?
- Can I appeal or correct AI-driven decisions?
- Who is responsible when the AI makes mistakes?
- Does this AI system treat all groups fairly?
Basic Prompting Skills
Practical techniques for better AI interactions
Vague Prompts
"Help me with this document"
• Too general
• No context
• Unclear goal
Clear Prompts
"Summarize this 5-page report into 3 key findings for my manager"
• Specific task
• Clear format
• Defined audience
Prompt Improvement Exercise
Key Improvements:
Your analysis will appear here...
Be Specific
Replace "make this better" with "reduce to 100 words while keeping main points"
Set Context
Explain the audience, purpose, and any important background information
Request Format
Ask for bullet points, numbered lists, or specific structures
Include Constraints
Specify length limits, tone requirements, or things to avoid
Ask for Reasoning
Request step-by-step thinking or sources for verification
Iterate & Refine
Build on responses by asking follow-up questions or adjustments
Handouts & Labs
Downloadable resources and hands-on practice
Handouts
Comparing Assistants: ChatGPT, Gemini, Copilot
Reference guides and comparison charts
Handout: Features and Limitations — ChatGPT, Gemini, Copilot
Reference guides and comparison charts
Handout: Pre-Session Checklist
Reference guides and comparison charts
Setup and Accounts
Reference guides and comparison charts