DATA4800 - Understanding the fundamentals
Experienced Lecturer for 4 years at:
AI Specialist at leading tech companies:
Attendance is mandatory for all sessions
80% Pass | 20% Fail
End-of-class quiz provides bonus points for assessments
Visual programming for data analysis
Cloud-based coding environment
Collaborative documentation
AI is the simulation of human intelligence in machines that are programmed to think, learn, and make decisions like humans.
Alan Turing asked "Can machines think?" and proposed the famous Turing Test to evaluate machine intelligence.
The Dartmouth Conference officially established AI as an academic field. Researchers predicted machines would match human intelligence within 20 years.
Initial optimism led to basic programs solving algebra problems. However, limited computing power and unrealistic expectations led to reduced funding.
Expert systems became commercially viable. Neural networks showed promise but faced limitations. Another AI winter occurred in the late 1980s.
IBM's Deep Blue defeated world chess champion Garry Kasparov, marking a milestone in AI's ability to outperform humans in complex tasks.
Big data, powerful computing, and deep learning breakthroughs led to practical AI applications in every industry, from healthcare to finance to retail.
Businesses generate massive amounts of data daily. AI can process and analyze this data at speeds impossible for humans, uncovering insights that drive better decisions.
AI frees humans from mundane, repetitive work, allowing employees to focus on creative problem-solving and strategic thinking that adds more value.
AI analyzes patterns and predicts outcomes with high accuracy, helping businesses make data-driven decisions and reduce costly errors.
AI enables businesses to understand individual customer preferences and deliver personalized recommendations, increasing satisfaction and loyalty.
Expected AI contribution to global economy by 2030
The Turing Test is a method proposed by Alan Turing in 1950 to determine if a machine can exhibit intelligent behavior indistinguishable from a human.
Human judge asks questions
Human
Machine
Interrogator: "What is your favorite season?"
Responder: "I love autumn because of the beautiful colors and perfect weather for outdoor activities."
Interrogator: "Can you solve 234 + 789?"
Responder: "Let me think... I believe it's 1023."
Result: If the interrogator cannot reliably distinguish the machine from the human, the machine is said to have passed the Turing Test.
Designed for specific tasks. Exceeds human performance in one area.
Examples: Chess AI, Image recognition
Human-level intelligence across all domains. Can apply intelligence to any problem.
Status: Not yet achieved
Surpasses human intelligence in all fields including creativity and wisdom.
Status: Theoretical
ML is a subset of AI that enables computers to learn and improve automatically from experience without being explicitly programmed for every task.
Sales, customers, operations
Pattern recognition, predictions
Actionable recommendations
Increased profit, efficiency
Key Insight: AI/ML turns vast amounts of raw business data into strategic advantages. What once took analysts weeks can now be done in minutes with higher accuracy.
Challenge: Understanding diverse customer groups
AI/ML Solution: Clustering algorithms group customers by behavior, enabling targeted marketing campaigns with 30-40% higher conversion rates.
Challenge: Identifying fraudulent transactions
AI/ML Solution: Real-time anomaly detection models flag suspicious activities, reducing fraud losses by up to 70%.
Challenge: Predicting patient volumes
AI/ML Solution: Time series models predict demand using historical data and external factors, optimizing staff allocation and reducing wait times.
Challenge: Balancing stock levels
AI/ML Solution: Predictive models forecast demand and recommend optimal inventory levels, reducing holding costs by 20-30%.
Business question
Collect information
Build models
Extract insights
Business recommendations
Problem: Which customers should we target for our next marketing campaign?
Data: Customer purchase history, demographics, email engagement
AI/ML Model: Classification model predicts likelihood of response
Result: Target high-probability customers, achieving 3x better ROI than random targeting
of leading businesses have ongoing AI investments
Companies using AI/ML make faster, more accurate decisions than competitors who rely only on human analysis.
Automation of routine tasks and optimized operations can reduce costs by 20-40% in many business processes.
Personalized recommendations and targeted marketing increase customer lifetime value by 15-30%.
Predictive models identify potential issues before they occur, preventing costly problems and reducing operational risks.
Learns from labeled examples
Examples: Email spam detection, house price prediction
Finds patterns in unlabeled data
Examples: Customer segmentation, anomaly detection
Learns through trial and error with rewards and penalties
Examples: Game playing AI, autonomous driving
We have labeled data - we know the outcomes!
| Student ID | Study Hours per Week | Class Attendance | Assignment Score | Result (Target) |
|---|---|---|---|---|
| 001 | 25 | 95% | 85 | PASS |
| 002 | 8 | 60% | 45 | FAIL |
| 003 | 20 | 88% | 78 | PASS |
| 004 | 5 | 40% | 35 | FAIL |
| 005 | 30 | 100% | 92 | PASS |
Supervised Learning: The algorithm learns from this labeled data to predict outcomes for new students.
But now we have NO labels - we don't know the outcomes!
| Student ID | Study Hours per Week | Class Attendance | Assignment Score | Result (?) |
|---|---|---|---|---|
| 001 | 25 | 95% | 85 | ? |
| 002 | 8 | 60% | 45 | ? |
| 003 | 20 | 88% | 78 | ? |
| 004 | 5 | 40% | 35 | ? |
| 005 | 30 | 100% | 92 | ? |
High study hours, good attendance, high scores
Likely to pass
Low study hours, poor attendance, low scores
Likely to fail
Unsupervised Learning: We group students by similar patterns and estimate which group a new student belongs to!
Think of AI as the destination and ML as one of the vehicles to get there.
AI: Computer vision, NLP, robotics, expert systems, ML
ML: Decision trees, neural networks, clustering, regression
Deep Learning: Multi-layered neural networks
Medical diagnosis, drug discovery, personalized treatment
Autonomous vehicles, traffic optimization, route planning
Fraud detection, algorithmic trading, credit scoring
Recommendation engines, price optimization, chatbots
Content recommendation, game AI, music generation
Quality control, predictive maintenance, robotics
Remember: AI is a powerful tool that should augment human capabilities, not replace human judgment in critical decisions.
Understanding how AI/ML fits into business decision-making
What should we do?
What will happen?
Why did it happen?
What happened?
Traditional: Monthly sales reports showing revenue by product category
AI-Enhanced: Automated discovery of seasonal patterns, customer purchase clusters, and unusual sales spikes with intelligent alerts
Traditional: Manual analysis of why customers left using surveys and interviews
AI-Enhanced: ML models identify that customers who experienced more than 3 service calls plus more than 5 day response time have 80% churn probability
Traditional: Linear forecasting based on historical patient volumes
AI-Enhanced: ML models predict patient demand using weather, flu trends, demographics, and social media sentiment to optimize staffing
Traditional: Manual route planning and inventory decisions based on experience
AI-Enhanced: Real-time optimization considering traffic, weather, fuel costs, and demand predictions to automatically route deliveries and manage inventory
Understand current state
Find root causes
Forecast outcomes
Recommend actions
Key Insight: Each level builds on the previous one. You need to understand "what happened" before you can determine "what should happen."
Next Week: We will explore supervised learning algorithms in depth, starting with Decision Trees and how to build your first prediction model!
Remember: AI and ML are tools to enhance your analytical capabilities. The goal is to make you a more effective business analyst who can leverage these technologies to solve real-world problems.
Q: Do I need to be good at math for this course?
A: Basic statistics helps, but we focus on practical applications using visual tools like Orange Data Mining.
Q: Will we learn programming?
A: We will use Python in Google Colab, but we start with fundamentals and build gradually.
Q: How is AI different from traditional programming?
A: Traditional programming uses explicit rules. AI/ML learns patterns from data without being explicitly programmed.
Q: Can I use AI tools like ChatGPT for assignments?
A: AI tools can help with learning, but your work must demonstrate understanding of concepts. Check the assessment guidelines.
Don't forget to complete the end-of-class quiz for bonus points!
Supervised Learning: Decision Trees and Classification