Artificial Intelligence

Application, Problems and Techniques

Dr. Dhaval U Patel • 2025

Part 1: Foundation

Why Intelligence Matters to Humans

Why We Are Called "Homo Sapiens"

Let's start with ourselves! We humans proudly call ourselves "Homo Sapiens" - and this name tells our entire story.

Homo = Latin word meaning "human" or "man"
Sapiens = Latin word meaning "wise" or "intelligent"
Together: "Wise Human" - intelligence defines who we are!

                    For thousands of years, we've been obsessed with understanding how we think - how we perceive, understand, predict, and manipulate our world. This curiosity led us to create AI!
                

Evolution

Think about how computers have evolved alongside human ambition:

Early Computers (1940s-1960s): Only dealt with numerical calculations - fancy calculators!
Modern Computers: Now involved in reasoning with knowledge, not just crunching numbers
With AI: Computers transform from "useful tools" to "thinking partners"

The Goal: Make computers perform tasks that humans are naturally good at - thinking, reasoning, learning, and problem-solving!

⏭️ Next Question: But what exactly is "Artificial Intelligence"?

Part 2: Definition

Artificial Intelligence?

The Great Challenge: Defining Intelligence

The Core Problem: How can we build "artificial intelligence" when we can't even agree on what "intelligence" means?

The Dictionary Says: Intelligence is "capacity for learning, reasoning, understanding and similar forms of mental activity."

The Problem: This just gives examples, not a true definition!

Intelligence is the ability to acquire knowledge, adapt to new situations, solve problems, and achieve goals effectively in a changing environment.

The Solution: Computer scientists created four different approaches to defining AI...

The Classic 2×2 Framework

Computer scientists AI definitions:

🧠 THINKING

“[automation of] activities that we associate with human thinking, activities such as decision making, problem solving, learning…” (Bellman 1978)

🎯 ACTING

“The study of mental faculties through the use of computational models” (Charniak& McDertmott 1985)

👥 HUMANLY

“The study of how to make computers do things at which, at the moment, people are better” (Rich & Knight 1991)

⚖️ RATIONALLY

“The branch of computer science that is concerned with the automation of intelligent behavior” (Luger & Stubblefield 1993)

The Four Traditional Definitions

🧠👥 Thinking Humanly

Goal: Automate human thinking processes

Example: Decision-making, problem-solving, learning like humans

🎯👥 Acting Humanly

Goal: Make computers do things humans are better at

Example: Turing Test - can you tell if it's human or AI?

🧠⚖️ Thinking Rationally

Goal: Study mental processes through logical models

Example: Perfect logical reasoning and deliberation

🎯⚖️ Acting Rationally

Goal: Automate intelligent behavior rationally

Example: Always taking the "right" action for the goal

🤔 Question: Which definition sounds best to you? Let's examine each one...

Why Three Definitions Don't Work

🧠👥 Thinking Humanly - The Problem:

Humans can't explain their own thinking
We don't want to copy human flaws (bias, emotions, irrationality)

🎯👥 Acting Humanly (Turing Test) - The Problem:

Human behavior should be a milestone, not the goal
We want AI to be better than humans, not just equal

🧠⚖️ Thinking Rationally - The Problem:

Not all intelligence is logical (sometimes you need quick reflexes)
Logic without purpose isn't intelligence

✅ That leaves us with: Acting Rationally - doing the right thing!

🎯⚖️ Acting Rationally

The Goal: Focus on taking the "right action" regardless of the internal process.

                    Why This Works Best: It's about achieving goals optimally, whether through deliberation or reflexive action!
                

Key Advantages:

Goal-oriented: Actions are in pursuit of something valuable
Flexible: Can include both deliberative and reflexive actions
Measurable: We can objectively evaluate if actions are "right"
Practical: Focuses on outcomes, not internal processes

Our Working Definition: AI is a branch of science that develops machines to find solutions to complex problems by taking optimal actions to achieve their goals.

🔄 This leads us to the modern view: The Rational Agent

The Rational Agent - Modern AI Definition

Building on "Acting Rationally," we arrive at the Rational Agent View - the most widely accepted definition of AI today.

                    Core Definition: For each possible percept sequence, does whatever action is expected to maximize its performance measure on the basis of evidence perceived so far and built-in knowledge.
                

The Components:

🔍 Sensors: How the agent perceives the world
🎯 Effectors: How the agent acts upon the world
🧠 The AI Brain: The computational part that decides actions
🌍 Environment: The world the agent operates in
📊 Performance Measure: How we define "success"

💡 Now we know WHAT AI is - let's see what types exist today!

Part 3: Current Reality

What Types of AI Exist Today?

The Three Levels of AI Intelligence

🎯 Narrow AI (Current Reality)

What We Have Today

Performs ONE specific task very well
Cannot go beyond its training
Examples: Chess programs, facial recognition, Siri, ChatGPT

Think: A genius chef who can only make one perfect dish!

🧠 General AI (The Goal)

The Holy Grail

Can perform ANY intellectual task like humans
Thinks and learns like humans
Currently doesn't exist!

This is what researchers are trying to achieve - true human-level intelligence in machines.

Super AI: Beyond Human Intelligence

The final frontier - and the most controversial topic in AI!

                    Super AI (Artificial Superintelligence): Machines that surpass human intelligence in every possible way and can perform any task better than humans.
                

Current Status: Purely hypothetical - doesn't exist yet
The Big Question: Would this be humanity's greatest achievement or greatest threat?
Research Focus: How to ensure AI remains beneficial to humanity

🌟 Now let's see where AI is making a real impact today!

Part 4: Applications

AI in the Real World

Why Do We Need AI? The Perfect Partnership

🖥️ Where Computers Excel

1. Numerical Computations:

Computers can multiply 1.89254 × 8.29743 instantly
Humans need minutes and make errors
Perfect accuracy, lightning speed

2. Information Storage:

Store virtually unlimited information
Perfect recall, no forgetting
Humans have limited memory capacity

3. Repetitive Operations:

Never get bored or tired
Consistent performance
Humans make mistakes when bored

🧠 Where Humans Excel

Intelligent Tasks:

Creative problem solving
Understanding context and emotions
Making intuitive leaps
Learning from few examples
Dealing with ambiguity

The Goal of AI: Combine the computational power of machines with the intelligent reasoning abilities of humans!

What AI Can Give Us:

Solutions to complex real-world problems
Virtual assistants (Siri, Alexa)
Robots for dangerous environments
New technologies and opportunities

Real-World Applications of AI

AI isn't just science fiction - it's already transforming our world in amazing ways:

🏥 Healthcare: Providing better and faster diagnosis than human doctors
🚗 Transportation: Self-driving cars and traffic optimization
💰 Finance: Chatbots resolve consumer queries and detect fraud
📱 Social Media: Analyzing trends and understanding user requirements
🛒 E-commerce: Recommendation systems (Netflix, Amazon)
🎮 Gaming: AI machines can play strategic games at superhuman levels
📚 Education: Personalized learning and automated grading
🔭 Astronomy: Helping us understand the birth and workings of the universe

                    Remember: These aren't future possibilities - they're happening right now! You probably interact with AI multiple times every day without even realizing it.
                

⚙️ But how does AI actually work? Let's explore the techniques!

Part 5: AI Working - AI problems

Understanding AI Problems by Complexity

AI tackles problems at different levels of complexity. Think of it like a skill pyramid:

                    Mundane Tasks (Bottom Level)
                    
• Recognizing people and objects
                    
• Navigation and robot control
                    
• Game playing
                    
Skills needed: Perception, basic reasoning

Formal Tasks (Middle Level)
• Engineering analysis and design
• Scientific analysis
• Theorem proving
Skills needed: Logical reasoning, strategic planning

                    Expert Tasks (Top Level)
                    
• Medical diagnosis
                    
• Financial analysis
                    
• Complex engineering tasks
                    
Skills needed: Deep expertise, pattern recognition

🔧 Each level requires different AI techniques. Let's learn the three fundamental approaches!

The Three Fundamental AI Techniques

Think of these as the three pillars that support all AI systems:

                    🔍 1. Search
                    
When there's no direct solution path, AI explores different possibilities systematically.
                    
Example: Finding the best move in chess by exploring millions of possibilities

📚 2. Use of Knowledge
AI uses structured information about the world to solve complex problems.
Example: Medical AI uses knowledge about diseases, symptoms, and treatments

                    🎨 3. Abstraction
                    
AI focuses on important features while ignoring irrelevant details.
                    
Example: Recognizing a cat regardless of its color, size, or position

Key Insight: Knowledge is powerful but challenging - it's voluminous, poorly organized, constantly changing, and differs from raw data. AI techniques help us handle this complexity efficiently!

⚙️ Now let's see how we implement these techniques using Production Systems!

Part 6: Implementation

How We Build AI Systems

Production Systems: The Engine of AI

Production systems are like the "brain" of AI programs - they provide artificial intelligence through rules and logic.

A Production System = Rules + Database + Control Strategy + Rule Applier

The Four Components Explained:

📋 Rules (IF-THEN statements): Each rule has a condition - if the condition is met, the rule fires
🗄️ Database/Knowledge Base: Contains all the information relevant to the current problem
🎮 Control Strategy: Decides which rule to apply when multiple rules could fire simultaneously
⚙️ Rule Applier: The computational engine that actually executes the rules

Think of it like a sophisticated decision-making system: "IF this condition exists, THEN take this action" - but with thousands of such rules working together!

Why Production Systems Work So Well

Production systems are popular in AI because they have four amazing characteristics:

                    🎯 1. Simplicity
                    
Each rule uses clear IF-THEN structure. Every sentence is unique and easy to understand.
                    
Benefit: Anyone can read and understand the logic!

🧱 2. Modularity
Knowledge is coded in discrete, independent pieces. Like LEGO blocks!
Benefit: Easy to add, modify, or delete information without side effects.

                    🔧 3. Modifiability
                    
Rules can be modified easily. Start with basic rules, then refine them for specific applications.
                    
Benefit: Systems can evolve and improve over time!

🧠 4. Knowledge-Intensive
Stores pure knowledge in human-readable language, separate from programming logic.
Benefit: Domain experts can contribute knowledge without programming!

The Bottom Line: Production systems make AI development more manageable, understandable, and collaborative!

Understanding Problem Characteristics

Before solving any AI problem, we need to analyze it carefully. It's like choosing the right tool for the right job!

The Seven Key Questions Every AI Engineer Asks:

🧩 Decomposability: Can we break this into smaller, independent sub-problems?
↩️ Reversibility: Can we undo solution steps if they prove unwise?
🔮 Predictability: Is the universe (problem environment) predictable?
✅ Solution Quality: Is a good solution absolute, or do we need to compare all options?
🎯 Solution Type: Is the solution a final state or a path to get there?
📖 Knowledge Role: How important is domain knowledge for solving this?
👥 Human Interaction: Does the task require interaction with people?

                    Understanding these characteristics helps us choose the most appropriate AI method for each specific problem!
                

Summary: Our Journey Through AI

What We've Learned - Step by Step:

                    1. Foundation: Intelligence defines humanity (Homo Sapiens) - creating artificial intelligence is our natural next step.
                

2. Definition: AI = Acting Rationally. We build rational agents that take optimal actions to achieve their goals.

                    3. Current Reality: We have Narrow AI today, aiming for General AI, with Super AI being hypothetical.
                

4. Applications: AI combines human intelligence with computer power across healthcare, finance, transportation, and more.

                    5. Techniques: Search, Knowledge, and Abstraction are the three pillars of AI problem-solving.
                

6. Implementation: Production systems provide a practical way to build AI using rules, databases, and control strategies.

🚀 Next Steps: Now that you understand WHAT AI is and HOW it works, you're ready to start learning specific algorithms and building your own AI systems!