AI Basics: Definitions, History, Agents, and Applications

What is Artificial Intelligence (AI)?

Artificial Intelligence (AI) is the field of study that focuses on creating systems capable of performing tasks that typically require human intelligence. These tasks include reasoning, learning, problem-solving, perception, and decision-making.

Definitions

AI can be defined in multiple ways depending on the context:

• Turing Test (1950): A machine is considered intelligent if a human cannot distinguish its responses from those of a human in a text-based conversation.
• Weak AI (Narrow AI): AI systems designed to perform specific tasks, such as voice recognition or playing chess. These systems do not possess general intelligence.
• Strong AI (Artificial General Intelligence, AGI): Hypothetical AI that can perform any intellectual task that a human can, including abstract thinking and reasoning.
• Superintelligent AI: AI that surpasses human intelligence in nearly every domain.

Key Characteristics of AI

AI systems exhibit the following traits:

• Autonomy: Ability to operate without continuous human intervention.
• Adaptability: Capacity to learn and adapt to new situations.
• Reasoning: Logical thinking and problem-solving abilities.
• Perception: Understanding and interpreting sensory data (e.g., images, speech).
• Natural Language Processing (NLP): Ability to understand and generate human language.

Historical Evolution of AI

The journey of AI spans several decades, marked by periods of optimism, skepticism, and breakthroughs.

Timeline of AI Development

timeline
    title Historical Evolution of AI
    1943 : McCulloch and Pitts propose the first neural network model.
    1950 : Alan Turing publishes "Computing Machinery and Intelligence," introducing the Turing Test.
    1956 : Dartmouth Conference coins the term "Artificial Intelligence."
    1966 : ELIZA, the first chatbot, is developed by Joseph Weizenbaum.
    1974 : First AI winter begins due to overpromising and underdelivering.
    1980 : Expert systems (e.g., MYCIN) gain popularity.
    1997 : Deep Blue defeats Garry Kasparov in chess.
    2011 : IBM Watson wins Jeopardy!.
    2012 : AlexNet revolutionizes deep learning with convolutional neural networks.
    2020 : AI achieves breakthroughs in natural language processing (e.g., GPT-3).

Key Milestones

• 1950s: Early rule-based systems and the birth of AI as a field.
• 1960s-1970s: Development of expert systems and the first AI winter.
• 1980s-1990s: Revival of AI with advancements in machine learning and neural networks.
• 2000s-Present: Explosion of AI applications, driven by big data, cloud computing, and deep learning.

Types of AI

AI can be categorized based on its capabilities and functionalities.

Classification of AI

mindmap
  root((Types of AI))
    Weak AI
      Narrow AI
      Examples: Siri, Alexa, self-driving cars
    Strong AI
      Artificial General Intelligence (AGI)
      Hypothetical: Human-like reasoning
    Superintelligent AI
      Beyond human intelligence
      Hypothetical: Autonomous decision-making

Applications of AI

AI is used in various domains, including:

• Healthcare: Diagnosing diseases, drug discovery, and personalized medicine.
• Finance: Fraud detection, algorithmic trading, and customer service chatbots.
• Retail: Recommendation systems (e.g., Amazon), inventory management.
• Transportation: Autonomous vehicles, route optimization (e.g., Pathao, Uber).
• Entertainment: Content recommendation (e.g., YouTube, Netflix), virtual assistants.
• Education: Adaptive learning platforms, tutoring systems.

Intelligent Agents

An intelligent agent is an entity that perceives its environment and acts to achieve its goals. Agents can be simple or complex, reactive or proactive.

Components of an Intelligent Agent

classDiagram
    class Agent {
        +Percept: Input from environment
        +Action: Output to environment
        +Goal: Objective to achieve
        +Knowledge: Internal state or memory
        +Architecture: Decision-making process
    }
    class Environment {
        <<environment>>
        +State: Current conditions
        +Actions: Possible interactions
    }
    Agent --> Environment : Perceives
    Environment --> Agent : Acts upon

Types of Agents

In the Real World

AI is transforming industries and everyday life in Nepal and globally. Here are concrete examples:

1. eSewa (Nepal):

   • Idea Used: Natural Language Processing (NLP) and Rule-Based Systems
   • How: eSewa uses AI to process and verify transactions in Nepali and English. For example, when you pay your electricity bill via eSewa, the system automatically validates the customer ID, bill amount, and payment method using predefined rules and machine learning models to detect fraudulent activities.

2. Pathao (Nepal/Global):

   • Idea Used: Intelligent Agents and Optimization Algorithms
   • How: Pathao’s app uses AI to match drivers and riders efficiently. The system acts as a utility-based agent, calculating the best route, estimated time of arrival (ETA), and fare dynamically. It also learns from user behavior to improve future matches (e.g., predicting demand during peak hours in Kathmandu).

3. Ncell (Nepal):

   • Idea Used: Machine Learning for Predictive Maintenance and Customer Service
   • How: Ncell uses AI to predict network failures by analyzing data from cell towers. For example, if a tower in Bhaktapur frequently experiences outages, the AI system flags it for maintenance before it disrupts service. Additionally, Ncell’s customer service chatbot uses NLP to resolve common queries like checking balance or data usage.

4. Daraz (Nepal/Global):

   • Idea Used: Recommendation Systems and Reinforcement Learning
   • How: Daraz’s "Recommended for You" section uses collaborative filtering (a type of machine learning) to suggest products based on your browsing and purchase history. For example, if you frequently buy electronics, the system will prioritize showing you new gadgets or accessories.

5. NEPSE (Nepal Stock Exchange):

   • Idea Used: Time-Series Forecasting and Algorithmic Trading
   • How: AI models analyze historical stock prices and market trends to predict future movements. For instance, an AI agent might recommend buying shares of a company like NMB Bank if its stock price is trending upward based on past data and news sentiment analysis.

Worked Example: AI in Kathmandu Traffic Management

Imagine Kathmandu’s traffic is managed using an intelligent agent that optimizes traffic light timings. Here’s how it works:

1. Perception: Sensors on roads detect the number of vehicles at each intersection (e.g., 50 cars at Trivicharan, 20 at Thapathali).
2. Decision-Making: The agent uses a utility-based approach to maximize the flow of traffic. It prioritizes intersections with higher congestion.
3. Action: Traffic lights adjust dynamically. For example:
   • Trivicharan gets a green light for 45 seconds (longer due to higher traffic).
   • Thapathali gets a green light for 30 seconds.
4. Learning: Over time, the system learns patterns (e.g., rush hours at 8 AM and 6 PM) and adjusts timings automatically.

Visualization of Traffic Agent’s Decision:

graph TD
    A["Sensors Detect Traffic"] --> B["Agent Receives Data"]
    B --> C{"Is Congestion High?"}
    C -->|"Yes"| D["Adjust Timings: Longer Green for Busy Roads"]
    C -->|"No"| E["Maintain Default Timings"]
    D --> F["Traffic Flows Smoothly"]
    E --> F

Exam Tip

For the TU board exam on this unit, focus on the following:

1. Definitions: Be ready to explain AI, intelligent agents, and the Turing Test clearly.
2. History: Know key milestones (e.g., Dartmouth Conference, Deep Blue, Watson) and their significance.
3. Types of AI: Differentiate between Weak AI, Strong AI, and Superintelligent AI with examples.
4. Intelligent Agents: Understand the components (percept, action, goal) and types (reflex, model-based, goal-based).
5. Real-World Applications: Connect theoretical concepts to practical examples like eSewa, Pathao, or Ncell.
6. Diagrams: Practice drawing agent architectures and timelines. Examiners often test conceptual understanding through diagrams.

Common Pitfalls:

• Confusing Weak AI and Strong AI. Always provide examples (e.g., Siri is Weak AI; AGI is hypothetical).
• Forgetting to link theory to real-world applications. Always relate concepts to Nepalese or global products.
• Overcomplicating answers. Stick to the syllabus and use bullet points for clarity.