AI Basics: Definitions, History, Fields, Agents & Ethics

What is Artificial Intelligence?

Definition and Core Ideas

Artificial Intelligence (AI) is the study of how to create machines or systems that can perform tasks requiring human-like intelligence, such as reasoning, learning, problem-solving, perception, and decision-making. Unlike traditional computing, AI systems improve with experience and adapt to new data.

Key Characteristics of AI:

• Autonomy: Operates without continuous human intervention.
• Adaptability: Learns and improves from data or feedback.
• Reasoning: Makes logical decisions or predictions.
• Perception: Interprets sensory data (e.g., images, speech).

How AI Differs from Traditional Computing

History of AI: Milestones and Eras

AI has evolved through distinct phases, marked by breakthroughs and setbacks:

Timeline of AI Development

timeline
    title History of AI
    1950 : Turing Test (Alan Turing) - "Can machines think?"
    1956 : Dartmouth Conference - Birth of AI as a field
    1960s : Early AI programs (e.g., ELIZA, GPS)
    1970s : Expert Systems (e.g., MYCIN for medical diagnosis)
    1980s : Knowledge-based systems and decline of "AI Winter"
    1997 : Deep Blue defeats Garry Kasparov (Chess)
    2010s : Big Data + Machine Learning (e.g., Google’s AlphaGo)
    2020s : Generative AI (e.g., ChatGPT, DALL·E)

Key Figures:

• Alan Turing: Proposed the Turing Test (1950) to evaluate machine intelligence.
• John McCarthy: Coined the term "Artificial Intelligence" (1956).
• Marvin Minsky & Seymour Papert: Pioneered early AI research (e.g., frames, neural networks).

Subfields of AI

AI is divided into narrow (weak) AI and general (strong) AI, with specialized branches:

Classification of AI Subfields

mindmap
  root((AI Subfields))
    Narrow AI
      Machine Learning
        Supervised Learning
        Unsupervised Learning
        Reinforcement Learning
      Neural Networks
      Computer Vision
      Natural Language Processing (NLP)
    General AI
      Human-like reasoning
      Consciousness (theoretical)
    Applications
      Robotics
      Expert Systems
      Autonomous Vehicles

Examples of Subfields in Action:

1. Machine Learning (ML):
   • Used by eSewa (Nepal) to predict customer behavior for personalized offers.
   • Google’s recommendation system (YouTube, Gmail) uses collaborative filtering.
2. Natural Language Processing (NLP):
   • WhatsApp’s chatbots (e.g., customer service) understand and generate text.
   • Khalti’s fraud detection analyzes transaction patterns in real-time.
3. Computer Vision:
   • Pathao’s driver-app uses image recognition to verify driver IDs.
   • NTC’s traffic monitoring cameras detect violations via AI.

Intelligent Agents

An intelligent agent is an entity that perceives its environment and acts to achieve goals. Agents can be:

• Simple reflex agents (react to immediate inputs).
• Model-based reflex agents (use internal state to decide actions).
• Goal-based agents (pursue objectives, e.g., a robot vacuum cleaner).
• Learning agents (improve over time, e.g., Netflix recommendations).

How Agents Work: A Worked Example

Scenario: A Khalti payment agent processes a transaction.

1. Perception: Receives input (user ID, amount, merchant details).
2. Decision: Checks fraud rules (e.g., "Is the transaction amount > $1000?").
3. Action: Approves/rejects the payment and updates the ledger.

flowchart LR
    A["User Input"] --> B["Agent Perception"]
    B --> C{"Fraud Check?"}
    C -->|"Yes"| D["Reject"]
    C -->|"No"| E["Approve & Update Ledger"]
    E --> F["Send Confirmation"]

Applications of AI in Nepal and Globally

Real-World Examples

1. eSewa (Nepal)

• Idea Used: Recommendation Systems (ML)
• How: Predicts which bills (electricity, phone) a user is likely to pay next based on past behavior.
• Impact: Reduces user effort by suggesting frequent payments.

2. Pathao (Ride-Hailing App)

• Idea Used: Optimization Algorithms (Search)

• How: Uses A* search to find the fastest route for drivers, balancing distance and traffic.

• Worked Example:

  • State Space: All possible paths from point A to B.
  • Heuristic: Estimated time to destination (ignoring obstacles).
  • Optimal Path: Chosen as the one with the lowest cost (time + fuel).

A search tree for Pathao route optimization (Image: Herbert Glarner, CC BY 3.0, via Wikimedia Commons) ```

3. NEPSE (Stock Market)

• Idea Used: Predictive Analytics (ML)
• How: Uses time-series forecasting to predict stock prices based on historical data.
• Example: If past data shows a 10% drop after a holiday, the system flags potential risks.

4. Ncell’s Customer Service Chatbot

• Idea Used: NLP (Natural Language Processing)
• How: Understands queries like "My bill is wrong" and routes them to the correct department.
• Trace:
  1. User: "Why is my data usage high?"
  2. Chatbot: "Checking your usage... You exceeded 5GB by 2GB. Here’s how to upgrade."
  3. System logs the interaction for future improvements.

Ethical Considerations in AI

AI raises critical ethical questions:

1. Bias and Fairness:
   • Example: Nepal’s civil service exam scoring must avoid bias toward certain regions or demographics.
2. Privacy:
   • Khalti’s data collection must comply with Nepal’s Data Privacy Act (2018).
3. Accountability:
   • Who is responsible if an autonomous vehicle (e.g., self-driving Pathao car) causes an accident?
4. Job Displacement:
   • AI in NTC’s automated toll booths may reduce human jobs but improve efficiency.

AI Ethics Framework

mindmap
  root((AI Ethics))
    Transparency
    Accountability
    Fairness
    Privacy
    Safety
    Sustainability

Exam Tip: How to Score Full Marks

1. Define Clearly: Start with precise definitions (e.g., "AI is the simulation of human intelligence...").
2. Use Diagrams: Draw timelines, flowcharts, or mindmaps for history/subfields.
3. Relate to Nepal: Always link examples to local apps (eSewa, Khalti, NTC) or global tech (Google, WhatsApp).
4. Compare Tables: For differences (e.g., AI vs. traditional computing), use 2-column tables.
5. Worked Examples: Show step-by-step traces (e.g., Khalti’s fraud check or Pathao’s route).
6. Ethics: Mention at least one ethical concern (bias, privacy, accountability) in your answer.

Summary Checklist for Revision

• Can you define AI and distinguish it from traditional computing?
• Do you know the 4 eras of AI and key figures (Turing, McCarthy)?
• Can you list 3 subfields of AI and give a Nepalese example for each?
• Understand intelligent agents and their types (reflex, goal-based).
• Recall 2 real-world AI applications in Nepal and explain their techniques.
• Identify 3 ethical issues in AI deployment.