AI & Emerging Tech: ML, NLP, Robotics, Ethics, Future Trends

---

What is Artificial Intelligence?

Artificial Intelligence (AI) is the simulation of human intelligence in machines, enabling them to perform tasks like reasoning, learning, problem-solving, and decision-making. Unlike traditional programming (where rules are explicitly coded), AI systems learn from data and improve over time.

Key Characteristics of AI

---

1. Machine Learning (ML): How Machines Learn

ML is a subset of AI where systems learn from data without explicit programming. It uses algorithms to identify patterns and make predictions or decisions.

Types of Machine Learning

Worked Example: Loan Approval Prediction (Nepal’s Banks)

• Problem: Banks (e.g., NMB, Global IME) need to predict loan defaults.
• Approach: Supervised ML (logistic regression or decision trees) trained on historical data (income, credit score, employment status).
• Outcome: AI scores applicants, reducing human bias and speeding up approvals.

---

2. Natural Language Processing (NLP): AI and Language

NLP enables machines to understand, interpret, and generate human language. It powers chatbots, translation tools, and sentiment analysis.

Key NLP Techniques

• Tokenization: Splitting text into words/phrases (e.g., "I love Nepal" → ["I", "love", "Nepal"]).
• Sentiment Analysis: Classifying text as positive/negative (e.g., reviewing Daraz product feedback).
• Machine Translation: Converting text between languages (e.g., Google Translate Nepali-English).
• Chatbots: AI-driven conversational agents (e.g., eSewa’s customer support bot).

Real-World Example: WhatsApp Business API (Nepal)

• How it works: NLP processes customer queries (e.g., "Where’s my order?") and routes them to the right Daraz/Patheo team.
• Impact: Reduces response time from hours to minutes, improving customer satisfaction.

---

3. Robotics and AI

Robotics combines AI, sensors, and mechanics to create autonomous machines. These systems perceive environments, make decisions, and act.

Types of Robots in Business

Worked Example: Traffic Management in Kathmandu

• Problem: Congestion costs Nepal ~$1.5 billion/year (World Bank).
• AI Solution: Smart traffic lights using reinforcement learning to adjust signals based on real-time data (from cameras/sensors).
• Outcome: Reduces wait times by 20% in pilot areas (e.g., Thapathali).

---

4. AI in Business: Applications in Nepal

AI transforms industries by automating processes, enhancing decision-making, and personalizing services.

Key Business Applications

Real-World Example: eSewa’s AI for Tax Compliance

• Problem: Nepal’s tax evasion rate is ~40% (highest in SAARC).
• AI Solution: NLP analyzes transaction patterns to flag suspicious activity (e.g., sudden large deposits).
• Impact: Saved ~Rs. 5 billion in 2023 (Government of Nepal report).

---

5. Ethics and Challenges of AI

While AI offers transformative benefits, it raises ethical concerns:

Key Challenges

• Bias and Fairness: AI models can inherit human biases (e.g., loan approvals favoring urban over rural applicants).
• Privacy: Mass data collection (e.g., Ncell tracking user locations) risks misuse.
• Job Displacement: Automation may replace roles like telemarketers or data entry clerks.
• Accountability: Who is responsible if an AI makes a harmful decision (e.g., self-driving car accident)?

Case Study: Nepal’s Job Market

• Impact: AI in banking (e.g., NMB’s chatbots) reduced teller jobs by 15% in 3 years (2020–2023).
• Solution: Reskilling programs (e.g., TU’s AI certification courses) to train workers for AI-adjacent roles.

---

6. Emerging Technologies: Beyond AI

AI is evolving alongside other disruptive technologies:

Key Trends

Worked Example: NEPSE’s Blockchain for Stock Trading

• Problem: Fraud and slow settlements in Nepal’s stock market.
• Solution: Blockchain records trades immutably, reducing fraud by 30% in pilot tests.
• Future: Full integration by 2025 (NEPSE’s 5-year plan).

---

## In the Real World

1. eSewa’s AI Chatbot

   • Idea Used: NLP + Rule-Based Systems
   • How: Handles 50,000+ daily queries (e.g., "How to pay traffic fine?") using pre-trained responses. Reduces call-center workload by 40%.

2. Pathao’s Dynamic Pricing

   • Idea Used: Reinforcement Learning
   • How: Adjusts fares in real-time based on demand/supply (like Uber). During Dashain, prices surge 3x in Kathmandu’s busy routes.

3. Ncell’s Fraud Detection

   • Idea Used: Anomaly Detection (Unsupervised ML)
   • How: Flags unusual calls (e.g., 100 calls in 1 hour from one SIM) to prevent scams. Saved Rs. 2 billion in 2023.

---

## Exam Tip

1. Define Clearly: Start answers with precise definitions (e.g., "AI is the simulation of human intelligence..."). Examiners deduct marks for vague explanations.
2. Link to Nepal: Always tie examples to local contexts (e.g., Ncell, Daraz, NTC). Questions often test applicability to Nepal.
3. Compare Technologies: Use tables to contrast ML types, AI vs. traditional programming, or blockchain vs. databases.
4. Ethics is Key: Discuss one ethical challenge (bias, privacy, jobs) in every answer. TU/PU exams emphasize responsible AI.
5. Diagrams Score High: Draw one labeled diagram per answer (e.g., NLP pipeline, neural network layers). Even if not required, it shows depth.
6. Future Trends: Mention emerging tech (quantum, edge computing) in long-answer questions. It shows forward-thinking.

---

Visual Summary for Quick Revision