Recap of Lecture 1 and Homework Check
LAST LECTURE · 5 MIN
Intelligence
AI
Four approaches
Learn, reason, solve problems, perceive, use language - and adapt to new situations.
Machines that do tasks which normally need human intelligence. Father of AI: John McCarthy (1956).
Think/act × human/rational. Our course follows acting rationally (rational agents).
Today's Plan (90 minutes)
ROADMAP
0-5
Recap of Lecture 1 and homework check
S02
5-8
Plan and learning outcomes
S03-S04
8-30
Module 1: Foundations and history of AI (1943 - today)
S05-S10
30-50
Module 2: Applications of AI, AI vs robotics, advantages
S11-S15
50-62
Module 3: Types of AI (capability and functionality)
S16-S18
62-75
Module 4: Risks, ethics and responsible AI
S19-S22
75-84
Class activity: type and ethical issue
S23
84-90
Mistakes, recap quiz, homework
S24-S26
Learning Outcomes (SLOs)
BY THE END OF THIS LECTURE YOU WILL BE ABLE TO
SLO 2.1
Describe the foundations of AI and the main milestones of its history, including the AI winters
→ CLO1 · C1 Remember
SLO 2.2
Explain applications of AI in different sectors, with local examples
→ CLO1 · C2 Understand
SLO 2.3
Differentiate between AI and robotics, and list advantages of AI
→ CLO1 · C2 Understand
SLO 2.4
Classify AI systems by capability (narrow, general, super) and functionality
→ CLO1 · C2 Understand
SLO 2.5
Explain the risks and ethical issues of AI and principles of responsible AI
→ CLO1 · C2 Understand
- Course Learning Outcome(s) served
- CLO1: Explain the applications of AI and agent-based approach to AI. (Ethics also supports PLO9.)
Slide 5
MODULE 1
Foundations and History of AI
Which subjects built AI, and how did it grow from 1943 to today?
Foundations of AI: Subjects That Built It
MODULE 1 · FOUNDATIONS
Philosophy
Mathematics
Economics
Neuroscience
Logic, reasoning, mind. Can thinking follow rules?
Logic, probability, what can be computed
How to choose the best decision (utility = value)
How neurons in the brain work → neural networks
Psychology
Computer eng.
Linguistics
Control theory
How humans think, learn and remember
Fast hardware (CPUs, GPUs) to run AI
Grammar and meaning of language → NLP
Self-correcting machines, like an AC thermostat
Key idea: AI is a mix of many subjects. That is why it needs maths, logic and programming together.
History 1: The Birth of AI (1943 - 1966)
MODULE 1 · HISTORY
1943
1950
1956
1958
1966
- Artificial neuron
- McCulloch and Pitts make the first maths model of a brain neuron.
- Can machines think?
- Alan Turing's paper proposes the Turing Test.
- Birth of AI
- Dartmouth workshop. McCarthy coins the term 'Artificial Intelligence'.
- LISP and Perceptron
- McCarthy creates LISP; Rosenblatt builds a learning machine (Perceptron).
- ELIZA chatbot
- Weizenbaum's program chats like a counsellor using simple patterns.
Mood: Great hope. Researchers said a machine as smart as a human would come within 20 years.
History 2: AI Winters and Expert Systems (1969 - 1993)
MODULE 1 · HISTORY
1969
1974
1980s
1987
1990s
- Limits shown
- Minsky and Papert: a simple perceptron cannot learn XOR.
- 1st AI winter
- Lighthill report (1973): funding cut. Lasts to about 1980.
- Expert systems
- IF-THEN rule programs: MYCIN (medicine), XCON (computer orders).
- 2nd AI winter
- Expert systems too costly and could not learn. Lasts to about 1993.
- Quiet return
- Probability and machine learning grow slowly in the background.
AI winter: a period when people lost trust in AI and research money (funding) was cut.
History 3: The Modern AI Boom (1997 - today)
MODULE 1 · HISTORY
1997
2012
2016
2017
2022 +
- Deep Blue
- IBM's computer beats world chess champion Kasparov.
- Deep learning
- AlexNet wins an image contest using GPUs. The boom starts.
- AlphaGo
- DeepMind's program beats Go champion Lee Sedol.
- Transformer
- Google's new neural network design - base of ChatGPT.
- Generative AI
- ChatGPT (Nov 2022), then Gemini, Claude, Meta AI and AI agents.
Also: 2011 - IBM Watson wins Jeopardy! and Apple launches Siri, the first popular voice assistant.
Why Did AI Fail Before and Succeed Now?
MODULE 1 · HISTORY
1. Big data
2. Computing power
3. Better algorithms
Billions of photos, videos, posts and web pages to learn from. In 1970 there was almost no digital data.
GPUs (graphics cards, first made for games) do thousands of calculations at the same time.
Deep learning (2012) and Transformers (2017) learn far better than old methods.
Analogy: a fast bowler needs fitness, a good ball and technique. Old AI had good ideas but no data and no power.
Lesson: over-promising leads to winters. Judge AI claims carefully, even today.
Slide 11
MODULE 2
Applications of AI
Where is AI used in Pakistan and the world?
AI Applications by Sector
MODULE 2 · APPLICATIONS
Healthcare
Agriculture
Banking
AI reads chest X-rays for TB signs; predicts dengue outbreaks from weather and case data.
Phone apps detect cotton or wheat disease from a leaf photo; smart irrigation advice.
Fraud alerts and chatbots in banks and wallets like JazzCash and Easypaisa.
Transport
Government
Education
Careem and inDrive match drivers and set fares; Google Maps predicts traffic.
NADRA biometric checks; Safe City cameras read number plates for e-challans.
AI tutors (ChatGPT, Gemini), automatic grading and plagiarism checks.
Three Applications Up Close
MODULE 2 · APPLICATIONS
Pattern: every application = INPUT (percepts) → AI → OUTPUT (action). This is the agent idea from Lecture 1.
AI vs Robotics: Not the Same Thing
MODULE 2 · APPLICATIONS
Both together
Intelligent robots: a robot vacuum, a self-driving car, a drone that sprays only the sick plants in a field. AI is the brain, robotics is the body.
Remember: AI = brain (software). Robot = body (hardware). Not every robot is intelligent.
Advantages of AI
MODULE 2 · APPLICATIONS
Fewer errors
Risky jobs
Available 24×7
Decisions from data reduce mistakes, e.g. more accurate weather forecasts.
Robots defuse bombs, work in mines and explore the deep sea or Mars.
A bank chatbot answers your question at 2 a.m. without getting tired.
Faster decisions
Repetitive work
Expert help
Fraud checks in milliseconds; a chess engine picks the best move in seconds.
Reading thousands of forms or sorting e-mails, so humans do creative work.
A farmer far from a crop expert can still get disease advice on his phone.
Types of AI by Capability
MODULE 3 · TYPES OF AI
Remember: every AI system you have ever used is Narrow AI - even very clever chatbots.
Types of AI by Functionality
MODULE 3 · TYPES OF AI
Reactive machines
Limited memory
Theory of mind
Self-aware
- React to the present only. No memory.
- Example: IBM Deep Blue chess (1997).
- Status: exists.
- Use recent past data for a short time.
- Example: self-driving car, chatbot remembering this chat.
- Status: most AI today.
- Would understand people's feelings, beliefs and intentions.
- Status: research only.
- Would have its own consciousness (awareness of itself).
- Status: imagined only.
Link: types 1 and 2 are Narrow AI. Types 3 and 4 would need General or Super AI.
Quick Check: Name the Type
MODULE 3 · TYPES OF AI · 2 MIN
Hint: ask 'does it use memory of the past?' and 'can it do only one job?'
Slide 19
MODULE 4
Risks, Ethics and Responsible AI
AI is powerful. What can go wrong, and who is responsible?
Disadvantages and Controversies
MODULE 4 · RISKS
High cost
Job changes
Over-dependence
Large models need huge computers, electricity and experts.
Routine jobs (data entry, call centres) shrink; new AI jobs appear.
Copying AI answers stops students from thinking and learning.
No real emotions
Hallucination
Energy use
AI has no feelings and limited true creativity; it mixes old patterns.
Chatbots can state false facts confidently. Always check the source.
Data centres use a lot of electricity and water for cooling.
Ethical Issues in AI
MODULE 4 · ETHICS
Ethics: the principles of right and wrong. Engineers must think about them while building AI (PLO9).
Responsible AI: Principles and Rules
MODULE 4 · ETHICS
Fairness
Transparency
Privacy
Treat all groups equally; test for bias.
Explain how and why the AI decides.
Collect only needed data, with permission.
Safety
Accountability
Human control
Test well; stop harm before release.
A human or company is responsible.
People can check and override AI.
- Laws are coming: the EU AI Act (2024) controls high-risk AI; Pakistan's cabinet approved a National AI Policy in 2025.
- Your part as a student: do not submit AI text as your own work, check AI facts, and never upload others' private data.
Activity: Type, Sector and Ethical Issue
CLASS ACTIVITY · PAIRS · 9 MIN
In pairs, for each case write: (i) sector, (ii) Narrow or General AI, (iii) one main ethical risk.
A Safe City camera reads car number plates and sends e-challans
A student submits a ChatGPT-written assignment as his own
A bank's AI decides who gets a loan from mobile phone data
A video shows a politician saying words he never said
A crop app tells a Burewala farmer which pesticide to spray from a leaf photo
A hospital AI reads X-rays and marks possible TB cases
Common Mistakes
WATCH OUT
WRONG IDEA
CORRECT IDEA
✗ AI started with ChatGPT in 2022.
✓ AI started in 1956 (Dartmouth). ChatGPT is a recent result of 70 years of work.
✗ ChatGPT is General AI (strong AI).
✓ ChatGPT is Narrow AI. General AI does not exist yet.
✗ Every robot is AI.
✓ A robot that repeats one fixed movement is not AI. AI is the brain, robotics the body.
✗ AI winter means AI was used in winter.
✓ AI winter = a period when trust and research funding for AI were cut.
✗ AI is neutral, so it cannot be unfair.
✓ AI learns from human data and can copy human bias.
Recap Quiz
CHECK YOUR UNDERSTANDING · 5 MIN
- Q1. The 1956 workshop called the birth of AI was held at:
- (a) MIT (b) Dartmouth (c) Stanford (d) Cambridge
- Q2. A period when AI funding and trust fell is called:
- (a) AI spring (b) AI winter (c) AI boom (d) AI freeze-out
- Q3. IBM Deep Blue (1997) is an example of which type?
- (a) Self-aware (b) Theory of mind (c) Reactive machine (d) Super AI
- Q4. All AI systems in use today are:
- (a) Narrow AI (b) General AI (c) Super AI (d) Self-aware AI
- Q5. AI copying unfair patterns from its training data is called:
- (a) Hallucination (b) Bias (c) Deepfake (d) Transparency
- Q6. Which is NOT a reason for today's AI boom?
- (a) Big data (b) GPUs (c) Better algorithms (d) Fewer computers
Homework and Next Lecture
BEFORE THE NEXT CLASS
- Task 1: Make a timeline of 8 AI milestones (1943 - today) with one line on each. Mark the two AI winters.
- Task 2: Choose one Pakistani AI application. Write (half page): sector, input → AI → output, type of AI, one advantage and one ethical risk.
- Read: Kaplan, Ch 1 and 7; Govers, Ch 1.
- Reminder: Assignment 1 (Gemini AI tool report) is due on 06-10-2026.
NEXT LECTURE
- Lecture 3:
- Turing Machine and the Turing Test