Skip to content

CO3519 Artificial Intelligence
CO3519 Lecture 1 - Introduction To Artificial Intelligence

Lecture Documents

CO3519 Lecture 2.pdf

Written Notes

CO3519 Lecture 2 - Note 1.png
CO3519 Lecture 2 - Note 2.png

Lecture Contents

  • Types of AI
  • Based on Capacity
  • Based on Functionality

  • Based on Applications

  • Working Areas

  • Facial Expression Recognition

  • Object Detection

  • Assessment 1 Released (Guide)

Artificial Intelligence

graph LR
%% --- Title ---
%% Artificial Intelligence Classification Overview

%% --- Capability Classification ---
    subgraph Capability
        C1["Weak AI"]
        C2["Strong AI"]
        C3["Super AI"]
    end

%% --- Functionality Classification ---
    subgraph Functionality
        F1["Reactive Machines"]
        F2["Limited Memory AI"]
        F3["Theory of Mind AI"]
        F4["Self-Aware AI"]
    end

%% --- Real-World Applications ---
    subgraph "Real-World Applications"
        R1["Narrow Intelligence"]
        R2["General Intelligence"]
        R3["Super Intelligence"]
    end

%% --- Technologies ---
    subgraph Technologies
        T1["Machine Learning"]
        T2["Deep Learning"]
        T3["Natural Language Processing"]
        T4["Computer Vision"]
        T5["Robotics"]
    end

%% --- Connections (Conceptual flow) ---
    C1 --> F1
    C1 --> F2
    C2 --> F3
    C3 --> F4

    C1 --> R1
    C2 --> R2
    C3 --> R3

    R1 --> T1
    R1 --> T2
    R1 --> T3
    R1 --> T4
    R1 --> T5

    R2 --> T1
    R2 --> T2
    R2 --> T3
    R2 --> T4
    R2 --> T5

    R3 --> T1
    R3 --> T2
    R3 --> T3
    R3 --> T4
    R3 --> T5

Capacity

Weak AI

  • Trained and designed for a specific task or a narrow range of tasks
  • Can't perform beyond its defined tasks

Example

Voice Assistants (Siri, Alexa, Google)
Recommendation Systems (Netflix, YouTube, Spotify)

Strong AI

  • Can perform any intellectual task that a human can.
  • Ability to learn, reason, and understand across various domains.
  • Not restricted to any specific task.

Imagine

Imagine an AI system that can not only assist with voice commands like Siri but:
- write essays
- solve complex math equations
- understand human emotions
- drive a car
- create art

Example

This would be an AI that could perform any task you ask of it, similar to a human with broad knowledge and abilities.

Super AI

  • Far exceeds human intelligence in all respects.
  • AI could learn, adapt and evolve much faster than humans.
  • Solving problem in innovative ways

Example

The Matrix or Ex Machina
- AI Depictions not only outperform human intelligence but also gain consciousness and potentially control the human environment.
- A Super AI might for example cure all diseases or design technologies far beyond current human capacities but it might also pose ethical risks if it behaves unpredictability.
- AI can categorise into several types based:
- Functionality
- Capacity
- Real-World Applications
- Other Technologies.

Functionality

Reactive Machines

  • Simplest AI that perform specific tasks by reacting to current conditions without storing any past experiences.
  • Cannot 'learn' from the previous actions or improve their performance based on experience.

IBM's Deep Blue (Chess-Playing AI)

  • Famously beat the chess champion Garry Kasparov is a reactive AI.
  • It analysed the chessboard's current state, evaluated possible moves, and chose the best one based on programmed algorithms.

The Problem?

It couldn't learn or adapt between games and was purely reactive to the current game's conditions.

Limited Memory AI

  • Can learn from historical data to make decisions.
  • Can store past experiences or data for a short time and use it to improve their predictions or performance.

Example

Self-Driving Cars (e.g. Tesla or Waymo)
- Use limited memory AI
- Analyse real-time data from sensors, cameras, and radar (like road conditions, traffic, and other vehicles' behaviour) to make decisions.
- Use previous experiences (like patterns of traffic flow) to navigate better in future.

Theory of Mind AI

  • Designed to understand emotions, intentions and social situations.
  • To interpret human needs and respond accordingly.
  • To predict how humans will behave in certain situations and adjust its behaviour accordingly.

Currently Theoretical

  • In development however remains theoretical for now.
  • Research into emotional AI is underway.
  • No current AI system fully understands human emotions.

If Developed

  • Will be able to detect sad, or stressed based on facial expressions, voice tone, or physiological data.
  • Responds accordingly offering calm music if stressed or providing motivational content if down.

Self-Aware AI

  • Exists only in theory
  • Own consciousness and self-awareness
  • Understand its own state of existence
  • Recognise it's own identity
  • Have independent desires or motivations

Example

  • In films like Ex Machina AI systems become self-aware developing independent thoughts, emotions and even desires.
  • Systems start to question their purpose and existence. (Much like humans)

Types of AI Technologies

  • Machine Learning
  • Deep Learning
  • Natural Language Processing
  • Computer Vision
  • Robotics

Facial Expression Recognition

CO3519 Lecture 2 - Facial Expression Recognition.png
Human communication consists of two types:D
- Verbal
- Non-Verbal cues

Non-verbal is through facial expression, eye contact, and subconscious head movements.

Feature Extraction

CO3519 Lecture 2 - FER Feature Extraction.png

Object Detection

Object Detection in 2024: The Definitive Guide - viso.ai

Assignment 1 Released (Guides)

  • Released - 06 October 2025 10:00 GMT
  • Located CO3519 Artificial Intelligence (2025 - 2026 Semester 1 and 2) blackboard.
  • Assessment folder contains:
  • CO3519 Assignment 1 Brief 2025-2026
  • Assignment 1 Submission Link - 14th December

Next Week

  • Machine Learning & Key Concepts
  • Data, Types of Data
  • Supervised Learning
  • Key Supervised Algorithms

CO3519 Lecture 3 - Machine Learning I