Introduction to AI
Artificial and Computational Intelligence
Dr T. J. Alalibo
Department of Computer Engineering
Rivers State University
Introduction to AI & CI
• Artificial Intelligence (AI) and Computational Intelligence (CI) are
often used interchangeably but differ conceptually.
• AI focuses on replicating human intelligence in machines.
• CI emphasizes designing systems that solve problems intelligently,
often inspired by nature or data patterns.
• Both fields aim to create systems capable of reasoning, learning, and
decision-making.
2
What is Artificial Intelligence (AI)?
• AI refers to the ability of machines to perform tasks requiring
human-like intelligence.
• Tasks include problem-solving, learning, language understanding,
and decision-making.
• AI can be embedded in software, robots, or physical devices.
• Modern AI leverages massive data, computational power, and
advanced algorithms.
3
What is Computational Intelligence (CI)?
• CI is both a scientific and engineering discipline focused on
understanding and building intelligent systems.
• It often utilizes techniques inspired by natural phenomena (e.g.,
evolution, swarm behavior).
• CI emphasizes learning and adaptation over hard-coded instructions.
• Common CI approaches include neural networks, fuzzy systems, and
evolutionary algorithms.
4
AI vs CI – Differences and Similarities
Aspect Artificial Intelligence (AI) Computational Intelligence (CI) Similarity
Definition Simulates human intelligence in Develops adaptive algorithms for solving Both aim to replicate aspects of
machines complex problems intelligent behavior
Scope Broad field with many subfields Specialized subfield of AI focused on Both are part of intelligent system
adaptive methods design
Techniques Machine learning, NLP, Neural networks, fuzzy logic, genetic Use data-driven and biologically
computer vision, robotics algorithms, swarm intelligence inspired methods
Learning Approach Requires large labeled datasets; Learns from data continuously; highly Both involve learning from data
less flexible once trained flexible and adaptive
Problem-Solving Task-specific (e.g., image General-purpose, adaptable to many Both solve real-world, often
Strategy recognition, translation) problem types complex, problems
Adaptability Limited adaptability after Designed for continual learning and Both strive to improve system
training adaptation performance based on experience
Rule Dependency Often based on predefined rules Avoids rigid rules; emphasizes learning Both aim to reduce the need for
or structured models and self-organization manual programming
Application Areas Widely used across industries Also used across similar areas, Both have overlapping application
(healthcare, finance, particularly in optimization and evolving domains
automotive, etc.) environments
Inspiration Human cognitive functions Biological and nature-inspired systems Both are inspired by intelligent
(e.g., evolution, swarms) behavior in nature/humans
Interdisciplinary Combines computer science, Combines computer science with biology, Both are highly interdisciplinary
Approach logic, and psychology neuroscience, and mathematics
5
Note that, CI can be considered a subset or complementary approach within AI, particularly focused on
adaptability and learning, but both share the overarching goal of intelligent problem-solving using
computational methods.
6
Steps in Learning Artificial Intelligence
• Theory: Understanding mathematical and conceptual foundations of
intelligence.
• Modeling: Abstracting real-world problems into computational
representations.
• Algorithm: Designing step-by-step methods to solve modeled problems.
• Application: Implementing AI in real-world tasks (e.g., navigation,
healthcare, automation).
7
Importance of Modeling
• Modeling simplifies complex real-world problems into manageable
forms.
• Allows computational tools to simulate and solve real-world
challenges.
• Enhances problem understanding by identifying key components and
relationships.
• Essential in AI system design, enabling effective algorithm
development.
8
Real-World Example – Google Maps
• Google Maps helps users navigate unknown areas using AI principles.
• Streets and intersections are modeled as nodes and edges in a graph.
• Algorithms compute optimal routes based on distance, time, or
traffic data.
• This transformation from geography to computational problem
demonstrates AI's practical power.
9
Google Map contd’
Steps in Modeling Google Maps Problem Algorithms in Route Optimization
• Identify Nodes: Intersections or • Dijkstra's Algorithm: Finds shortest
landmarks in the map. paths in weighted graphs.
• Define Edges: Road segments • A Search:* Combines path cost and
connecting nodes. heuristic estimates for efficiency.
• Assign Weights: Time or distance • Traffic Data Integration: Real-time
associated with each edge. updates adjust routes dynamically.
• Apply Algorithms: Use graph search • Machine Learning: Learns from
methods to find shortest or fastest historical patterns to improve
path. predictions.
10
Traditional AI
Characteristics of Traditional AI
Traditional AI Overview (1950 – 2008)
• Also called Weak AI or Narrow AI. • Operates based on explicitly defined
rules and logic.
• Focuses on solving specific tasks using
pre-programmed rules. • Cannot adapt to new situations without
manual intervention.
• No capacity for creativity or general
problem-solving. • Performs well in structured, predictable
environments.
• Examples: chess-playing programs,
early expert systems, simple chatbots. • Limited to the scope of its programming
and training data.
11
Example of Traditional AI
Some examples of traditional AI are:
• Expert Systems: These systems are designed to emulate the knowledge and
expertise of human experts in a specific field. For example, an expert system
could be used to diagnose diseases, troubleshoot technical problems, or provide
financial advice.
• Decision Trees: These systems are used to make decisions based on a set of
rules. For example, a decision tree could be used to decide whether or not to
approve a loan application or to recommend a product to a customer.
• Natural Language Processing (NLP): These systems are used to understand and
generate human language. For example, in search engines, chatbots, and
machine translation systems.
12
Natural Language Processing (NLP) in Traditional AI
• Early NLP used strict grammar rules for language translation.
• Limited understanding of context and nuance.
• Enabled simple chatbots and machine translation systems.
• Modern NLP has evolved with deep learning for more natural
communication.
13
Real-World Applications of Traditional AI
• Expert Systems: Medical diagnosis tools like MYCIN.
• Decision Trees: Rule-based decision-making (e.g., loan approvals).
• Early NLP: Basic language translation and keyword recognition.
• Spam Filters: Rule-based identification of unwanted emails.
14
Limitations of Traditional AI
• Requires large labeled datasets for training.
• Struggles with ambiguity, uncertainty, or incomplete information.
• Cannot generalize knowledge beyond predefined tasks.
• Performance degrades when faced with unexpected scenarios.
15
Modern AI Introduction (2008+)
• Shift from rule-based to data-driven approaches.
• Heavy reliance on machine learning, particularly deep learning.
• Enabled breakthroughs in image recognition, language processing,
and decision-making.
• Data is fundamental to AI
• AI systems learn patterns from vast amounts of data.
16
Modern AI Models
Probabilistic Models Statistical Models
• Use probability theory to model • Analyze data to identify patterns,
uncertain events. trends, and relationships.
• Assign likelihoods to different • Enable inference and hypothesis
outcomes. testing.
• Common examples: Markov Chains, • Examples: Linear Regression, Time
Bayesian Networks. Series Analysis, Hypothesis Testing.
• Useful for prediction, risk • Foundational for data-driven AI
assessment, and decision-making applications.
under uncertainty.
17
The Rise of Data Science
• The word ‘Data Science’ was coined in 2008 by LinkedIn and
Facebook data scientists.
• Combines statistics, probability, algebra, and calculus.
• Provides the foundation for AI model development.
• Data Science drives insights and predictions in AI applications.
18
Modern AI – ML and DL
• Modern AI relies heavily on
Machine Learning (ML) and
Deep Learning (DL) methods.
• Modern AI learn from data
trained by these methods
• The figure shows subset of AI.
• Note: All ML is AI, but not all
AI is ML.
19
Types of Learning in ML/DL
20
21
AlexNet and the ImageNet Breakthrough
(2012)
• Key success in Modern AI is Convolutional Neural Network (CNN)
architecture.
• AlexNet is a CNN model, which is a deep learning architecture
• AlexNet outperformed competitors in image recognition with
dramatically lower error rates.
• Demonstrated the power of deep learning.
• Sparked widespread adoption of neural networks in AI
22
Role of GPUs in Modern AI
• Graphical Processing Units (GPUs) allow for massive parallel
computations.
• Essential for training deep learning models efficiently.
• Enabled handling of large datasets and complex algorithms.
• Fueled AI research, innovation, and real-world deployment.
23
Generative AI – The Next Frontier (2014+)
• AI that creates new content (text,
images, music, code).
• Learns patterns from vast datasets and
produces original outputs.
• Goes beyond task automation to
content generation.
• Examples: ChatGPT for text, DALL·E for
images, MuseNet for music.
24
Generative AI contd’
Evolution of Generative AI How Generative AI Works
• 1960s: Eliza chatbot – basic
• Input Prompt: Text, image, music, or
conversational AI.
code.
• 2014: Ian Goodfellow's GANs enabled
realistic image/audio generation. • Model Processing: AI analyzes prompt
using trained patterns.
• Recent Years: Transformer models
(e.g., GPT-4) revolutionize language • Content Generation: New output
and content generation. created (e.g., text response, image).
• Generative AI now applies to text, • User Refinement: Feedback fine-tunes
images, audio, and even code style, tone, or detail.
25
How Generative AI Works
26
Generative AI vs Predictive & Conversational AI
• Generative AI: Creates original content such as text, images, audio,
or code.
• Predictive AI: Analyzes historical data to forecast future trends or
events.
• Conversational AI: Powers natural, human-like interactions via
chatbots or virtual assistants.
• All three are complementary but serve distinct purposes in AI
systems.
27
Generative AI Tools – Text, Image, Music, Code and Voice
• Text Tools: GPT (OpenAI), Jasper, AI-Writer generate human-like written content.
• Image Tools: DALL·E, Midjourney, Stable Diffusion produce realistic or artistic images.
• Music Tools: Amper, Dadabots, MuseNet create original music compositions.
• Code Tools: GitHub Copilot, Codex automate code generation and debugging.
• Voice Tools: Descript, Podcast.ai create synthetic voices or realistic speech.
These tools enhance creativity and automate content production. The tools enhance the
following:
• Efficiency: Speeds up software development and media production.
• Accessibility: Makes technical tasks more approachable for non-experts.
28
Applications of Generative AI
• Content Creation: Articles, essays, marketing materials.
• Visual Generation: AI art, design prototypes, deepfakes.
• Music & Audio: AI-composed songs, synthesized voices.
• Code Assistance: Automatic code generation and debugging tools.
29
Generative AI in Everyday Applications
• AI-powered chatbots (e.g., Siri, Alexa) use conversational and
generative AI.
• Tools like Google Translate apply generative models to language
translation.
• AI enhances game development, 3D modeling, and virtual
simulations.
• Content creation tools help with social media, marketing, and
personalized experiences.
30
Benefits and Limitations of Generative AI
Benefits Limitations
• Productivity: Automates tedious • Can produce misleading or incorrect
content creation tasks. information (hallucinations).
• Creativity: Assists designers, artists, and • Lacks transparency in identifying
musicians in idea generation. information sources.
• Accessibility: Democratizes content • Generated content may contain
creation for non-experts. biases from training data.
• Innovation: Accelerates product design, • Difficult to control outputs in highly
prototyping, and media development sensitive applications (e.g., legal or
medical).
31
The Future of AI Beyond 2023
• AI impacting industries: healthcare, finance, manufacturing,
education, and more.
• Neuromorphic processing mimics the human brain for efficient AI
computations.
• Focus on AI ethics: fairness, accountability, privacy, and trust.
• Increased adoption of AI as a standard component in business and
daily life.
32
