CSC 3402

Introduction to AI

Department of Computer Science

School of Natural Sciences
University of Zambia
Intro to AI - Overview
• The field of artificial intelligence, or AI, is concerned building intelligent
entities—machines that can compute how to act effectively and safely in a
wide variety of novel situations.

• AI as one of the most interesting and fastest-growing fields, and it is already

generating over a trillion dollars a year in revenue.

• OpenAI is set to hit $2 billion in revenue (12 Feb 2024) - https://qz.com/openai-will-make-2-

billion-revenue-soon-tech-ai-chatgpt-1851247985

• Nvidia - First-quarter revenue was a record $22.6 billion, up 23% from the previous quarter
and up 427% from a year ago (22 May 2024) - https://nvidianews.nvidia.com/news/nvidia-
announces-financial-results-for-first-quarter-fiscal-2025

• AI industry is vast, amounting to around $200 billion in 2023 and is expected to grow well
beyond that to over $1.8 trillion by 2030 - https://www.statista.com/outlook/tmo/artificial-
intelligence/worldwide
Intro to AI - Overview
• AI will impact the word as we know it 30, 20 or even 10 years from now.

• What's more, whereas older sciences such as physics might have the best
ideas already discovered…

• …AI still has much more to explore, much more to discover, much more to
realize.

• AI encompasses a huge variety of subfields, ranging

• from general (learning, reasoning, perception, and so on)

• to specifics, such as playing chess, proving mathematical theorems, writing poetry, driving a
car, or diagnosing diseases.

• AI is relevant to any intellectual task; it is truly a universal field.

Intro to AI - What is AI?
• AI is not Machine Learning – the latter is a sub-field of the former

• Definition of intelligence

1. fidelity to human performance

2. rationality

• Characterization of intelligence

1. as property of internal thought processes and reasoning

2. as externally observable behavior

Intro to AI - What is AI?
• the study of human-like intelligence is part an empirical science related to psychology,
involving observations about human behavior and hypotheses about thought processes;

• the study of rational intelligence involves a combination of mathematics and

engineering, and connects to statistics, control theory, and economics to characterize
behavior and thought.

• Results in four approaches to studying intelligence:

1. Acting humanly: The Turing test approach

2. Thinking humanly: The cognitive modeling approach

3. Thinking rationally: The “laws of thought” approach

4. Acting rationally: The rational agent approach

Intro to AI - Acting Humanly
• Can a machine think?

• philosophical vague of the question

• The Turing Test

• a thought experiment proposed by Alan Turing (1950)

• a computer passes the test if a human, after posing some written questions, cannot
tell whether the written responses come from a person or from a computer.
Intro to AI - Acting Humanly
• Building a computer system to pass the Turing Test would need consideration of
the following:

• natural language processing to communicate successfully in a human language;

• knowledge representation to store what it knows or hears;

• automated reasoning to answer questions and to draw new conclusions;

• machine learning to adapt to new circumstances and to detect and extrapolate patterns.
Intro to AI - Acting Humanly
• Total Turing Test

• This test requires that a computer system interacts with objects and people in the real world.

• Building a computer system / robot to pass the Total Turing Test will need, in
addition to what's needed for the Turing Test, the following

• computer vision and speech recognition to perceive the world;

• robotics to manipulate objects and move about.

• Each of these six criteria used in these tests make up most of the study in
Artificial Intelligence.
Intro to AI - Thinking Humanly
• To say that a program / computer system thinks like a human, we must know how
humans think (human cognition).

• Three ways mainly to learn about human thought:

• Introspection—trying to catch our own thoughts as they occur;

• Psychological Experiments—observing a person in action;

• Brain Imaging—observing the brain in action.

• Once we have a sufficiently precise theory of the mind, it becomes possible to express
the theory as a computer program.

• If the program’s input–output behavior matches corresponding human behavior, that is

evidence that some of the program’s mechanisms could also be operating as humans do.
Intro to AI - Thinking Rationally
• Greek philosopher Aristotle attempted codify “right thinking”— that is, irrefutable
reasoning processes.

• He provided patterns for argument structures that always yielded correct conclusions when
given correct premises.

• Simple logic example

• Premise: Tom is a man

• Premise: All men are mortal

• Conclusion: Tom is mortal

Intro to AI - Thinking Rationally
• Logicians developed this further into a precise notation for statements about
objects in the world and the relations among them.

• Programs can, in principle, solve any solvable problem described in logical notation.

• Artificial intelligence hopes to build on such programs to create intelligent systems.

• Limitations of logic-based reasoning

• Logic requires knowledge of the world that is certain— in reality, it is not entirely possible.

• For example: It is easy to create logic-based rules for a game of chess but not so
straightforward politics.
Intro to AI - Thinking Rationally
• The theory of probability fills this gap

• rigorous reasoning with uncertain information.

• The theory of probability provides for

• the construction of a comprehensive model of rational thought,

• leading from raw perceptual information

• to an understanding of how the world works

• to predictions about the future.

Intro to AI - Acting Rationally
• All computer programs do something, but intelligent agents are computer
programs that are expected to:

1. operate autonomously,

2. perceive their environment,

3. persist over a prolonged time period,

4. adapt to change, and

5. create and pursue goals.

• A rational agent is one that acts so as to achieve the best outcome or, when
there is uncertainty, the best expected outcome.
Intro to AI - Acting Rationally
• The rational-agent approach to AI has two advantages over the other
approaches.

1. it is more general than the “laws of thought” approach because correct inference is
just one of several possible mechanisms for achieving rationality.

2. it is more amenable to scientific development.

• The standard of rationality is mathematically well-defined.

• We can often work backwards to derive agent designs that probably achieve it—
something that is somewhat impossible if the goal is to imitate human behaviour or
thought processes.

• For these reasons, the rational-agent approach to AI is more common.

Intro to AI - Acting Rationally
• Methods based on probability theory and machine learning allow the creation of
agents that can make decisions under uncertainty to attain the best-expected
outcome.

• The Standard Model in AI - construction of agents that do the objective that we

provide to the agent (aka the right thing).

• Perfect Rationality

• always taking the exactly optimal action

• not feasible in complex environments

• computational demands are just too high

• Read on Beneficial Machines in the 4th edition of the prescribed book

Intro to AI - Foundations of AI
Disciplines that contributed ideas, viewpoints, and techniques to AI.

1. Philosophy

2. Mathematics

3. Economics

4. Neuroscience

5. Psychology

6. Computer Engineering

7. Control Theory

8. Linguistic
 (1) Philosophy
• Philosophers like Aristotle developed early logical systems, forming the basis for
reasoning processes in AI aka formal logic

• Philosophical debates on how knowledge is acquired influenced AI approaches to

learning and reasoning with the main themes being Rationalism vs. Empiricism

• Philosophers like Descartes explored the relationship between mind and matter,
shaping discussions on the nature of artificial minds

• Philosophical questions about free will and ethical behavior informed AI decision-
making and autonomy

• The study of knowledge (epistemology) and belief in philosophy directly impacts AI's
understanding of truth, inference, and reasoning

• Debates on whether the mind is purely physical or involves non-physical elements

(Dualism vs. Materialism) influence AI's approach to modeling consciousness
 (2) Mathematics
• Mathematics provided the tools for formalizing logic, essential for reasoning and
deduction in AI

• Early mathematical algorithms laid the groundwork for computational processes used in
AI

• Mathematics introduced probability theory, crucial for decision-making and handling

uncertainty in AI systems

• Mathematical studies on computability (e.g., Turing machines) and complexity (e.g.,

NP-completeness) define what problems AI can solve efficiently

• Mathematical optimization techniques are used in AI to find the best solutions to

complex problems

• Mathematics contributes statistical methods for data analysis, machine learning, and
inference in AI
 (3) Economics
• Economics introduced the concept of agents maximizing their utility (rational agents),
which parallels AI's design of goal-driven systems

• Economics introduced

• Decision Theory which combines probability and utility theory to provide a framework for making
decisions under uncertainty, influencing AI algorithms

• Game Theory which explores strategic interactions between agents, informing AI models on multi-
agent systems and competitive environments

• Economics contributes to AI through Markov decision processes, which model decision-

making over time aka Sequential Decision Making

• Economic principles guide AI in efficiently allocating resources and optimizing

outcomes.

• Insights into human decision-making behaviors (Behavioral Economics) help refine AI

models to better mimic real-world decision processes
 (4) Neuroscience
• Neuroscience studies how the brain processes information, providing a biological model
for AI systems

• Inspired by the structure and function of neurons in the brain (Neural Networks), AI
developed artificial neural networks for learning and pattern recognition

• Research on specific brain regions aka localized brain functions guides the development
of specialized AI models for tasks like speech and vision

• The brain's ability to adapt and reorganize (aka Brain Plasticity) informs AI algorithms on
learning and adapting to new data

• Insights into memory, perception, and decision-making in the brain (Cognitive

Processes) influence AI's approach to replicating human-like intelligence

• Advances in understanding brain activity (e.g., EEG, fMRI) drive AI's development of
systems that interact with or simulate brain functions
 (5) Psychology
• Psychology studies human behavior, cognition, and mental processes, providing insights
into how humans think, learn, and solve problems

• Early psychological theory emphasizing observable behavior (Behaviorism), influenced

early AI models like simple rule-based systems

• Cognitive Psychology shifted focus to understanding internal mental processes, leading

to AI developments in areas like problem-solving, memory, and perception

• Concepts like Reinforcement Learning in psychology have directly influenced AI

techniques for training models

• Insights from psychology help improve AI interfaces, making them more intuitive and
user-friendly – this stems from studies in Human-Machine Interaction

• AI seeks to emulate aspects of human thought or in other words to Model Human

Intelligence, inspired by psychological theories on intelligence and decision-making
 (6) Computer Engineering
• Computer engineering in AI focuses on designing and optimizing the hardware necessary
for AI, enabling faster processing and more efficient data handling

• Advances in computer engineering have led to parallel processing technologies, crucial

for training complex AI models like deep learning networks

• Specialized microprocessor designs, such as GPUs and TPUs, have been developed
to meet the high computational demands of AI applications

• AI is increasingly integrated into embedded systems, from smartphones to autonomous

vehicles, made possible by advancements in computer engineering

• Engineering innovations enable the scaling of AI systems, allowing them to handle large
datasets and complex computations efficiently

• Ongoing research in computer engineering aims to reduce the energy consumption of AI

systems (energy efficiency), balancing performance with sustainability
 (7) Control Theory and Cybernetics
• Control Theory and Cybernetics studies self-regulating systems that use feedback to achieve
stable, goal-directed behavior, using both biological and mechanical processes

• While the concept of self-regulation is foundational for AI's feedback and control systems, the
feedback loops theories of cybernetics provide a mechanism for goal-oriented behavior in AI

• Control theory introduced the idea of minimizing the difference between the current state and
the desired goal state, key in AI optimization and learning algorithms

• W. Ross Ashby’s work on homeostatic systems contributed to AI's understanding of adaptive

behavior and stability in dynamic environments

• Advanced control theory provides mathematical frameworks for designing systems that optimize
performance over time (Stochastic Optimal Control), aligning closely with AI objectives

• To design optimal systems, AI focuses on logical inference and computation

 (8) Linguistics
• Linguistics focuses on the structure, meaning, and context of human language,
providing essential insights for AI in natural language processing (NLP)

• Noam Chomsky introduced transformational grammar, revolutionizing the

understanding of syntax and influencing AI models for language processing.

• Linguistics provides frameworks for understanding meaning (semantics), crucial for

developing AI systems that comprehend and generate meaningful language.

• The study of language in context (pragmatics) informs AI systems on how to interpret

and respond to language in real-world situations.

• Insights from phonetics and phonology in linguistics have driven advancements in AI's
ability to recognize and process spoken language - Speech Recognition

• Linguistics explores the relationship between language and cognition, informing AI

models that simulate human thought processes
 History of AI
• The inception of artificial intelligence (1943–1956): AI began with foundational work in
formal logic, computation, and neural networks, leading to the first AI programs and
the concept of machines simulating human intelligence

• Early enthusiasm, great expectations (1952–1969): Early AI research was marked by

optimism and ambitious goals, including the development of symbolic reasoning and
problem-solving programs

• A dose of reality (1966–1973): The initial excitement was tempered by challenges, as AI

systems struggled with complex problems, leading to a period of slower progress and
more realistic expectations

• Expert systems (1969–1986): AI shifted focus to creating expert systems that encoded
specialized knowledge, achieving commercial success and demonstrating the practical
applications of AI.
 History of AI
• Neural networks (1986–present): Neural networks regained attention as computational
power increased, leading to breakthroughs in learning algorithms and pattern
recognition.

• Probabilistic reasoning and machine learning (1987–present): AI advanced through

the integration of probabilistic models and machine learning, enabling systems to handle
uncertainty and improve with experience.

• Big data (2001–present): The rise of big data provided AI with vast amounts of
information to analyze, significantly enhancing machine learning and decision-making
capabilities.

• Deep learning (2011–present): Deep learning emerged as a powerful technique in AI,

enabling significant advances in image recognition, natural language processing, and
other complex tasks.

• For a detailed account READ “History of AI” in 4th ed. of the prescribed book
State of the Art
Natural Language Processing (NLP)

1. Transformer Models: Advanced models like GPT-4, BERT, and T5 are leading NLP
tasks such as text generation, sentiment analysis, and translation e.g.
https://translate.google.com/?sl=bem&tl=en&text=AI%20yaliwama%20kabili%20ikapilibula%20ubwikashi&op=translate

2. Language Understanding: Models like OpenAI's GPT-4 and Google's LaMDA are
setting new standards in conversational AI

3. Speech Recognition: Systems like Google's Speech-to-Text and Apple's Siri leverage
cutting-edge speech recognition for tasks such as virtual assistants, smart devices,
customer service and support, and hands-free operations

Natural Language Understanding (NLU)

1. Advanced NLU Models: Progress in understanding context, sentiment, and intent

beyond just text processing, with applications in chatbots, and customer service
State of the Art
Computer Vision

1. Object Detection and Recognition: Models like YOLO (You Only Look Once) and
EfficientDet are among the best for real-time object detection

2. Image and Video Analysis: Convolutional Neural Networks (CNNs) and Vision
Transformers (ViTs) excel in tasks like image classification and video understanding

3. Generative Models: GANs (Generative Adversarial Networks) and diffusion models

are leading in image synthesis and style transfer

AI in Agriculture

1. Precision Farming: AI optimizes crop management, pest control, and resource

use, increasing efficiency and sustainability in agriculture
State of the Art
Deep Learning

1. Deep Neural Networks: Used extensively in various AI applications, from image

recognition to game playing

2. Reinforcement Learning: Algorithms like Deep Q-Networks (DQNs) and AlphaGo

have achieved superhuman performance in complex games and decision-making
tasks

3. Transfer Learning: Techniques allowing pre-trained models to be fine-tuned for

specific tasks, significantly improving performance with less data

Meta-Learning (Learning to Learn)

1. Adaptive Learning Algorithms: Meta-learning focuses on creating models that can

adapt to new tasks with minimal data, enhancing AI’s flexibility and efficiency.
State of the Art
Autonomous Systems

1. Self-Driving Cars & Drones: Companies like Tesla and Waymo are at the forefront of
autonomous vehicle technology, while Amazon has rolled out autonomous drones

2. Robotics: Advanced robotics are being developed for tasks ranging from industrial
automation to household chores, leveraging AI for perception and manipulation e.g.
https://www.youtube.com/watch?v=XPVC4IyRTG8

Probabilistic Models

1. Bayesian Networks: Used for reasoning under uncertainty in domains like medical
diagnosis and risk assessment

2. Hidden Markov Models (HMMs): Applied in finance to model and predict time series
data for example in algorithmic stock market trading
State of the Art
Ethics and Fairness in AI

1. Bias Mitigation: Research focuses on detecting and reducing biases in AI systems

to ensure fairness and equity

Recommendation Systems

1. AI-Powered Recommenders: Companies use AI to recommend products (Amazon,

eBay), media (Netflix, Spotify), and more (filter SPAM) based on user behavior

Big Data and AI Integration

1. Data-Driven AI: Leveraging large datasets to improve machine learning models,

particularly in domains like finance, healthcare, and marketing

2. Edge AI: Bringing AI processing closer to data sources (e.g., IoT devices) for faster,
more efficient decision-making
State of the Art
AI in Healthcare

1. Medical Imaging: AI models are revolutionizing radiology by assisting in the detection

and diagnosis of diseases

2. Drug Discovery: AI is accelerating the identification of new drug candidates and

optimizing clinical trials

Quantum AI

1. Quantum Machine Learning: Exploration of quantum computing to enhance machine

learning algorithms, promising faster processing for complex tasks

Federated Learning

1. Decentralized Learning: Enables AI models to be trained across multiple devices

or servers while maintaining data privacy, particularly in healthcare and finance
State of the Art
AI in Creative Arts

1. Generative Art: AI is being used to create music, visual art, and literature, with tools
like DALL-E generating images from text descriptions

2. AI-Assisted Design: AI tools are aiding in fashion design, architecture, and other
creative fields, pushing the boundaries of innovation

Game Playing

1. AI Superiority: AI surpassed human champions in chess (Deep Blue), Go (AlphaGo),

and other games like Jeopardy!, poker, Dota 2, and StarCraft II

2. Self-Learning: ALPHAZERO learned to master games like Go and chess through self-
play without human input
Benefits of AI
• Automation and Efficiency: AI can automate repetitive tasks, leading to increased
productivity and efficiency in industries such as manufacturing, logistics, and customer
service.

• Enhanced Decision-Making: AI systems can analyze large datasets quickly, providing

insights that help in making more informed decisions, particularly in fields like finance,
healthcare, and marketing.

• Healthcare Advancements: AI-powered diagnostic tools can detect diseases at early

stages, improving patient outcomes. AI also aids in drug discovery, personalized
medicine, and optimizing treatment plans.

• Improved Personalization: AI algorithms can tailor products, services, and content to

individual preferences, enhancing user experience in sectors like e-commerce,
streaming services, and social media.
Benefits of AI
• Safety and Risk Management: AI systems can predict and mitigate risks in areas like
cybersecurity, natural disasters, and fraud detection, enhancing safety and security.

• Scientific and Technological Innovation: AI accelerates research in various fields,

including climate science, materials science, and biotechnology, leading to new
discoveries and innovations.

• Accessibility: AI technologies, such as speech recognition and machine translation,

make information and services more accessible to people with disabilities or those
who speak different languages.

• Economic Growth: AI can drive economic growth by creating new industries, jobs,
and opportunities, especially in tech-driven sectors.

• Environmental Monitoring: AI can be used to monitor and address environmental

challenges, such as tracking wildlife populations, optimizing energy use, and predicting
climate change impacts.
Risks of AI
• Job Displacement: Automation driven by AI may lead to job losses in certain sectors,
particularly for routine and manual jobs, leading to economic and social challenges

• Bias and Discrimination: AI systems can perpetuate or amplify biases present in

training data, leading to unfair outcomes in areas like hiring or lending, or applying law

• Privacy Concerns: AI's ability to analyze vast amounts of data raises significant privacy
issues, as personal information can be misused or inadequately protected

• Security Threats: AI can be exploited for malicious purposes, such as developing

more sophisticated cyberattacks, deepfakes, and autonomous weapons

• Loss of Human Control: As AI systems become more autonomous, there is a risk that
humans may lose control over critical decisions, leading to unwanted consequences

• Economic Inequality: The benefits of AI may not be evenly distributed, exacerbating

economic inequality between countries, industries, and individuals
Risks of AI
• Ethical Dilemmas: AI raises complex ethical issues, such as the use of AI in warfare,
the development of sentient machines, and the implications of AI in making life-and-death
decisions

• Concentration of Power: The development and deployment of AI can lead to

increased power concentration among a few large tech companies or
governments, potentially leading to monopolistic practices or authoritarian control

• Dependence on AI: Over-reliance on AI systems may lead to skill degradation in

humans and vulnerabilities in critical systems if AI fails or is compromised

• Unpredictable Outcomes: As AI systems become more complex, their behavior can

become less predictable, leading to unintended and potentially harmful
consequences in critical applications

• Artificial Superintelligence (ASI): If AI surpasses human intelligence it could lead to

loss of human control over.
Possible Mitigations
Governance and Regulation

• AI research communities and corporations have developed self-governance principles for AI

activities

• Governments and international organizations are setting up advisory bodies to create appropriate
regulations

Human-Level AI (HLAI)

• Create machines that think, learn, and create like humans

• The focus on achieving human-level AI rather than Superintelligence (ASI)

Built-in Failsafe

• E.g. build machines that can be switched off voluntarily or learn human preferences through
observation