COURSE : ARTIFICIAL INTELLIGENCE (AI)

BCSE306L (L-3,T-0,P-0,C-3)

Instructor : Dr. Tahir Mujtaba (53598)

Assistant Professor,
SCOPE, VIT Chennai
[email protected]
Cabin 5, Annexure Ground Floor AB IIIrd

1
COURSE
OUTLINE
 DEFINITION OF AI ?
Intelligence: The ability to learn and solve problems”

–Webster’s Dictionary.

“Artificial intelligence (AI) is the intelligence exhibited by machines

or software’

–Wikipedia

“The science and engineering of making intelligent machines”

–McCarthy

“The study and design of intelligent agents, where an intelligent

agent is a system that perceives its environment and takes actions
that maximize its chances of success.”

– Russel and Norvig AI book.

AI APPROACHES/ VIEW OF
AI ?
 AI APPROACHES /VIEWS OF AI
(contd)

Four main approaches to AI:

1. Systems that think like humans: These systems aim to replicate human
cognitive functions, such as reasoning, problem-solving, and learning. This
approach involves modeling human thought processes.
2. Systems that act like humans: These systems are designed to mimic human
behavior. The Turing Test, proposed by Alan Turing, is a classic example of
evaluating whether a machine's behavior is indistinguishable from that of a
human.
3. Systems that think rationally: These systems use logic to emulate human
reasoning. They focus on creating algorithms that reason correctly and make
logical inferences.
4. Systems that act rationally: These systems aim to achieve the best outcome
or perform actions that are expected to maximize the performance measure.
This approach is often associated with intelligent agents that perceive their
environment and take actions to achieve specific goals.
 THINKING HUMANLY (THE COGNITIVE MODELLING
APPROACH)

• This approach aims to create systems that think in ways similar to

human beings. It involves cognitive modeling, understanding how
humans think, and replicating those processes in machines.
• The "Thinking Humanly" approach to AI focuses on creating systems that
simulate human thought processes. The goal is to understand and
replicate how humans think, reason, and solve problems. This approach
is deeply rooted in cognitive science and psychology.

• For Example :
• Neural Networks
• Virtual Assistants: AI systems like Siri or Alexa can understand and
respond to natural language commands by mimicking human
language processing.
• Educational Tools: AI tutors can adapt their teaching methods based
on how students learn, providing personalized learning experiences.
THINKING HUMANLY (CONTD)

To understand the exact process of how we think, we need to go inside

the human mind to see how this giant machine works.
 THINKING HUMANLY (CONTD)

Introspection method – Catch our thoughts and see how it flows.

Psychological Inspections method – Observe a person on the action.
Brain Imaging method (MRI (Magnetic resonance imaging) or fMRI
(Functional Magnetic resonance imaging) scanning) – Observe a
person’s brain in action.
Using the above methods, if we are able to catch the human brain’s
actions and give it as a theory, then we can convert that theory
into a computer program.

If the input/output of the computer program matches with human

behavior, then it may be possible that a part of the program may
be behaving like a human brain.

The General Problem Solver (GPS) programme was created by Allen

Newell and Herbert Simon to simulate human thought and test
whether it can answer issues by using the same steps in reasoning
as a person. The program’s goal is to solve the problem using the
same process that a human brain would use, not only to solve it
 ACTING HUMANLY (THE TURING TEST APPROACH )
• This approach focuses on creating systems that behave like humans. It
involves passing the Turing Test, where a machine's ability to exhibit
intelligent behavior is indistinguishable from that of a human.
• Example: A chatbot designed to pass the Turing Test, where it
converses with humans in natural language and aims to be
indistinguishable from a human. Examples include advanced
chatbots like OpenAI's GPT-3 and GPT-4, which can engage in
realistic and meaningful conversations with users.

The Turing Test Approach

• It is proposed by Alan Turing (1950),
was designed to provide a satisfactory
operational definition of intelligence.
• A computer passes the test if a human
interrogator, after posing some written
questions, cannot tell whether the written
responses come from a person or from a
computer.
ACTING HUMANLY (THE TURING TEST APPROACH )

• According to the Turing test the computer would need to possess the
following capabilities

o Natural language processing : to enable it to communicate

successfully in English
o Knowledge representation : to store what it knows or hears;
o Automated reasoning : to use the stored information to answer
questions and to draw new conclusions;
o Machine learning to adapt to new circumstances and to detect
and extrapolate patterns.
o However, the so-called total Turing TOTAL TURING TEST includes
a video signal so that the interrogator can test the subject’s
perceptual abilities, as well as the opportunity for the
interrogator to pass physical objects “through the hatch.” To
pass the total Turing Test, the computer will also need:
o COMPUTER VISION • computer vision to perceive objects,
and
o ROBOTICS • robotics to manipulate objects and move
about.
THINKING RATIONALY (LAWS OF THOUGHT APPROACH)
This definition is based on the laws of thought, specifically logic. It involves
creating systems that use logical reasoning to solve problems and make decisions.

• Example: An AI that helps with course scheduling by using logical rules. For
instance, the system can be programmed with rules like "If a student has taken
Calculus I, they can enroll in Calculus II," and it uses these rules to help students
plan their classes.

• More Examples:

“Socrates is a man; all men are mortal; therefore, Socrates is mortal.”

"All TVs use energy; Energy generates heat; ∴ all TVs generate heat.”
"All dogs are animals; all animals have four legs; ∴ all dogs have four legs".

• These arguments (like above) initiated the field called logic. Notations for
statements for all kinds of objects were developed and interrelated between them
to show logic. By 1965, programs existed that could solve problems that were
described in logical notation and provides a solution. In order to develop
intelligence systems, programs, or computational models, AI aspires to expand
on such Programs.
ACTING RATIONALLY (THE RATIONAL AGENT
APPROACH)

• This definition emphasizes rational behavior, aiming to create

systems that act to achieve the best outcome or expected utility.
Rational agents take actions to maximize their performance measure
given the available information.

• Example: An autonomous vehicle designed to navigate and drive

safely by making rational decisions based on its sensors and
environment. The vehicle uses algorithms to maximize safety and
efficiency, such as avoiding obstacles, obeying traffic laws, and
optimizing routes.
• Another example is a chess-playing AI like Deep Blue, which acts
rationally to make the best possible moves to win the game.
AI
SUBFIELDS
 AI SUBFIELDS (CONTD)

1. Natural Language Processing (NLP)

NLP focuses on the interaction between computers and humans through
natural language. It encompasses various tasks and applications that enable
machines to understand, interpret, and respond to human language.
a. Text Generation
The creation of coherent and contextually relevant text by machines,
often using techniques like neural networks and transformers.
b. Text Classification
Assigning predefined categories or labels to text documents based on
their content, used in applications like spam detection and sentiment analysis.
c. Question Answering
Systems that automatically answer questions posed by humans in natural
language, often using large databases or pre-trained models to find relevant
information.
d. Machine Translation
The automatic translation of text from one language to another, using
statistical, rule-based, or neural network-based approaches.
e. Context Extraction
Identifying and extracting relevant context or meaning from text, such as
named entity recognition (NER) and sentiment analysis.
 AI SUBFIELDS (CONTD)

2. Planning
Planning involves formulating and executing strategies or actions to
achieve specific goals, often in dynamic and uncertain environments.
a. Fuzzy Logic
A form of many-valued logic dealing with reasoning that is
approximate rather than fixed and exact, used to handle the concept of
partial truth in planning and decision-making processes.
3. Computer Vision
Computer vision enables machines to interpret and make
decisions based on visual data from the world, such as images
and videos.
a. Object Detection
Identifying and locating objects within an image or video, often
using techniques like convolutional neural networks (CNNs).
b. Image Recognition/Classification
Classifying images into predefined categories based on their
content, such as identifying objects, scenes, or activities.
 AI SUBFIELDS (CONTD)

4. Speech Recognition
Speech recognition technology enables machines to
recognize and process human speech, converting
spoken language into text and vice versa.
a. Text to Speech
The conversion of written text into spoken words using
synthetic speech generation technologies.
b. Speech to Text
The conversion of spoken language into written text, using
algorithms and models to transcribe audio input accurately.
5. Robotics
Robotics involves the design, construction, operation, and
use of robots, which are capable of performing tasks in the
physical world. This includes areas like autonomous
navigation, manipulation, and interaction with humans and
environments.
 AI SUBFIELDS (CONTD)

6. Expert Systems
Expert systems are AI programs that mimic the decision-
making abilities of a human expert. They use rule-based systems to
solve specific problems within a particular domain, providing
solutions or recommendations based on a set of predefined rules
and knowledge.
7. Machine Learning (ML)
Machine learning is the study of algorithms and statistical
models that enable computers to perform tasks without explicit
instructions, by learning from data and making predictions or
decisions.
a. Deep Learning
A subset of machine learning involving neural networks with
many layers (deep neural networks), capable of learning complex
patterns and representations from large datasets. Deep learning
is widely used in tasks like image and speech recognition,
natural language processing, and autonomous driving.
THE BUZZ
WORDS
APPLICATIONS OF
AI
 APPLICATIONS OF AI (CONTD)

1. Healthcare
a. Medical Diagnosis
AI systems analyze medical data to diagnose diseases and
conditions, often with high accuracy and speed.
b. Drug Discovery
AI accelerates the discovery of new drugs by predicting how different
compounds will interact with targets in the body.
2. Finance
a. Fraud Detection
AI algorithms detect unusual patterns and behaviors to identify and
prevent fraudulent activities in financial transactions.
b. Algorithmic Trading
AI models analyze market data and execute trades at optimal times,
often faster and more accurately than human traders.
 APPLICATIONS OF
AI
3. Transportation

a. Autonomous Vehicles
AI enables vehicles to navigate and operate without human
intervention, improving safety and efficiency.
b. Logistics
AI optimizes supply chain management, including routing,
warehousing, and delivery processes.

4. Customer Service

a. Chatbots
AI-powered chatbots provide instant responses and support to
customers, handling inquiries and issues efficiently.
b. Personalized Recommendations
AI analyzes user data to offer tailored product or service
suggestions, enhancing the customer experience.
 APPLICATIONS OF
AI
5. Education

a. Personalized Learning
AI systems adapt educational content to the individual
needs and pace of each student, improving learning
outcomes.
b. Automated Grading
AI tools grade assignments and exams quickly and
accurately, freeing up time for educators.

6. Entertainment

a. Content Creation
AI generates new content, such as stories, music, and art,
often indistinguishable from human-created works.
b. Recommendation Systems
AI recommends movies, music, and other media based on
user preferences and viewing habits.
 APPLICATIONS OF
AI
7. Manufacturing

a. Predictive Maintenance
AI predicts equipment failures before they occur, allowing for
timely maintenance and reducing downtime.
b. Quality Control
AI systems inspect products for defects during manufacturing,
ensuring high quality and consistency.

8. Retail

a. Inventory Management
AI optimizes stock levels, reducing excess inventory and
preventing stockouts.
b. Customer Insights
AI analyzes consumer data to understand purchasing behavior
and preferences, informing marketing strategies.
 APPLICATIONS OF
AI
9. Security

a. Surveillance
AI analyzes video footage in real-time to detect suspicious activities and
enhance public safety.
b. Cybersecurity
AI identifies and mitigates cyber threats by analyzing network traffic and
identifying anomalies.

10. Agriculture

a. Precision Farming
AI optimizes crop management by analyzing data on soil conditions, weather,
and plant health.
b. Automated Harvesting
AI-powered machinery automates the harvesting process, improving efficiency
and reducing labor costs.

11. Robotics

a. Industrial Robots
Robots perform repetitive and hazardous tasks in manufacturing, improving
efficiency and safety.
 APPLICATIONS OF
AI
12. Generative AI

a. Art and Design

AI creates original artworks and design concepts, pushing the boundaries
of creative expression.
b. Text Generation
AI produces human-like text for applications such as writing articles,
stories, and social media posts.
c. Music Composition
AI composes original music pieces, often collaborating with human
musicians.
d. Synthetic Media
AI generates realistic images, videos, and audio, often used in media
production and entertainment.
 CLASSIFICATION/TYPES OF AI
Artificial Intelligence is all about building intelligent machines from
vast volumes of data and getting them to perform human-like tasks,
leading to an enhancement in speed, precision, and effectiveness of
human efforts.
CLASSIFICATION OF AI (CONTD)

Based on Capabilities - Narrow AI :

 It also called as Weak AI, focuses on one narrow task and cannot
perform beyond its limitations.
 It targets a single subset of cognitive abilities and advances
in that spectrum.
 Narrow AI applications are becoming increasingly common in
our day-to-day lives as machine learning and deep learning
methods continue to develop.
 Examples : Apple Siri, Alexa, IBM Watson, Google Translate,
Image Recognition, Recommendation Systems, Spam Filtering,
Google Page Ranking .....
CLASSIFICATION OF AI (CONTD)

Based on Capabilities - General AI

• General AI, also known as strong AI
• It can understand and learn any intellectual task that a human
being can It can learn, perceive, understand, and function entirely
like a human
• It allows a machine to apply knowledge and skills in different
contexts.
CLASSIFICATION OF AI (CONTD)

Based on Capabilities - Super AI

• Super AI surpasses human intelligence and can perform any
task better than a human.
• The concept of artificial super intelligence sees AI evolved to be
so akin to human sentiments and experiences that it doesn’t
merely understand them;
• It also evokes emotions, needs, beliefs, and desires of its own. Its
existence is still hypothetical.
• Some of the critical characteristics of super AI include thinking,
solving puzzles, making judgments, and decisions on its own.
CLASSIFICATION OF AI (CONTD)

Based on Functionalities - Reactive Machines

• It is the primary form of AI that does not store memories or
use past experiences to determine future actions.
• It works only with present data. They perceive the world and
react to it.
• Reactive machines are provided with specific tasks, and they
don’t have capabilities beyond those tasks.
• Example :
• IBM’s Deep Blue that defeated chess grand master Garry
Kasparov in 1997. It could see all the chessboard pieces and
kept reacting to each situation with a new move without
storing data for the previous move.
• Based on Functionalities - Limited Memory
• Limited Memory AI trains from past data to make decisions.
The memory of such systems is short-lived.
• They can use this past data for a specific period of time, but
they cannot add it to a library of their experiences. [ Example :
Self Driving Vehicles]
• It observes how other vehicles are moving around them, at
CLASSIFICATION OF AI (CONTD)

Based on Functionalities - Theory of Mind

 Theory of Mind AI, although mostly hypothetical at the moment,
represents an exciting potential. By definition, it shall understand
humans as entities who have emotions, sentiments, and thoughts.
While many improvements have been made in the field, the Theory
of Mind remains an undeveloped area.
 Examples
 Kismet(real world example) is a robot head made in the late
90s by a Massachusetts Institute of Technology researcher.
 It can mimic human emotions and recognize them.
 Both abilities are key advancements in theory of mind AI, but
Kismet can’t follow gazes or convey attention to humans.
 Sophia from Hanson Robotics is another example where the
theory of mind AI was implemented. Cameras present in
Sophia’s eyes, combined with computer algorithms, allow her to
see. She can sustain eye contact, recognize individuals, and
follow faces.
CLASSIFICATION OF AI (CONTD)

Based on Functionalities - Self Awareness

• The most advanced form of AI, these systems have their
own consciousness, self-awareness, and the ability to
understand their own existence. This type of AI does not
yet exist and remains a theoretical concept.

• Such systems understand their internal traits, states, and

conditions and perceive human emotions.
GOAL AND HISTORY OF
AI
Goal of AI
To Create Expert Systems : The systems which exhibit
intelligent behavior, learn, demonstrate, explain, and
advice its users.
To Implement Human Intelligence in Machines : Creating
systems that understand, think, learn, and behave like
humans.
History/ Evolution of AI
Year Milestone / Innovation
1923 Karel Čapek play named “Rossum’s Universal
Robots” (RUR) opens in London, first use of the
word "robot" in English.
1943 Foundations for neural networks laid.
1945 Isaac Asimov, a Columbia University alumnus,
coined the term Robotics.
 History/Evolution of
AI
Alan Turing introduced Turing Test for evaluation of
1950 intelligence and published Computing Machinery and
Intelligence. Claude Shannon published Detailed Analysis
of Chess Playing as a search.
John McCarthy coined the term Artificial Intelligence.
1956
Demonstration of the first running AI program at Carnegie
Mellon University.
1958 John McCarthy invents LISP programming language for AI.
Danny Bobrow’s dissertation at MIT showed that
1964
computers can understand natural language well enough
to solve algebra word problems correctly.
1965 Joseph Weizenbaum at MIT built ELIZA, an interactive
problem that carries on a dialogue in English.
1969 Scientists at Stanford Research Institute Developed
Shakey, a robot, equipped with locomotion, perception,
and problem solving.
The Assembly Robotics group at Edinburgh University built
1973
Freddy,
the Famous Scottish Robot, capable of using
vision to locate and assemble models.
 Goal and History/Evolution
of AI
1979 The first computer-controlled autonomous vehicle, Stanford Cart,
was built.
1985 Harold Cohen created and demonstrated the drawing program,
Aaron.
Major advances in all areas of AI
Significant demonstrations in machine learning Case-
based reasoning
Multi-agent planning Scheduling
1990 Data mining, Web Crawler
natural language understanding and translation Vision,
Virtual Reality
Games
1997 The Deep Blue Chess Program beats the world chess champion,
Garry Kasparov.
Interactive robot pets become commercially available. MIT
2000 displays Kismet, a robot with a face that expresses emotions.
The robot Nomad explores remote regions of Antarctica and
locates meteorites.
 Goal and History/Evolution
of AI
Early 2000s: Rise of Machine Learning and Big Data

2001 IBM's Watson development begins, focusing on natural language

processing and question answering.
2006 Geoffrey Hinton introduces the concept of Deep Belief Networks,
revitalizing interest in deep learning
2009 Google Brain project begins, aiming to create large-scale
artificial neural networks using Google's vast computing
infrastructure.

2010s: Breakthroughs in Deep Learning and AI

Applications
2012 AlexNet, a deep convolutional neural network, wins the
ImageNet Large Scale Visual Recognition Challenge, significantly
outperforming previous approaches and sparking a surge in
deep learning research.
2012 Google acquires DeepMind, which later demonstrates the
success of deep reinforcement learning with its AI system
playing Atari games
2015 OpenAI is founded with the goal of promoting and developing
friendly AI for the benefit of humanity.
 Goal and History/Evolution
of AI
2016 DeepMind's AlphaGo defeats world champion Go player Lee
Sedol, marking a major milestone in AI's ability to handle
complex strategy games.
2017 The transformer model is introduced in the paper "Attention is
All You Need," revolutionizing natural language processing and
leading to the development of powerful language models like
BERT and GPT.
Late 2010s to 2020s: AI Integration and Advancements in
Generative Models
2018 Google Duplex demonstrates an AI system that can make
human-like phone calls to schedule appointments, showcasing
advancements in conversational AI.
2019
OpenAI releases GPT-2, a language model capable of generating
coherent and contextually relevant text, raising both excitement
and ethical concerns
2020 OpenAI releases GPT-3, one of the largest and most powerful
language models to date, capable of performing a wide range of
tasks with impressive human-like text generation.
2021 DeepMind's AlphaFold achieves significant breakthroughs in
protein folding, predicting the 3D structure of proteins with
 Goal and History/Evolution
of AI
2022 The launch of DALL-E 2 by OpenAI, an advanced generative
model that creates highly realistic images from textual
descriptions, demonstrating the potential of AI in creative fields.
2023 Google Bard and Microsoft Copilot integrate large language
models into everyday applications like search engines and
productivity tools, making advanced AI capabilities more
accessible to the public.

Ongoing Trends and Future Directions

AI Ethics and Governance
Fairness and Bias Mitigation: Developing methods to ensure AI systems make
fair and unbiased decisions.
Transparency and Explainability: Creating AI systems whose decision-making
processes can be easily understood and explained.
Regulation and Standards: Establishing global standards and regulations for
the ethical use of AI.

Edge AI
On-Device Processing: Implementing AI algorithms directly on devices (like
smartphones and IoT devices) to reduce latency and enhance privacy.
Energy Efficiency: Designing AI models that require less computational power
and energy.
 Goal and History/Evolution
of AI
AI in Healthcare
Personalized Medicine: Tailoring medical treatments to individual
patients based on their genetic information and lifestyle.
Predictive Analytics: Using AI to predict disease outbreaks and patient
outcomes.
Robotic Surgery: Enhancing precision in surgeries with AI-driven
robotic systems.

Natural Language Processing

Multilingual Models: Developing models that can understand and
generate text in multiple languages.
Contextual Understanding: Improving AI's ability to understand
context and nuance in human language.

Autonomous Systems
Autonomous Vehicles: Advancing self-driving cars, trucks, and drones
for commercial and personal use.
Industrial Automation: Increasing the use of AI-driven robots in
manufacturing and logistics.
 Goal and History/Evolution
of AI AI
Quantum
Quantum Computing: Leveraging quantum computing to solve complex
problems more efficiently than classical computers.
Quantum Machine Learning: Developing new machine learning algorithms
that run on quantum computers.
Generative AI
Creative Applications: Using AI to create art, music, and literature, pushing
the boundaries of human creativity.
Synthetic Data Generation: Generating realistic synthetic data for training
machine learning models, addressing data privacy and scarcity issues.
AI in Security
Cybersecurity: Enhancing threat detection and response with AI-driven
systems.
Surveillance: Improving the accuracy and efficiency of surveillance systems
with AI.
AI in Finance
Automated Trading: Developing more sophisticated algorithms for high-
frequency trading.
Fraud Detection: Using AI to detect and prevent financial fraud.
AI in Education
Personalized Learning: Creating adaptive learning systems that tailor
educational content to individual student needs.
Automated Grading: Implementing AI systems to assist with grading and
providing feedback.