1

Modue-4 Trends in AI

1) AI and Ethical Concerns

a) Ethics of AI:The ethics of Artificial Intelligence (AI) deals with moral obligations and duties
of humans that arise from creating and interacting with robots and other artificially intelligent
agents

AI ethics is be divided into two groups.

i. Roboethics: Roboethics considers moral behavior of humans as they design, construct, use and
treat artificially intelligent beings.

ii. Machine ethics: These ethics manage the moral behavior of artificial moral agents (AMAs)
that play an important role in delivering artificial general intelligence (AGIs) within existing legal
and social frameworks.

b) Root of Ethical Problem:

 AI is basically build on data + mathematical model + training based on data +

predictions.

 What if the data provided for the training is unintentionally wrong/biased? Issues like
these are endless. Hence, getting a good understanding of them is critical to the success of
this technology.

d) Trusted AI Principles: Figure below shows some guiding principles of a trusted AI Tool we
should learn.

Figure: Trusted AI Principle

As a citizen of the AI society, we must be aware about how AI works and the framework of AI
ethics.
 2

1.1 Ethical Use of Artificial Intelligence

AI systems undoubtedly provide wide range of functionality for businesses, but their use also raises
ethical questions.

1) An ideal AI system must not be biased (it should be fair) especially when AI uses inherently
unexplainable deep learning and generative adversarial network (GAN) algorithms.

2) Their internal decision-making is often difficult to interpret, even by experts.

This lack of transparency becomes a major obstacle in industries that require strict regulatory
compliance, where every decision must be justified.

3) A responsible AI system is Explainable, monitorable, reproducible, secure, human-

centered, unbiased, and justifiable. [Characteristics of Responsible AI]

 Explainable: Able to clearly justify how and why a decision was made.
 Monitorable: Its behavior can be tracked and checked for errors, drifts, or biases.
 Reproducible: If the same input is given, the system should produce the same decision
consistently.
 Secure: Protected against attacks, data breaches, and manipulation.
 Human-centered: Designed to support and enhance human decision-making, not replace it
blindly.
 Unbiased: Decisions must be fair, free from discrimination or prejudice.
 Justifiable: The outcomes must be reasonable and defensible when audited.

Figure: Features of Ethical AI System

Formulating laws to regulate AI is not at all easy as AI uses multiple technologies. Any strict law
can hamper its progress and development.

4) Formulating laws to regulate AI is not at all easy as AI uses multiple technologies. Any strict
law can hamper its progress and development.

5) The AI technology is changing so rapidly that new technology breakthroughs sometimes

make existing laws obsolete.
 3

Additional Risks of AI
1. Job Automation: AI-powered robots are becoming more efficient and capable. As a result,
many predictable and repetitive jobs are at risk of being replaced.
Examples:

 Restaurants using robots as waiters

 Warehouse and retail jobs replaced by automation

2. Biased Algorithms:AI works on the principle of Garbage In – Garbage Out (GIGO).If an AI

model is trained on biased or unfair data, the output will also be biased. This can cause major
problems when such systems are used for real-world decisions.

Example: A smart advertising billboard in Utrecht had to be switched off after the installed
surveillance software caused public outrage due to bias concerns.

3. Too little Privacy:AI and IoT systems generate huge amounts of data.
Every day, nearly 2.5 quintillion bytes of data are added worldwide.
Nearly 90% of all digital data was created in the last two years.

Since smart systems require a lot of data to function properly, personal information is constantly
collected. Once this data is collected, people have little control over how it is used, who uses it, or
for what purpose.

Example:
 Cameras equipped with facial recognition can capture details such as age, gender, ethnicity,
expressions, and emotional state.
 In China, some police officers wear AI-powered facial recognition glasses to identify
“suspects.”

4. Everything becomes unreliable (Spread of Fake Information and Digital Manipulation)

Today, AI systems can:

 Generate realistic fake faces
 Compose text
 Create fake tweets
 Manipulate images
 Clone voices
 Produce fake videos (Deepfakes)

These technologies make misinformation and digital fraud easier.

Example:
 Deepfake software can superimpose a person’s face onto another person’s body in a video.
This has led to the creation of fake celebrity videos, blackmail, and pornographic content.
 4

1.3 Ethics in AI (Ethical AI)

Artificial Intelligence is transforming the world rapidly. People who use AI-based systems often
enjoy a more comfortable and efficient lifestyle. However, this is true only when AI is used
ethically.

So, we focus not on just "AI", but on ethical AI.

The term ethical AI consists of two words:

 Ethical → something that follows moral principles

 AI → systems that perform tasks requiring human-like intelligence

So, ethical AI refers to developing and using AI systems in a way that is fair, just, and morally
right. It focuses on the fairness and correctness of the output, outcome, and overall impact of an
AI system.

AI can produce biased (harmful) results if it is trained on biased data.

So, any AI system that works on biased data is actually not ethical (or unethical).

Sources of Bias in AI
AI systems usually become biased due to the following three major reasons:

1. Data: AI learns from the training data provided to it. If the data itself is biased, the AI system
will also produce biased results.

2. Algorithm: Even when the data contains slight biases, the algorithm may amplify them and
create stronger biases.
3. People: Developers, engineers, and scientists who design AI systems can unintentionally
introduce their own biases.

Humans design AI systems to give the most accurate results with the available data. So, data
feeders that may be humans may again hit the success unknowingly.

Therefore, it is rightly said that ethics and bias are not the problem of the machine but that of
the humans behind the machine.

Why is Ethical AI Important

 AI is used in medicine, law enforcement, recruitment, data privacy, military defense,

and self-driving vehicles, so it must give accurate, transparent, and understandable
results aligned with ethical standards.
 A biased or incorrect AI output can:
o Violate or misuse personal data
o Make decisions humans cannot understand
o Create difficulty in fixing responsibility when harm occurs
 5

 Ethical AI must be unbiased and accurate (Fig. 8.5).

 The trustworthiness of AI also depends on how it is developed and used.
 Although AI is expected to help humanity, it can also be misused to create systems harmful
to society.

1.7 Impact of AI on Jobs

 AI is one of the most in-demand skill areas today.
 World Economic Forum: AI/ML will displace 75 million jobs but create 133 million new jobs.
 Many human tasks are now automated, creating fear of job loss.

 McKinsey Global Institute:

o By 2030, AI/robots may replace 30% of global human labor.
o 400–800 million jobs may be displaced.
o 375 million people may need to switch job categories.

 Similar to the Industrial Revolution: some jobs disappear, but new ones arise.
 AI brings a paradigm shift rather than unemployment.
 New emerging IT areas (shown in Fig. 8.6) will create many opportunities.

FIGURE 8.6 Emerging Areas in IT

Jobs that Will Grow with AI

1. Creative Jobs
o Artists, doctors, scientists → AI enhances efficiency and reduces complexity.
2. Management Jobs
o Cannot be replaced easily because managing needs human emotions, ethics, and
communication.
o AI may assist, not replace human managers.
3. Tech Jobs
o High demand for programmers, data scientists, analysts, AI engineers.
o Some tech roles may change but remain vital.
4. New Job Titles by 2030
o Chief Bias Officer
o Data Detective
 6

o Man–Machine Teaming Manager

o AI Business Development Manager
o AI-Assisted Medical Professional
o AI Tutor

 AI adoption has grown 270% in four years.

 AI drives innovation → creates new industries and new job sectors.
 AI systems still have limitations:
o Can solve only one specific problem at a time.
o Lack generalized intelligence like humans.
o Require massive data (may be biased or protected).
o Limited by computing power and high electricity cost.

 Question is not “humans OR computers”, but “humans AND computers” working together.
 90% of major companies have already invested in AI.
 AI/ML skills expected to grow 71% in next 5 years.

Sectors Strongly Impacted by AI

1. Medical
o Uses huge data for prediction and diagnosis.
o Potential to reduce diseases and global poverty.
2. Automotive
o Autonomous vehicles, navigation systems, automated manufacturing.
3. Cybersecurity
o Detect and predict cyber threats.
o AI helps prevent financial fraud.
4. E-commerce
o Chatbots, product recommendations, targeted ads, warehouse automation.
5. HR Applications
o AI filters 75% of irrelevant resumes.
o Saves recruiters hours of manual work.
6. Legal Profession
o AI supports document handling, summarization, classification, and knowledge
extraction.
 7

2. AI as a Service (AIaaS)

AI as a Service (AIaaS) refers to using ready-made, cloud-based AI tools instead of developing a
full in-house AI system.
Companies can access AI features like ML models, chatbots, vision tools, etc., through cloud
platforms.
Building an AI system from scratch is expensive and requires expert teams, hardware, and large
datasets.
 AIaaS offers off-the-shelf solutions at a much lower cost.
 Just like Software as a Service (SaaS), AIaaS uses cloud computing to deliver AI tools over the
internet.
 This helps even small businesses use powerful AI services to improve performance without
heavy investment.

2.1 Factors Triggering Growth of AIaaS

AIaaS has rapidly grown in recent years due to major advancements in IT. Key factors include:

1. Growth of Cloud Platforms:Cloud services now offer affordable and flexible options for
enterprise data storage and management.
2. Improved Data Storage Technologies:Modern storage solutions are cheaper, faster, and more
reliable, making AI implementation easier.

3. Explosion of IoT & Streaming Data:IoT devices and streaming systems generate huge volumes
of data, which can be analysed for competitive insights.

4. Rise of Automated Analytics Tools:Semi-automated and fully automated tools for data
management, analytics, and business intelligence (BI) have transformed business analysis.
Moreover, some major factors that have motivated business to use AIaaS are listed below.

1. Expensive infrastructure
2. Lack of trained programmers
3. Very high charges charged by programmers for implementing AI in a particular
organization
4. Many companies lacked sufficient data to analyze

2.2 The Growth of AIaaS

 With cloud services becoming easily accessible, AI adoption has increased rapidly.
 AI can now be used by a large number of companies, allowing them to collect and store massive
amounts of data.
 Earlier, companies hesitated to build their own cloud infrastructure to develop and test AI
systems. With AIaaS, companies can now gain data insights without heavy investment in talent
and resources.

1. Allows companies to focus on core business areas instead of becoming experts in machine
learning.
 8

2. Reduces operational and infrastructure costs; AI systems require expensive hardware

like parallel machines and GPUs, which AIaaS provides at lower cost.
3. Minimizes investment risk, especially for small and new companies.
4. Short deployment time for AI-based solutions.
5. Pre-built algorithms make complex data management accessible even to small
organizations.
6. On-demand services give businesses a competitive advantage.
7. Improves benefits from data trend analysis.
8. Enables cost-effective, flexible, and transparent solutions using templates, pre-built
models, and drag-and-drop tools, reducing the need for in-house AI experts.
9. Improves strategic flexibility, as companies pay only for what they use, saving costs when
AI usage is limited or occasional.
10. Scalability is a major advantage—companies can start small and scale services up or down
based on project needs.

Figure 8.7 summarizes the advantages and shortcomings of AI as a Service.

2.3 Challenges of AIaaS

1. Reduced Security
o To avail services over cloud, companies must share data with third-party cloud vendors.
o Strong security measures are required to protect data during storage, transmission, and to
prevent illegal access.
2. Reliance on Third Parties
o When working with third parties, organizations become dependent on service providers to
meet information needs.
o This dependency may cause delays in responses.
3. Reduced Transparency
o AIaaS provides services without revealing internal processes.
o The system works like a black box, where inputs are given and outputs are received.
o Lack of clarity about algorithms may lead to confusion about data stability and results.
 9

4. Data Governance Issues

 Organizations follow strict data policies.
 These policies may restrict storing data on third-party clouds, limiting the effective use
of AIaaS.

2.4 Vendors of AIaaS

Many AI provider platforms offer machine learning and AI services.

o Organizations select AIaaS vendors by matching: Business goals, Company size, Budget, In-
house technical capability, amount of data to be processed
o Cloud AI providers offer: GPU-based processing for intensive workloads, Maintenance, staffing,
and hardware support.

Major AIaaS Vendors

1. Amazon Web Services (AWS)
2. Microsoft Azure
3. Google Cloud Platform (GCP)
[Link] Watson
AIPaaS (AI Platform as a Service)

Platform-as-a-Service is an end-to-end solution like a cloud platform, using which business

organizations can utilize required services on a pay-per-use or pay-per-service basis. With AIPaaS,
users can use third-party APIs that provides a complete, intelligent data management platform

3 Recent Trends in AI
3.1 Collaborative Systems
Collaborative systems focus on developing systems in which humans and machines (with AI) can
become effective partners by overcoming each other’s limitations and augmenting human abilities.

Examples of Machine- Human collaboration:

 Computer Game Foldit : Humans use intuition and creativity; AI handles complex
computations to fold proteins.
 Chess collaboration: Two amateur chess players + AI worked collaboratly and defeated
supercomputers and grandmasters.
 Business: In 2014, a Japanese company appointed an AI system to its board of directors.
 Military:
o UAV drones: Machines operate on front lines; humans provide virtual support.
o Talos suit: Power exoskeleton enhances soldiers with monitoring systems and HUD.
 10

3.2 Machines Assisting Humans

Smart machines assist humans in three ways: amplify, interact, and embody.

i. Amplify

 AI systems enhance human creativity, analytics, and decision-making.

 They provide the right information at the right time, improving efficiency and accuracy.

ii. Interact
 AI systems interact effectively with employees and customers.
 Digital assistants also help humans perform tasks at home using voice instructions.

Example: SEB Bank (Sweden)

 Uses Aida, a virtual assistant, to:
o Help customers open bank accounts
o Make cross-border payments
o Detect customer emotions (satisfied or frustrated) through voice tone
 In 30% cases, unresolved issues are transferred to human agents.
 The system learns from human interactions to improve future responses.

iii. Embody (Cobots & Robots)

 AI is embedded in machines like robots to augment human abilities.
 These machines use sensors, motors, and actuators to:
o Recognize people and objects
o Work safely in factories, warehouses, and labs

3.3 Algorithmic Game Theory and Computational Social Choice

With increased internet connectivity and speed (especially after the pandemic), more people play
games such as chess, checkers, poker, and solitaire. These games follow clearly defined rules,
creating the need to embed algorithms in machines so they can play ethically according to
predefined rules. This requirement led to the development of game theory.

Categories of Game Theory

i. Cooperative vs Non-cooperative Games

 Cooperative games: Players can form alliances to improve winning chances.

 Non-cooperative games: Players cannot form alliances.

ii. Symmetric vs Asymmetric Games

 Symmetric games: All players have the same goals, though strategies may differ.
 Asymmetric games: Players have different or conflicting goals.

iii. Perfect vs Imperfect Information Games

 Perfect information games: All players can see each other’s moves (e.g., chess).
 Imperfect information games: Some information is hidden (e.g., card games).
 11

iv. Simultaneous vs Sequential Games

 Simultaneous games: Players act at the same time.

 Sequential games: Players act one after another, knowing previous moves.

v . Zero-sum vs Non-zero-sum Games

 Zero-sum games: One player’s gain equals another player’s loss.

 Non-zero-sum games: Multiple players can gain benefits.

Important Concepts in AI-based Game Theory

i. Nash Equilibrium
ii. Inverse Game Theory

i. Nash Equilibrium

 The Nash equilibrium is a condition in which all the players playing the game accept that
there is no better solution to the game than the current situation.
 No player gains anything by changing their decision unilaterally.
 Any new decision made by a player will not improve the result.

Example: Prisoner’s Dilemma

 Two criminals are arrested and kept in isolation without communication.

 Possible outcomes:
o If one confesses and the other does not:
 Confessor goes free, the other gets 10 years in prison.
o If neither confesses:
 Both get 1 year in prison.
o If both confess:
 Both get 5 years in prison.
 Nash equilibrium occurs when both criminals betray each other (i.e. both confess).

cooperate=not confessing, defect=confess

 12

ii. Inverse Game Theory

 Game theory analyzes the game to optimize the possible outcome of its players.

Inverse game theory works in the opposite direction It designs a game based on players’
strategies and aims.

 It is important for designing AI agent environments.

Generative Adversarial Network (GAN):

A GAN is an ML model where two neural networks compete with each other. GAN consists of
two models a Generative model and a Discriminative model.

FIGURE 8.22 GAN Architecture

Generative model accepts some features as inputs to examine their distributions and understand
how they have been produced. Discriminative model takes the input features to predict the class to
which our sample belong.

For example, in a GAN, the generative model uses input features to create new samples that
resemble the main features of the original samples. The original samples and the newly generated
samples are then passed to the discriminative model to identify which samples are genuine and
which are fake.

A common application of GANs is to generate images (as given in Fig. 8.23) and then distinguish
between real and fake ones.

3.4 Multi-Agents Reinforcement Learning (MARL)

Reinforcement Learning (RL) enables an agent to learn by interacting with its environment using
a reward–punishment mechanism.

RL is based on Markov Decision Processes, where an agent in a stochastic environment learns a

policy and finally converges to a satisfactory policy.

 When multiple agents exist in the same environment, learning becomes more complex
because:
o The agent’s behavior depends not only on the environment
o But also on interactions with other agents
 13

 As the number of agents increases, the number of interactions grows exponentially,

making modeling difficult.
 To reduce complexity, Mean Field Scenarios (MFS) are used:
o Assume all agents have similar reward functions
o Helps simplify multi-agent modeling

Example: Traffic Control

To improve the traffic flow in a city, AI-powered self-driving cars can perfectly interact with the
external environment. But things can get complicated when multiple cars choose the same optimal
route. In such a case, a car might get in conflict with another when both of them find it most
convenient to follow a certain route.

The above situation can be easily modelled using game theory. Our cars would be different players
and the Nash equilibrium would be the point between the collaboration between the different cars.

3.5 Neuromorphic Computing

Neuromorphic computing is a technique in which a computer (both hardware and software) is

modelled based on the architecture of the human brain and its nervous system.

The main aim of neuromorphic computing is to create artificial neural systems inspired by biological
structures. Such a structure would enable the machine to behave like a human brain in the following
ways. First, to learn, retain information and even make logical decisions as the human brain makes.
Second, to acquire new information and work like a human brain.

It is a science that draws concepts from multiple disciplines including computer science, biology,
mathematics, electronic engineering and physics.

Neuromorphic computing builds artificial neural systems using:

o Neurons – process information

o Synapses – connect neurons and transfer electrical signals

It uses analog circuitry to control signal strength, similar to the human brain.

This neuron–synapse structure forms a Spiking Neural Network (SNN).

SNNs transmit electric pulses (spikes), unlike traditional neural networks that use digital signals.

Key Features of Neuromorphic Systems

1. Massively parallel – can handle multiple tasks simultaneously

2. Event-driven – respond dynamically to environmental changes
3. Power-efficient – consume very little energy
4. Flexible and adaptive – able to generalize learning
5. Strong and Fault-tolerant – continue functioning even if some components fail

Architecture Advantage

 Neuromorphic systems do not follow the Von Neumann architecture.

 14

 Memory and processing are combined within each neuron.

 Eliminates the need for separate:
o Memory Unit (MU)
o Central Processing Unit (CPU)
o Data paths
 This avoids the Von Neumann bottleneck, where data movement causes delays and high
 Neuromorphic systems overcome this by providing fast computation with low power
usage and by mimicking the human brain.

4. Expert System

An expert system is a computer program that solves complex problems in a particular

domain, at the level of extra-ordinary human intelligence and expertise. That is, it is
designed to provide decisions just like a human expert.

Expert systems are the computer applications developed to solve complex problems. Expert
systems are designed for a specific domain, such as medicine, science, etc

To perform its task, the expert system analyses user queries to extract knowledge from its
knowledge base using the reasoning and inference rules.

The first expert system was developed in 1970. The performance of an expert system depends
on the knowledge stored in its knowledge base. More the amount of knowledge stored in the
KB, better is the performance of the expert system.

An expert system is not used to replace the human experts; instead, it is used to assist the human
in making a complex decision.

The working of an expert system can be understood through the block diagram given in Fig.
9.1.

Figure 9.1 Working of an expert system

The main components of an expert system include user interface, knowledge base and inference
engine.

Popular Examples of the Expert System

1. DENDRAL- Used in organic chemistry to detect unknown organic molecules

2. MYCIN: Backward chaining expert systems that was designed to diagnose blood clotting
diseases
 15

3. PXDES: Expert system that determines the type of lung cancer by analyzing picture(s) from
the upper body.

4. CaDeT: Diagnostic support system to detect cancer at early stages.

5. R1/XCON: Selects specific software to generate a customized computer system

6. DXplain: Clinical support that suggests the presence of one or more diseases based on the
findings of the doctor

4.2 Characteristics of an Expert System

Sl. No. Characteristic Description

1 High Performance Solves complex domain-specific problems efficiently and

accurately using an updated knowledge base.

2 Understandable Accepts input and provides output in a human-understandable

language.

3 Reliable Produces consistent and accurate results, making it dependable

for decision-making.

4 Highly Generates solutions quickly, often faster than a human expert.

Responsive

5 No Memory Can store and utilize a large amount of knowledge and data
Limitations without loss.

6 Expertise in a Combines knowledge from multiple human experts to provide

Domain comprehensive solutions.

7 Not Affected by Performs consistently without being influenced by fatigue, stress,

Emotions or emotions.

8 Not Biased Considers all available facts logically before providing

conclusions.

9 Reduced Cost Provides expert-level consultation at a much lower cost compared

to human experts.

10 Non-Perishable Once developed, it can be used indefinitely without loss of

expertise.

11 Intelligent Uses a knowledge base and inference engine with if–then rules to
deduce new facts.
 16

4.3 Components of an Expert System

An expert system mainly consists of three components—user interface, inference engine and
knowledge base (as shown in Fig. 9.3).

FIGURE 9.3 Components of an Expert System

i. User Interface

The user interface enables interaction between the user and the expert system. Users enter queries in
a simple and readable format, which are processed by the inference engine. The results are then
displayed to the user. The interface is designed to be user-friendly so that even non-experts can use
the system. The user interface also explains how the expert system derived a particular conclusion.

ii. Inference Engine (Rules of Engine)

The inference engine is the main processing unit and hence known as brain of the expert system.
The engine derives conclusion or deduces new information by applying inference rules to the
knowledge base.

a. Forward Chaining:

Forward chaining starts from the known facts and rules (refer Fig. 9.4).

FIGURE 9.4 Forward Chaining

b. Backward Chaining:

Backward reasoning, on the other hand, starts from the goal and works backward to prove the known
facts (as shown in Fig. 9.5).
 17

FIGURE 9.5 Backward Chaining

iii) Knowledge Base

The knowledge base contains information and rules of a particular domain or subject. The two main
components of a knowledge base are factual knowledge and heuristic knowledge.

a) Factual knowledge is based on facts and accepted by knowledge engineers in that domain.

b) Heuristic knowledge is based on practice, probability, evaluation and experiences

Knowledge in a knowledge base is usually represented using the If-else rules and is acquired by
extracting, organizing and structuring the domain knowledge under the guidance of human experts.
This knowledge is domain-specific and of high-quality

4.4 Participants in the Development of Expert System

In Fig. 9,7, we see that there are three primary participants in an expert system:

Figure: 9.7 Participants in an expert system

a) Expert:

The performance of any expert system depends on the knowledge fed in its knowledge base. This
knowledge is provided by human experts who are specialized in that specific domain.

b) Knowledge Engineer:

The term ‘knowledge engineering’ refers to the process of building an expert system and its
practitioners are called knowledge engineers. The main task of a knowledge engineer is to ensure
that the computer possesses all the knowledge required to solve a problem. For this, he/she gathers
 18

the knowledge from the domain experts and then represent that knowledge in the format accepted by
the knowledge base.

c) End User:

The end user who uses the expert system may or may not be well versed in that domain. He/she just
uses the expert system to find solution or advice for his queries, which are usually complex.

4.5 Capabilities of the Expert System (Read any 5)

An expert system can do the following things:

1. Advise: The system can give its advice to users for any query—simple or complex— that fits its
domain.
2. Provide decision-making capabilities: An expert system can be designed to make complex
decisions in any field, be it finance, medical science or any other.
3. Demonstrate a device: It is can be used to demonstrate features, specifications, usage of any new
product.
4. Problem-solving: It can be used to solve complex problems.
5. Explaining a problem: An expert system can be used to provide a detailed description of an input
problem.
6. Interpreting the input: It can interpret the input given by the user.
7. Predicting results: An expert system can predict results using past data.
8. Instructing and assisting human in decision-making
9. Diagnosing: Expert systems, especially in the field of medical science, are used to diagnose any
disease.
10. Justifying the conclusion: Expert systems not only find solutions to complex real-world
problems by drawing conclusions from known facts but also justify those conclusions.
11. Suggesting: The best part is that an expert system also suggests alternative options to a problem.

However, expert systems cannot do the following:

1. Substitute human decision-makers

2. Possess human capabilities
3. Produce accurate output for inadequate knowledge base
4. Refining knowledge base
5. Use emotions to make decisions

4.6 Advantages of Expert Systems (Read any 5)

1. Results obtained by an expert system are highly reproducible.

2. They can be used in risky places where it is not safe for humans to work.
3. They give accurate results, especially when the knowledge base is updated regularly.
4. They give consistent performance at a fast speed as these systems are not affected by emotions,
tension or fatigue.
5. They are easily available due to mass production of software.
6. They are affordable as the production cost is reasonable.
7. They work at great speed and reduce human effort.
8. They are capable of explaining how the solution was obtained.
 19

9. They improve the quality of decisions.

10. They gather scarce expertise knowledge in good amount and use it efficiently.

4. 7 Limitations of Expert Systems (Read any 5)

1. Results obtained from an expert system may get wrong if the knowledge base contains
wrong/obsolete information.
2. Each problem is different; therefore, the solution from a human expert may be strikingly different
from that of an expert system.
3. They lack creativity for different scenarios, especially in extraordinary situations.
4. They have high development and maintenance costs.
5. Knowledge acquisition is a challenging task.
6. Success depends on knowledge acquired from human experts.
7. These systems cannot learn from themselves and hence require manual updates.
8. They cannot produce correct results from less amount of knowledge.
9. They need excessive training.
10. They do not have any emotions.
11. They do not possess any common sense.
12. They are developed only for a particular domain.
13. They are not capable of making decisions in extraordinary situations.
14. They work on Garbage-in Garbage-out (GIGO) principle. This means that if there is an error in
the knowledge base, the expert system will give incorrect results.
15. They have high maintenance cost.

4. 8 Applications of Expert Systems (Read any 5)

Expert systems are one of the prominent research domains of AI. Some common applications of
expert system are as follows:

1. Designing and manufacturing devices such as VLSI systems, camera lenses and automobiles.
2. Publishing relevant knowledge content to users. An advisor and a tax advisor are two popularly
used expert systems in this domain.
3. Finance industries use expert systems to detect any type of possible fraud, suspicious activity,
stock market trading and advise bankers on whether they should approve the loans or not.
4. Expert systems are extensively used for medical diagnosis and treatment. In fact, it was the first
area where these systems were used.
5. Expert systems are used for planning and scheduling tasks for achieving the project’s goal.
Companies also use expert systems for scheduling airlines, cargo customer order, computer
resources and various manufacturing task.
6. Tracking the progress of a software development project.
7. Monitoring applications compare data continuously with observed system or with prescribed
behaviour to detect leakage in long petroleum pipeline using expert systems.
8. Process control systems use expert systems to control a physical process based by continuously
monitoring it.
9. The automobile and electronics industries use expert systems to identify faults in vehicles and
computers manufactured.
10. Diagnosing complex electronic, electromechanical and diesel-electric locomotive system.
11. Forecasting crop damage or in general predicting results of any experiment/activity.
12. Identifying the structure of a chemical compound.
 20

13. Assessing geologic structure from dip meter logs and space structure through satellite and robot.
14. Evaluating civil cases, product liability, employee performance, etc. using logs.
15. Detecting virus attack in a computer system.
16. Repair and maintenance projects.
17. Optimizing warehouses.
18. Configurating manufactured objects

4.9 Development of Expert Systems

The development of an expert system is an iterative process involving the following steps:

1. Identify the Problem Domain

Identify the problem to be solved and ensure it is suitable for an expert system. Domain
experts are identified, and feasibility and cost-effectiveness are analyzed.
2. Design the System
Select appropriate hardware, tools, and expert system shells. Plan system integration and
decide how domain knowledge will be represented.
3. Develop the Prototype
Knowledge is acquired from experts and represented using if–then–else rules. Heuristic
knowledge and explanation methods are also defined.
4. Test and Refine the Prototype
The prototype is tested using sample cases and with end users to identify errors. Based on
feedback, the system is refined and improved.
5. Develop and Complete the System
The finalized system is tested with users and other systems. Proper documentation is
prepared and users are trained.
6. Maintain the System
The knowledge base is regularly updated, and the system is modified to meet new
requirements and interface changes.

TABLE : Traditional System vs Expert System

 21

[Link] of Things (IoT)

IoT means a system of interrelated computing devices, machines, objects, animals or people
that have unique identifiers for identification and also the ability to transfer data over a
network without requiring human-to-human or human-to-computer interaction (as shown in
Fig. below).

In IoT, each “thing” is assigned a unique identifier and the ability to communicate over the
Internet. For example, wearable health monitors used by people, biochip transponders install in
animals sensors used by vehicles to generate alerts, and other objects are connected using IP
addresses

5.1 Applications (examples) of IoT (Read any 5)

i) Everyday Life Applications of IoT
a. Smart vehicles suggest the best route and notify others in case of traffic delays.
b. Smart alarms trigger coffee machines automatically when you wake up.
c. Smartwatches track activity, productivity, and health data.
d. Self-driving cars use sensors to detect obstacles.
e. Smart sports equipment (e.g., football) tracks performance data for training.
f. Smart refrigerators notify about low stock or expired food items.
g. Smart air conditioners can be controlled remotely via smartphones.

ii) Smart Cities :IoT enables efficient use of resources in smart cities.
Helps in: Reducing traffic congestion, Lowering pollution and noise, Controlling crime, Improving
public services.

iii) Business and Industrial Applications

 IoT improves:
o Process efficiency, Asset utilization, Productivity
o Sensors help companies: Track customer behavior, Identify popular products
o Align supply with demand, Prevent stock shortages
 22

5. 2 IoT Products
Many global companies are actively developing IoT solutions, including: Honeywell, Hitachi, GE,
Cisco, AT&T, Apple, Google, IBM, Microsoft, Red Hat, Zebra Technologies, etc.

Examples of IoT Products

 Amazon Echo (Smart Home)
o Uses voice assistant Alexa
o Performs tasks like playing music, weather updates, ordering services, etc.
 Fitbit One (Wearable Device)
o Tracks steps, calories, sleep quality, and floors climbed
o Syncs data wirelessly with smartphones and computers
 Barcelona – Smart City
o Uses IoT for smart parking and environmental monitoring
o One of the leading smart cities globally
 IoT Platforms
o Amazon Web Services
o Microsoft Azure
o IBM Watson
o Cisco IoT Cloud Connect
o Salesforce IoT Cloud
o Oracle Integrated Cloud
o GE Predix
 AT&T – Connected Car
o Added 1.3 million cars in 2016
o Total connected cars reached 9.5 million
 Future Outlook
o IoT offers endless opportunities and challenges
o Nearly $6 trillion expected to be spent on IoT in the next five years

5. 3 Challenges of IoT (Read any 10)

Security is the biggest challenge in IoT, Sensitive data protection is a major concern. Many IoT
devices do not encrypt data over local Wi-Fi, Unsecured networks increase hacking risks, Faulty
wearable IoT devices in healthcare can be life-threatening

Major Challenges
1. Scalability
o Billions of devices generate massive data, requires big data analytics and cloud
storage
2. Interoperability
o Fragmented standards prevent seamless device communication
3. Lack of Standardization
o Difficulty in integrating IoT with legacy systems
4. Lack of Government Support
o Absence of strong regulations for safety and security
5. IoT Malware and Ransomware
o Hackers can lock devices and demand ransom
o Example: hijacking cameras or personal data
6. Connectivity Issues
o Poor internet affects IoT sensor performance
 23

7. Lack of effective and informed government regulations

o Legislators often lack understanding of IoT technology
8. No Privacy Policies
o Devices may collect and share data without user consent
9. Bandwidth Strain
o Increased IoT devices demand higher network bandwidth

10 IoT professional skills Gap: IoT suffers from a professional skills gap. The main reason
behind this is that IoT is growing rapidly and many people are yet to develop a skillset working
with this technology.

o 38% lack IoT network expertise

o 27% lack deployment experience

5. 4 Sensors (Read any 10)

A sensor is a device that detects changes in the physical environment such as heat, light, motion,
pressure, moisture, gas, etc. Sensors convert these physical inputs into signals that can be displayed
or transmitted for further processing.

 Oxygen sensors in vehicles monitor the fuel–oxygen ratio and help maintain efficient
engine performance.
 Motion sensors are used in security systems, automatic doors, and lighting to detect
movement using waves or light interruption.
 Vision and imaging sensors detect objects, colours, and images within their field of view.
 Temperature sensors measure thermal conditions of solids, liquids, and gases in industrial
processes.
 Radiation sensors detect alpha, beta, or gamma rays and are used in monitoring and safety
applications.
 Proximity sensors detect nearby objects without physical contact and are widely used in
manufacturing.
 Pressure sensors measure force per unit area in liquids or gases.
 Position sensors determine the position of mechanical components like valves and doors.
 Photoelectric sensors detect objects and properties like colour or cleanliness in automation
systems.
 Particle sensors detect dust and airborne particles, commonly used in environmental
monitoring.
 Metal detectors identify the presence of metal in objects or people.
 Level sensors measure the level of liquids, solids, or gases in tanks or containers.
 Leak sensors detect unwanted leakage of gases or liquids.
 Humidity sensors measure moisture content in the air.
 Gas and chemical sensors detect specific gases or chemicals for safety and control
purposes.
 Force sensors measure parameters such as weight, torque, and load.
 Flow sensors detect the movement of fluids or gases in processing industries.
 Flaw sensors identify surface or internal defects in materials.
 Flame detectors sense fire and are used in combustion control systems.
 Electrical sensors measure current, voltage, and related parameters.
 Contact sensors detect physical touch and are used in alarm and access systems.
 Non-contact sensors work without physical contact, such as radar speed sensors.
 24

 Speed sensors measure the speed of moving objects.

 Ultrasonic sensors use sound waves to measure distance or detect objects.

6. Artificial Intelligence of Things (AIoT)

AIoT, or the combination of artificial intelligence (AI) technologies and the Internet of Things
(IoT) infrastructure aims to make IoT operations more efficient, improve interactions between
humans and machines and facilitates better data management and analytics.

 AI simulates human intelligence in machines.

 IoT is a network of interconnected devices with unique identifiers that exchange data
without human intervention. IoT devices can include: Heart-monitor implants, Self-driving
cars, Sensors, smart objects, and everyday items

AIoT enhances IoT by adding machine learning for improved decision-making.

6.1 How Does AIoT Work?

 AI is embedded into programs and infrastructure components, like chips, which are all
connected using IoT networks.
 APIs ensure smooth data exchange between hardware, software, and platforms without human
involvement.
 IoT devices generate as well as collect data, that is then analyzed using AI techniques to provide
better insights into data. This improves efficiency and productivity.

Figure 9.12 shows an autonomous vehicle in which AI is embedded into chipsets, which are all
connected using IoT networks

FIGURE 9.12 Autonomous Vehicle

 25

6.2 Where Does AI Unlock IoT?

IoT services follow five steps:

1. Create 2. Communicate 3. Aggregate 4. Analyse 5. Act

Since the Act stage depends on the Analyse stage, using AI in the analysis stage significantly
improves:

 Efficiency
 Productivity
 Decision-making accuracy

AIoT helps in achieving the following [Benefits of AIoT]:

1. Manage, analyse and obtain meaningful insights from data

2. Provide fast and accurate data analysis
3. Maintain a balance between requirements for localized and centralized intelligence
4. Protect data privacy and confidentiality
5. Ensure security against cyber attack

6.3 Applications and Examples of AIoT (Read any 8)

AIoT is widely used across various domains where IoT data is analysed using AI to enable smart
decision-making.

1. Smart Cities
 Sensors, lights and meters collect data.
 AI processes data to:
o Improve operational efficiency
o Drive economic growth
o Enhance quality of life
 Used for traffic management, safety and resource optimization.
2. Smart Retail
 Smart cameras identify customers at self-checkout.
 Sensors track customer movement to predict checkout congestion.
 Helps in:
o Dynamic staffing
o Reducing waiting time
o Personalized marketing based on customer behaviour.
3. Smart Home
 Smart appliances analyse user interaction data.
 26

 Understand user habits and provide customized support and automation.

4. Manufacturing
 Smart chips monitor equipment health.
 AI predicts failures and identifies the right time for maintenance or replacement.
5. Hiring
 AIoT integrates data from HR platforms and social media.
 Helps identify suitable candidates without bias.
6. Autonomous Vehicles
 Vehicles use cameras and sensors to collect real-time data.
 AI analyses surroundings and driving conditions to take instant decisions.
7. Robots
 Autonomous robots use multiple sensors to sense the environment.
 AI helps robots make smart and optimized movements.
8. Healthcare
 Wearables and medical devices monitor real-time health data.
 AI analyses patient records, test results and imaging data.
 Improves diagnosis, enables telemedicine and reduces administrative burden.
9. Smart Thermostat
 Users control temperature remotely via smartphones.
 AI learns user preferences and schedules.
 Example: Nest smart thermostat.
10. Drone Traffic Monitoring
 Drones collect real-time traffic data.
 AI analyses data to manage traffic flow automatically.
 Example: Alibaba Cloud’s ET City Brain for accident detection and traffic control.
11. Smart Office Buildings
 Sensors detect occupancy and adjust lighting and temperature.
 Facial recognition used for access control and attendance management.
 Improves energy efficiency and security.
12. Fleet Management
 AIoT monitors vehicle location, fuel usage and driver behaviour.
 Helps reduce costs, improve maintenance and ensure safety.

6.4 Benefits and Challenges of AIoT (Read any 5)

Benefits of AIoT
1. Improved Operational Efficiency
AI analyses IoT data to identify patterns and insights and automatically adjusts system
operations, improving efficiency.
2. Real-time (On-the-fly) Decision Making
With AIoT and edge computing, data is analysed instantly, allowing quick decisions to
prevent failures or handle emergencies.
3. Reduced Workload
Since AI analyses large volumes of data automatically, it saves human effort, time and cost.
4. Scalability
AIoT systems can easily add more IoT devices as needed. AI summarizes data from sensors
before sharing it, enabling large-scale connectivity without overload.
5. Boosts Operational Efficiency
Continuous data analysis helps identify redundant processes and fine-tune operations for
 27

better results.
Example: Google uses AIoT to reduce data-centre cooling costs.
6. Better Risk Management
AIoT predicts risks and automates responses, helping prevent financial losses, improve
safety and protect against cyber threats.
Example: Fujitsu uses AI to analyse data from wearable devices to ensure employee safety.
7. New and Enhanced Products & Services
AIoT enables better human-machine interaction (e.g., using NLP) and improves existing
services by analysing real-time data.
Example: Rolls-Royce uses AIoT for aircraft engine maintenance.
8. Reduced Unplanned Downtime
Predictive maintenance detects failures in advance and schedules maintenance, reducing
downtime and costs.
Example: Deloitte reports reduced maintenance time, increased equipment uptime and
lower maintenance costs.

Challenges of AIoT
 System Failures Can Be Risky
If AIoT systems fail, the consequences can be serious.
o Autonomous delivery robots may delay deliveries
o Smart retail systems may misidentify customers
o Autonomous vehicles may cause accidents if sensors fail

Future of AIoT
 AIoT is formed by integrating Artificial Intelligence (AI) with the Internet of Things (IoT)
to create intelligent systems capable of making decisions without human intervention.
 The deployment of 5G networks will significantly boost AIoT by enabling faster data
transfer through high bandwidth and low latency.
 AIoT can revolutionize supply chains and delivery models by efficiently solving operational
problems.
 Many organizations are investing in AI and IoT to improve operational efficiency and gain
competitive advantages.
 IoT devices generate massive volumes of data every day, with around 42 billion connected
devices expected globally by 2025.
 The rapid increase in data presents both new opportunities and challenges for AIoT systems.
 AI enables IoT devices to learn from past data, predict future events, and continuously
improve decision-making.
 Local data processing using AI reduces network congestion and delays, improving system
performance.

Question Bank

1) What is Ethics of AI, describe 2 groups of Ethics, What are the key ethical concerns of AI
2) Describe the features of Ethical AI
3) Describe the sources of Bias in AI
4) Explain the concept of AI as a Service (AIaaS). Describe the factors triggering the Growth
of AIaaS
5) Describe the benefits and challenges of AIaaS
6) Describe the following recent trends in AI
 28

i) Collaborative System, ii) Machines Assisting Humans iii) Algorithmic Game Theory and
Computational Social Choice along with iv) Nash Equilibrium concept v) Multi Agent
Reinforcement Learning (MARL) vi) Neuromorphic Computing.
7) Explain Neuromorphic Computing.
8) What is an Expert System? Describe its components with neat diagram.
9) Describe the Characteristics, capabilities and advantages of Expert Systems
10) What is IoT? Describe the applications of IoT
11) Explain the challenges of IoT
12) What is a sensor? Describe any 10 Sensors
12) What is AIoT? Describe the applications of AIoT
13) Describe the benefits and challenges of AIoT