8

Current Trends in Artificial Intelligence

8.1 AI and Ethical Concerns

We know that ethics define the discipline that deals with moral obligations
and duties of humans.

Correspondingly, ethics of AI deals with ethics specific to robots and other

artificially intelligent beings. These ethics can be divided into two groups.

1. Roboethics: Roboethics consider moral behaviour of humans as they

design, construct, use and treat artificially intelligent beings.
2. Machine ethics: These ethics manage the moral behaviour of artificial
moral agents (AMAs) that play an important role in delivering artificial
general intelligence (AGIs) within existing legal and social frameworks.
FIGURE 8.1 Trusted AI principles

The bigger concerns are as follows:

1. If AI generates human-like output, can it also make human-like

decisions?
2. If AI makes human-like decisions, then are these decisions human-like?
3. If AI takes decisions of sanctioning or rejecting a bank loan, is that
decision justified?
4. If AI decides whether a student should be enrolled in a college or not, is
AI taking a fair decision?
5. If AI makes a human-like decision, is it human-like trustworthy?

AI is basically 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. Refer Fig. 8.1 to learn about
some guiding principles of a trusted AI Tool. As a citizen of the AI society,
we must be aware about how AI works and the framework of AI ethics.

8.1.1 Ethical Use of Artificial Intelligence

AI systems no doubt offer a wide range of functionality for businesses, but

their use also raises ethical questions.

An ideal AI system must not be biased. This is especially true when AI uses
inherently unexplainable deep learning and generative adversarial network
(GAN) algorithms.
FIGURE 8.2 Features of Ethical AI System

The ability to explain is a potential stumbling block to using AI systems in

industries that operate under strict regulatory compliance requirements. For
example, financial institutions must explain their credit-issuing decisions.
The main problem with such systems is that it will be difficult to explain
how the decision was taken to refuse a credit as such decisions are made by
teasing out subtle correlations between thousands of variables. In such a
scenario, the AI program that is not able to explain decisions is known as
black box AI.

A responsible AI system is explainable, monitorable, reproducible, secure,

human-centered, unbiased and justifiable (refer Fig. 8.2).
Formulating laws to regulate AI is not at all easy as AI uses multiple
technologies. Any strict law can hamper its progress and development. The
AI technology is changing so rapidly that new technology breakthroughs
sometimes make existing laws obsolete. For example, existing laws
regulating the privacy of conversations and recorded conversations do not
cover the challenge posed by voice assistants like Amazon’s Alexa and
Apple’s Siri. These digital assistants collect data not to distribute them but
to improve its machine learning algorithms. This is a golden chance for
criminals having malicious intent.

8.1.2 Is AI Dangerous? Will Robots Take Over the World?

Rapid growth and increase in the capabilities of AI systems have made us

wondering about the risks involved in using AI. Some of us think that AI
will quickly takeover tasks performed by humans. But most researchers
believe that super-intelligent AI is unlikely to exhibit human emotions, so
there is no point thinking that AI will become malevolent. AI can become a
risk only in two conditions.

AI system is specifically programmed to do something devastating. For

example, autonomous weapons are AI systems that are programmed to kill.
If acquired by scrupulous people having ill intentions, such weapons could
inadvertently lead to an AI war, and mass casualties, potentially even the
end of humanity.
Autonomous weapons are designed to be extremely difficult to ‘turn off’,
and once they become operational, humans could themselves rapidly lose
control over them. This risk is also present with narrow AI when given
autonomy.

Just imagine what would happen if an autonomous drone with facial

recognition as well as a 3D-printed rifle, pistol or other gun becomes
available. Or if a self-driving car, connected to the Internet is hacked to get
into some serious accident.

Even in hospitals, more and more equipment are now connected to the
Internet. If any or all of these are hacked, then can you imagine what the
hacker can do with the patient’s body?

AI could be programmed to do something beneficial, but the method

used to achieve its goal can be highly destructive. This is especially true
when the AI programmer asks the machine to complete a task, but does not
clearly outline the goals. For example, you can instruct a self-driven car to
program to take you to a particular destination as soon as possible.
However, the instruction ‘as soon as possible’ fails to address safety, road
rules, etc. The smart car may successfully complete its task, but after
creating a great havoc. So, there must be some provision to continuously
monitor and control the machine, its goals and process being performed to
achieve that goal.
Of 9 million low-skilled services and BPO roles, 30 per cent or around 3
million will be lost by 2022, principally driven by the impact of robot
process automation or RPA.

As of now, there is no liability for actions on machines. There is no clarity

on what legal aspects bind machines when they become increasingly smart.
The following questions may arise:

1. Do we judge AI systems the same way as we judge a human?

2. Who is responsible if the AI system becomes self-learning and
autonomous to a greater extent?
3. Is there any error margin for AI machines, even if it has fatal
consequences?

Apart from these two serious concerns, AI also poses some additional
threats/risks that calls for attention. These threats are discussed below.
 Credit: PaO_STUDIO / Shutterstock

FIGURE 8.3 Effect of AI on jobs

1. The immediate risk posed by job automation: As AI robots become

smarter and more dexterous, they will quickly replace humans. For
example, you must have heard of restaurants having robots as waiters.
Just imagine, where the less educated people will go if machines start
taking their jobs (as seen in Fig. 8.3)?
It is often rightly feared that the deployment of AI systems for job
automation will eventually replace certain types of jobs, especially those
that are predictable and repetitive. According to a 2019 Brookings
Institution study, 36 million people work in jobs that they may soon lose
owing to automation as at least 70 percent of their tasks (varying from
retail sales, market analysis to hospitality and warehouse labour) will be
done using AI. In fact, a newer Brookings report even states that white
collar jobs may be worst affected. As per McKinsey & Company report
 (2018), the African American workforce will be hardest hit by
automation.
2. Biased algorithms: We know that computers work on GIGO concept
which says Garbage-In-Garbage-Out. This is a very serious limitation in
AI applications. If we feed our algorithms data sets that contain biased
data, then the output generated from such systems will only produce
biased results. This biased result, if applied to solve a real-world
problem, may lead to an even bigger problem.

A dynamic advertising billboard in Utrecht was switched off because the

spy software installed on these billboards aroused public outrage.
3. Too little privacy: When using IoT or AI systems, a lot of data is
generated. It is said that approximately 2.5 quintillion bytes (or 2.5
million terabytes) of data is added each day. It is interesting to note that
90% of the digital data has been created in the last two years. A lot of
data is required for the proper functioning of the smart systems. As a
result, much data about us is collected thereby eroding our privacy. Once
the systems collect our data, we have no means to know what data about
us is used by whom and for what purpose.
These days, cameras can easily be fitted with facial recognition software
to capture details about us (including our gender, age, ethnicity, gesture
and state of mind). Do you know that in China, some police officers
wear glasses with facial recognition technology having a database with
 facial pictures of thousands of ‘suspects’ (judged on the basis of certain
behaviour)?

FIGURE 8.4 AI Tools can be used for Face Swapping and Face
Manipulation
4. Everything becomes unreliable: These days, fake news and filter
bubbles are not uncommon. Smart systems can create faces, compose
texts, produce tweets, manipulate images, clone voices and engage in
smart advertising.

An AI system can be used to depict day into night or create highly realistic
faces of people who have never existed.

For example, open-source software Deepfake can easily stick pictures of

faces on moving video footage. Due to such fake videos, celebrities are the
worst affected. People with malicious intentions easily create pornographic
videos starring them. Even normal citizens are blackmailed with morphed
images and/or videos. This is often known as Faceswap video
blackmailing (refer Fig. 8.4).

Even in politics, fake or manipulated videos are produced at a high speed to

influence the opinions of the masses.

Did you know that the company Cambridge Analytica, was found to
manage access to data from 87 million Facebook profiles of Americans to
campaign for President Trump so that he could come to power?

8.1.3 Ethics in AI

The current world is quickly getting transformed by artificial intelligence

and those who use this technology are living a richer and easy-going life.
However, this is possible only when we talk about ethical AI, not just AI.

The term, ethical AI, is composed of two words—ethical and AI. We are
now already familiar with the term AI. Ethical means one that is according
to ethics. And ethics relates to moral principles that govern the behaviour
and actions of a group or an individual. So, ethical AI is the theory behind
developing computer systems that can perform tasks that require human
intelligence. It focusses on how right, how fair and how just is the AI’s
output, outcome and impact.
Example: We know that AI can produce biased results if it is trained with
biased data or if it does not do anything to avoid bias. So, any AI system
that works on biased data is actually not ethical (or unethical). For example,
Microsoft released a chatbot called Tay on Twitter in 2016, to learn by
engaging people in dialogue through tweets or direct messages. However,
trolls made Tay learn all negative words that spread hatred for women, a
particular race and all sorts of biased data within hours. Tay was also taught
to give repulsive and toxic responses. This aroused the anger of Twitter
users and finally Microsoft had to silence Tay forever in even less than 24
hours.

An ethical AI system designed to solve a particular problem, uses

unbiased data and is trained using the right learning model for the
problem and is monitored to evaluate if the results produced by it is right
and fair.

8.1.4 AI and Bias

Think about the following statements.

1. Most images that show up when you do an image search for ‘doctor’ are
white men.
2. Images and videos of a ‘doctor’ show usually shows a male character
and that of a ‘nurse’ is a woman.
3. All virtual assistants like Alexa, Siri, Google assistant, etc. work with
female voice.
4. Computer vision systems frequently give an error while recognizing
people of different colour.

Can you think of more such cases where we witness biasness towards a
particular gender, colour, caste, religion, country, etc? Technically, AI bias
occurs when an algorithm produces results that are systematically
prejudiced towards a particular group of people and thus produces skewed
output. AI applications may have built-in biases because ultimately, they are
created by humans who have conscious or unconscious preferences that
may go undiscovered until the algorithms are used at a massive scale. Let
us understand three basic sources of bias in AI systems.

Data

Any AI system is just as good as the data we put into them. Naturally,
adding biased or skewed data can never give fair results. Machines do not
have the intelligence to differentiate between right or wrong training data.

For example, many-a-time, you must have noticed that Google or any other
voice assistant replies in American accent and at times gives wrong results
when we give a command in Indian accent. This happens because the AI
system is trained on recorded speech from white, upper-middle class
Americans, making it difficult for the technology to understand commands
from people not belonging to this category.
 Algorithm

Algorithm can further amplify the biases injected by skewed data. For
example, if you Google Image Search ‘Teacher’, then the image classifier
algorithm trained on the images available in public domain shows more
women teaching as opposed to male teachers. AI algorithms must be
designed to maximize accuracy. But in this case, the AI algorithm may
decide that all people teaching are women, despite the fact the training data
has some images of men in the classroom. So, this is a case of gender bias
in the AI system that becomes a part of AI algorithm due biasness in data.

People

People who are developing the AI system, that is, engineers, scientists,
developers, etc. are the next big source of bias. 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.

8.1.5 Towards Ethical and Trustworthy AI

Today, several companies worldwide are using algorithm-based ‘emotional

AI’ hiring platforms to augment and lower the financial burden of their
recruitment processes. While these systems may hire in a fair and unbiased
manner, there are cases reporting women applicants being
disproportionately rejected based on years of biased data in a male-
dominated sector.

Let us now see few ways in which world is trying to deal with this situation.

Regulating a More Ethical AI

In April 2021, the European Commission launched its first ever legal
framework on AI to guarantee the safety and fundamental rights of people
and businesses, while strengthening AI uptake, investment and innovation
across the EU.

The new risk-based approach will not only set strict requirements for AI
systems but also immediately ban AI systems which are considered to ‘be a
threat to the safety, livelihood and rights of people’ including ‘systems that
manipulate human behaviour, circumvent users’ free will and allow social
scoring by governments’.

Later, in June 2021, Australia launched a similar AI ethics framework to

guide businesses, governments and other organizations to responsibly
design, develop and use AI.

Company and Organizational Engagement

Besides necessary regulation and standards, these days, organizations are

building trust in AI technologies through cultural and educational
programmes, risk assessments and third-party audit programmes. This
would help them to move towards a prevention, detection and response
framework (like anti-corruption, prevention of tax evasion frameworks).

Rights and Activist Groups

AI systems are yet to go a long way. In this journey, the relationship

between people and technology will be redefined continuously. To ensure
that our societies move forward on the path of ethics, it is critical that AI
systems are human-centric. Civil rights and activist groups need to
challenge the discourse and amplify the voices of those who are most
adversely affected by new technologies.

We must all ask the question—Is the technology necessary or is there any
alternative? Only if the benefits outweigh the harm, we must check form
whom the AI systems work.

Ensuring Data Privacy

Companies are now taking measures that are important to keep data safe
and protect it from potential misuse. Privacy protection is now becoming an
integral part of any ethical AI system. Data ethics and privacy in such
systems is ensured by the following:

1. Encrypting the data.

2. Ensuring that the AI algorithm cannot learn from outside its dataset.
3. Using secure computation so that even the people who developed or
work on those AI systems are not able to see or access the data.
4. Restrict reverse engineering the AI model so that no user can access data
used for model training.

Diversify Your Team

Since bias is an important issue and results from an AI tool will be used by
people from different racial, gender and economic identities, we need to
make sure that there should be no bias. For this, build diverse teams to
reduce the potential risk of bias falling through the cracks. A diverse team
has data scientists, business leaders, professionals with different educational
backgrounds and experiences, such as lawyers, accountants, sociologists
and ethicists. This would help the company to get their views on any sort of
bias so that it could be mitigated in a timely manner. The team should
continually analyse data and algorithms ensure fairness. There are many
tools like Bias Analyzer which automate this process and also show the
costs and benefits associated with a variety of possible mitigation actions.

8.1.6 Why is Ethical AI Important?

Since AI is used in areas like medicine, law enforcement, recruiting, data

privacy, military defense or self-driving vehicles, it makes it mandatory that
the AI system must produce accurate, transparent and understandable
results that is in synchronization with the ethical standards and norms of our
society. To understand the need of ethical AI systems, we must first
understand that a biased or incorrect output from AI system that assist in
law enforcement, job recruitment, defense work, self-driving vehicles can
result in eroding people’s privacy (by misusing data in unintended ways),
and taking decisions that are impossible for people to understand and to
access liability for damages when harm is caused.

Apart from being unbiased and accurate (as shown in Fig. 8.5), the
credibility of AI systems also depends how we choose to develop and use it.
We can just hope that AI will be used for purposes that benefits mankind.
But there is nothing that can stop a person from creating systems that can be
disastrous for humanity.

In October 2019, researchers found that an algorithm used in US hospitals

to predict which patients may require extra medical care heavily favoured
white patients over blacks. This bias creeped in because of the fact that
white patients paid more bills and availed extra medical facilities for the
same medical problem as compared to black.

Amazon’s hiring algorithm also suffered from bias related issues. The
algorithm was found to be biased against women. This was probably due to
the fact that the system saw a greater number of men applicants than
women.
FIGURE 8.5 An ethical system must be unbiased and accurate

Nonetheless, efforts are being made by numerous agencies, committees,

coalitions and expert groups to ensure that we are on the right track to make
AI systems do more to heal than harm. For example, the IEEE Global
Initiative on Ethics of Autonomous and Intelligent Systems that put out
Ethically Aligned Design, a publication from 2016 outlines the guiding
standards for developing and administering ethical AI solutions. Therefore,
ethical AI aims to ensure is that AI is doing good for the planet.

Going further on this, the European Commission High-Level Expert Group

on AI identified seven guiding principles for creating Trustworthy AI. These
guidelines are as follows:

1. AI systems should support human autonomy and decision-making.

2. AI systems should be technically robust and safe to use. They must have
a fallback plan that can be used when something goes wrong.
3. AI systems should ensure data privacy by protecting user’s data, and
providing adequate mechanisms to maintain the quality and integrity of
the data.
4. AI systems should not work on biased data. The system should be
clear, explicable and transparent about its underlying data and the
models.
5. AI system should be trained with diverse, non-discriminatory and
fair data and models to avoid unfair bias.
6. AI systems should be designed to benefit everyone, now and in future.
They must focus on societal and environmental wellbeing to be
sustainable and environmentally friendly.
7. An AI system should have the accountability to ensure accurate,
unbiased outcomes with utmost responsibility.
8. An AI system must give due importance to data privacy and data
security. A system that provides design confidentiality, transparency and
security into their AI programs must ensure that data is collected, used,
managed and stored safely and responsibly.
9. An AI system must be accountable for its decisions. There must be
someone in the organization who will be held responsible for any
problematic situation.
Have you ever wondered how much information Google has about you? In
an Android phone, Google has your entire Contact List. Google knows the
names and numbers of people you talk to. Google knows your location. It
knows where you go every day and on holidays. Google knows what email
you write to whom. Google has your photographs (Google Photos), knows
details about your friends (Google Hangouts). All this is done in lieu of free
services it provides. But imagine what big risk it poses to our privacy.
8.1.7 Impact of AI on Jobs

Today, AI is on the tip of everyone’s tongue and one of the most in-demand
areas of expertise for job seekers. The World Economic Forum has
predicted that AI and ML will displace 75 million jobs but would also
create 133 million new ones in a few months.

Some of our favourite sci-fi movies have shown that the advent of this
technology has created a fear that AI will one day make human beings
obsolete in the workforce. As the technology is advancing, many tasks that
were once executed by humans have now been automated.
FIGURE 8.6 Emerging Areas in IT

A two-year study from McKinsey Global Institute suggests that by 2030,

intelligent agents and robots could replace as much as 30% of the world’s
current human labour. That is, automation will displace between 400 and
800 million jobs and require approximately 375 million people to switch
job categories entirely.

However, job losses due to AI is a fear far from reality. It will just introduce
a paradigm shift, similar to the one which occurred after the Industrial
Revolution. When the Industrial Revolution took place, many feared that
people will lose jobs to machines. But now we see that revolution has
generated lot more opportunities and jobs. Similarly, with AI, many
professions will become obsolete and disappear, but other new emerging
occupations in areas listed in Fig. 8.6 will become popular. Learning AI
skills will help you gain a significant career advantage.

Jobs which will grow with the help of AI include the following.

Creative jobs will get refined and advanced by use of AI. These jobs are
usually performed by professionals like artists, doctors and scientists. AI
will not only enhance efficiency but also make these jobs less complex for
humans.

AI will probably not make human workers obsolete, at least not for a long
time.

Management jobs cannot be replaced by artificial managers because

anytime, human managers will have to control artificial managers.
Managing is a very complex task that involves deep understanding of
emotions, profession, ethics, people and communication. Human managers
can use certain tools to become more effective at their jobs.

Tech jobs including programmers, data scientists, data analysts, big data
engineers and all those who work on the creation and maintenance of AI
systems will always be in high demand. However, a few tech jobs which are
in demand today may become less common, while others may become
more vital.
A few job titles which can be expected to appear by 2030 include Chief
Bias Officer, Data detective, Man–Machine teaming manager, AI business
development manager, AI assisted medical professional, AI tutor, to name a
few.

In many sectors of business, AI has grown by 270% over the last four years.
Rather than promoting the obsolescence of humans, AI will continue to
drive massive innovation that will fuel many existing industries and could
have the potential to create many new sectors for growth, ultimately leading
to the creation of more jobs.

We must understand that although AI has done a lot to replicate the efficacy
of human intelligence in executing certain tasks, AI programs suffer from
major limitations. They can only solve one problem at a time. These
systems become rigid, and are unable to think outside their prescribed
programming. In contrast, humans have generalized intelligence which
helps them solve any kind of problem, think abstractly to make critical
judgement across all sectors.

Moreover, AI can learn only with massive amounts of relevant data that is
available at the right time. Many a time, this data is protected under issues
of privacy and security. In addition, data may have bias. And it is quite
obvious that biased data will never give accurate results.
AI Is becoming the standard in all businesses, not just in the world of tech

Machines may also suffer because of availability of limited computation

and processing power. For example, the cost of electricity required to power
one supercharged language model AI was estimated around $4.6 million.

So, AI has the potential to ultimately create more jobs, not less. The
question is not ‘humans or computers’ but ‘humans and computers’
involved in complex systems that advance industry and prosperity. Nobody
would want to be behind the curve when it comes to AI. In fact, 90% of
leading businesses have already invested in AI technologies. More than half
of businesses that have implemented some manner of AI-driven technology
report experiencing greater productivity.

AI and machine learning top the skills required by companies these days
and jobs in this area are expected to increase by 71% in the next five years.
AI is likely to have a strong impact on certain sectors in particular. These
are discussed below.

1. Medical: The medical industry has huge data which can be utilized to
create predictive models related to healthcare and by physicians for
diagnosis. And at a higher level, AI and automation will also help to
eliminate disease and world poverty.
2. Automotive: Autonomous vehicles, autonomous navigation and
manufacturing within the automotive sector have all been possible with
 AI technology.
3. Cybersecurity: During the pandemic, cyber frauds rose by 600% as
hackers capitalized on people working from home, on less secure
technological systems and Wi-Fi networks. AI and machine learning
provide powerful tools to identify and predict threats. AI is also playing
a vital role in the field of finance as it can process large amounts of data
to predict and catch instances of fraud.
4. E-commerce: On e-commerce websites (like Amazon), AI is used to
design chatbots, recommend products, use image-based targeted
advertising, and assist in warehouse and inventory automation.
5. HR applications: An HR department gets thousands of applications
every day. AI tools can be used to reject 75% of resumes that do not fit
the current context. Prior to AI, HR managers had to devote considerable
time to filter applications to select suitable candidates. Data from
LinkedIn shows that recruiters spend up to 23 hours looking over
resumes for one successful hire.
6. Legal profession: Support services with document handling -
classification, discovery, summarization, comparison, knowledge
extraction, and management use AI for efficiency and accuracy.

Thus, the AI revolution will lead to a new era of prosperity, creativeness

and well-being. Humans will no longer be asked to perform routine,
limited-value, full-time jobs. Rather, jobs will be flexible and offer selective
premium services. There will be more jobs related to programming,
robotics, engineering, etc. And, blue-collar and white-collar jobs will be
eliminated.

8.2 AI as a Service (AIaaS)

In artificial intelligence as a service (AIAAS), companies use off-the-shelf

AI tools to implement and scale AI techniques at a very low price as
compared to building a complete in-house AI system.

When a software is provided as a service across the network, it usually uses

cloud computing. Similarly, the idea of providing AI services using cloud
computing has allowed even small business organizations to use cost
effective solutions to improve their performance in a big way.

8.2.1 Factors Triggering Growth of AIaaS

In the last couple of years, AIaaS is growing fast as a business opportunity

and many startups have joined to lead the AI revolution. Some major
advancements in IT in the last few years have fostered the adoption of
AIaaS. Some examples are as follows:

1. Availability of cloud platform with a variety of affordable options for

enterprise data management.
2. Data storage technologies have improved and have become cheaper and
reliable.
3. Streaming devices and IoT technologies generate massive amounts of
data which when analysed can render vital information required to gain
competitive advantage.
4. Availability of semi or fully automated data management, analytics and
BI products changed the way business was analysed.

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 analyse

8.2.2 The Growth of AIaaS

With cloud services becoming incredibly accessible, AI has seen a big

boom. Now, AI can be made available to a greater number of companies
and even those companies can collect and store any amount of data.

Previously, companies were hesitant to build their own clouds to develop,

test and utilize their own AI systems. But now with AIaaS, every other
company can take advantage of data insights without making huge
investment in talent and resources. Moreover, AIaaS has been a lucrative
option as business can now easily accomplish the following goals.
1. Focus on its core area rather than getting involved in becoming an expert
in machine learning.
2. Reduce operational and infrastructural cost to a big extent. A
complex AI system may require many parallel machines and speedy
GPUs. Not many companies can afford such huge investment especially
in its initial years. But with AIaaS, every small company can also
harness the power of machine learning at significantly lower costs.
3. Minimize the risk of investment.
4. Time required to deploy a solution is short.
5. Pre-built algorithms make complex data management accessible to
even small business organizations.
6. Availability of on-demand services empower business with a
competitive edge.
7. Increase the benefits obtained by analysing data trends.
8. Develop cost effective, flexible and transparent solution. AIaaS provides
a software to the organization that enables them to access machine
learning capabilities using templates, pre-built models and drag-and-drop
tools to assist developers in building a more customized machine
learning framework. Companies no longer have to maintain a team with
AI skilled professionals, train them, upgrade their knowledge and spend
on areas that only partially support decision-making.
9. Improve its strategic flexibility. Companies can only pay for what they
actually use. Though machine learning requires a lot of computing power
to run, small companies may actually be using only in short amounts that
 too in spurts. So, in such a scenario, AIaaS is a huge saving.

FIGURE 8.7
10. Scalability is another big advantage of AIaaS. Now companies can
delve into AI technology with smaller projects. If they find it beneficial,
then probably, they can experiment more services. At any point in time
with the change in demands of the ongoing project, companies can scale
up or down their demand for services.

Figure 8.7 summarizes the advantages and shortcomings of providing AI as

a Service.
 8.2.3 Challenges of AIaaS

1. Reduced security: To avail services over cloud, companies have no

option but to share their data with third-party vendors. Lot of data
security techniques need to be incorporated to ensure that data is safe
during storage and transmission and illegal access is strictly avoided.
2. Reliance; When working with third parties, the organization becomes
dependent on them to satisfy their information needs. This results in
delay in getting response.
3. Reduced transparency: In AIaaS, organizations are paying to get
service, but not the access. Third party vendors take the input and give
the output. The entire process works like a black box. The organization
has no idea about the inner workings (like algorithms used). This may
result in confusion or miscommunication regarding the stability of data
or the output.
4. Data governance: Every industry and organization has its own data
policy. These policies may restrict data storage on a third-party cloud,
thereby preventing a company from taking full advantage of AIaaS.

8.2.4 Vendors of AIaaS

There are multiple AI provider platforms offering various machine learning

and AI services. Organizations can match their needs with the services and
compare the cost at which those services can be accessed to select the
offering that is best suited to their current requirements.
Cloud AI service providers allow users to use GPU-based processing for
intensive workloads along with their staffing and maintenance staff as well
as catering to the hardware changes for different tasks. Some major vendors
of AIaaS are Amazon Web Services (AWS), Microsoft Azure and Google
Cloud Platform (GCP). Each of these vendors offer different types of bots,
APIs and machine learning frameworks to cater to the organization’s
demands. Apart from these, countless start-ups are focusing on AIaaS. It is
not uncommon for larger companies to purchase smaller companies to add
the developed services to their portfolios.

An organization decides on an AI service based on its goals, business size,

available budget, technical capabilities of its in-house teams and the amount
of data that needs to be processed.

MonkeyLearn is an AI platform that simplifies text analysis through

intuitive, no-code tools. Users can either use pre-trained model like survey
analyzer to classify customer feedback by topic or build customized
machine learning models to detect sentiment, keywords, and topics in the
data. Once developed, these models can be integrated with apps through
point-and-click integrations or via the API.

IBM Watson hosts a suite of AI tools and offers many pre-built

applications, like Watson Assistant and Watson Natural Language
Understand to build virtual assistants and to perform advanced text analysis
tasks respectively. Similarly, IBM Watson Studio can be used to build, train
and deploy machine learning models across any cloud. With these offerings
even users with no expertise on machine learning or data science can make
the most of their data.

Microsoft Azure is Microsoft’s public cloud computing platform that

provides a range of AI and machine learning solutions for developers. For
example, Azure Cognitive Services can be used to add computer vision or
text extraction capabilities to apps using APIs. Similarly, Azure Bot Service
can be used to incorporate any type of bot in the application.

Google Cloud ML is Google’s AI platform that basically facilitates data

scientists and developers working with big data to create and deploy
machine learning projects.

AutoML is being extensively used to train custom machine learning models

for text analysis, image classification, translation and more. Data in the
datasets can be visualized, ‘what-if tools’ can be used and metrics can be
analysed to assess performance of the model.

AIPaaS or AI 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.
Software-as-a-Service (SaaS) market is valued at $133 billion and the
advanced technology platforms-as-a-service market globally is forecast to
reach about $11 billion by 2023 and surpass ‘$88,500 million by the end of
2025.’

8.3 Robotics

We have read that robot is a programmable machine that can complete a

task. Correspondingly, the term robotics is the study focused on designing,
developing and programming physical robots which are able to interact
with the physical world and automate one or more tasks. Each robot has a
different level of autonomy varying from human-controlled bots that carry
out tasks to fully-autonomous bots that perform tasks without any external
influences.

8.3.1 Artificially Intelligent Robot

Many people think that AI and robotics are synonymous. But this is not
true. Table 8.1 highlights the difference between an AI program and a robot
system.
The potential field acting on the robot is then computed as the sum of the
attractive field of the goal and the repulsive fields of the obstacles. Thus,

Correspondingly, the net force after considering attracting and repulsing

forces is calculated as,

8.4 Recent Trends in AI

In this section, we will read about some latest trends in AI. Extensive
research is being done in the terms discussed here.

8.4.1 Collaborative Systems

These days, a lot of research is being done to develop collaborative systems

in which humans and machines (with AI) can become effective partners by
overcoming each other’s limitations and augmenting human abilities. There
are many examples of machine-human collaboration as follows.
1. The computer game Foldit uses machine-human collaboration fold
simulated proteins. The folding task is performed to understand how
real-world proteins involved in the causes of human disease are formed.
In the game, both the player as well as AI performs operations. The AI is
put to use where it excels and humans’ intuition and imagination is
applied where they exceed machines.
2. Two amateur chess players and an AI system worked collaboratively and
won against a field of supercomputers and grandmasters.
3. In 2014, Knowledge Ventures, a Japanese venture capital company,
elected an AI system to its board of directors, to realize the benefits of
machine-human collaboration at the highest levels of business.
4. In the military, UAV drones are being used in which a machine takes the
front lines with human provides virtual support.
5. The Talos suit in the US military augments a front-line human with a
power exoskeleton. The exoskeleton’s integrated heating cooling, vital
monitoring and heads-up display (HUD) collaborate with humans much
similar to that seen in the Ironman movies.

While many people believe that humans and machines can be good
partners, others, however, fear that relying too much on AI will result in a
massive job loss and making common man people dumb and lazy.

However, the good news is that research involving 1,500 companies says
that many companies that have used AI to automate processes and displaced
employees will see only short-term productivity gains. Better performance
is expected only when humans and machines work together to actively
enhance each other’s complementary strengths- leadership, teamwork,
creativity, social skills of humans and speed, scalability and quantitative
capabilities of the machines.

AI can boost our analytic and decision-making abilities and heighten

creativity.

Humans Assisting Machines

In a human-machine collaborative system, humans need to perform three

important roles. First, they must train machines to perform a particular task;
explain the outcome of that task (especially when results are
counterintuitive or controversial) and sustain the responsible use of
machines so that machines do not harm mankind. For example, the
European Union’s new General Data Protection Regulation (GDPR)
empowers consumers with the right to get an explanation for any algorithm-
based decision (like reason for acceptance/rejection of a loan). This feature
will not able be a big step in delivering responsible AI but will also create
about 75,000 new jobs to administer the GDPR requirements.

Last but not the least, companies also need employees who can
continuously monitor to ensure that AI systems are functioning properly,
safely and responsibly. For example, a self-driving car must not hit objects
and any living being. Similarly, tech companies including Apple use AI to
collect personal details about users to improve the user experience, but
collecting too much data can compromise privacy, anger customers and
even go against the law. Therefore, employees make sure that AI use data
only for analysis and the privacy of individual users is not violated.

Points to remember:

Humans must also train AI systems to interact with humans in the best
possible way. For example, all virtual assistants were given extensive
training to develop to react as a confident, caring and helpful personality
without being bossy. Such training requires inputs from a team of experts
(even including a poet, a novelist and a playwright).

These days, AI assistants are also being trained to display sympathetic

feelings. For example, the start-up Koko, an offshoot of the MIT Media
Lab, is developing an empathetic assistant. For example, if a user had a bad
day, then Koko would feel sorry for it, ask for more information and offer
advice to help the person see his issues in a different light.

8.4.2 Machines Assisting Humans

Smart machines are helping humans enhance their abilities in three ways—
amplify, interact and embody. AI systems boost human’s creativity, analytic
and decision-making abilities by providing the right information at the right
time. They can also interact with company’s employees and customers in
more effective ways. Even at homes, digital assistants are assisting humans
to perform their instructions.
SEB, a Swedish bank, uses Aida, a virtual assistant to interact with millions
of customers and assist them in opening a bank account, making cross-
border payments and interpreting their voice tone (satisfied or frustrated) to
provide better service later. In 30% of the cases when the system is not able
to resolve an issue, it directs the call to a human customer-service
representative and then monitors that interaction to learn how to resolve
similar problems in the future.

Case study Question:

Collect information about how Autodesk’s Dreamcatcher AI enhances the

imagination of even exceptional designers.

Embodying

AI systems are embodied in machines (like robots) to augment the

capabilities of a human worker. Using their sophisticated sensors, motors
and actuators, these machines can recognize people/objects and work safely
in factories, warehouses and laboratories. Such robots also known as cobots
handle repetitive actions that require heavy lifting or some dangerous tasks.
In such a collaborative environment, humans are free to do other
complementary tasks that require dexterity.

For example, Hyundai has extended the concept of cobot with exoskeletons.
Exoskeletons are wearable robotic devices, which adapt to an industrial
worker’s location in real time to enable him to perform his job with
superhuman endurance and strength. Similarly, Mercedes-Benz, uses cobot
arms to customize cars as per user’s expectations according to the real-time
choices consumers make at dealer’s showroom.

8.4.3 Algorithmic Game Theory and Computational Social Choice

These days, (especially enhanced due to pandemic) with increased Internet

connectivity and speed, more people prefer to play games like chess,
checker, poker and solitaire. These games are played using a clear set of
defined rules. So, there was a need to embed algorithms in machines that
allowed machines to play ethically according to pre-defined rules. This led
to the onset of game theory. In a multi-agent situation (when more than one
person is involved in solving a logical problem), game theory is used to
choose from a set of options knowing that our choice would affect the
choices of the opponent and their decision would affect our choices.

Von Newman is credited for the invention of game theory.

We can understand the concept of game theory by studying the five

categories.
1. Cooperative vs Non-cooperative Games: In cooperative games,
participants can establish alliances to increase their chances of winning
the game. Correspondingly, in a non-cooperative game, participants
cannot form any alliance.
2. Symmetric vs Asymmetric Games: In a symmetric game, each
participant has the same goal(s). However, their strategy to win may be
different. In contrast to this, in an asymmetric game, the participants
have different (and at times conflicting) goals.
3. Perfect vs Imperfect Information Games: In perfect information
games like chess, all the players can see what moves the other player(s)
are making. But, in an imperfect information game (like card game),
other players’ moves are hidden.
4. Simultaneous vs Sequential Games: In simultaneous games, multiple
players can take actions concurrently. In a sequential game like in a
board game, each player is aware of the other players’ previous actions.
5. Zero-sum vs Non-zero Sum Games: In zero sum games, a player gains
something that causes a loss to other players but in a non-zero sum
game, multiple players get the advantages when another player gains.

When using AI techniques in game theory, two concepts are very important
—Nash equilibrium and inverse game theory.
 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 would gain anything after changing their current
strategy. So, any new decision made by the players will not result in
anything better.

Example: A typical example of Nash equilibrium is the Prisoner’s

Dilemma. If two criminals get arrested and are kept in confinement without
being allowed to communicate with each other, then look at Fig. 8.21 to
understand the possibilities.

If either of the two prisoners confess that the other committed a crime,
the first will be set free and the other will be sent to prison for 10 years.
If none of them confess, then they will be sent to prison only for 1 year.
If both confess, then both will be sent to prison for 5 years.
FIGURE 8.21 Nash equilibrium

In this case, the Nash equilibrium is reached when both criminals betray
each other.

Inverse Game Theory

Game theory is used to understand the dynamics of a game to optimize the

possible outcome of its players. In contrast to this, inverse game theory
designs a game based on players’ strategies and aims. It is key to designing
AI agents environments.

Practical Examples
A generative adversarial network (GAN) is a machine learning (ML) model
in which two neural networks compete with each other to become more
accurate in their predictions.

Nash equilibrium can be attained easily in symmetric than asymmetric

games. But, in real world, asymmetric games are more common.

GANs contains two models—a generative model and a discriminative

model (Fig. 8.22).

FIGURE 8.22 GAN Architecture [3]

Generative models accept some features as input to examine their

distributions and understand how they have been produced.

Discriminative models take 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.

Credit: doglikehorse / Shutterstock

FIGURE 8.23 Generate Images using GAN Models

Even in game playing, GANs are used extensively. The players (the two
models) challenge each other. While one player creates fake samples to
confuse the other, the other tries to identify the right samples. The game is
played iteratively until Nash equilibrium is reached. In each iteration, the
learning parameters are updated to minimize loss.
 8.4.4 Multi-Agents Reinforcement Learning (MARL)

Reinforcement learning (RL) is used to make an agent (model) learn

through the interaction with its environment. RL was initially developed
based on Markov decision processes. This means that the agent placed in a
stochastic stationary environment tries to learn a policy through a
reward/punishment mechanism. The agent will finally converge to a
satisfactory policy.

However, this does not hold true when multiple agents are placed in the
same environment. Therefore, instead of depending only on learning
through interaction between the agent and the environment, now it is also
dependent on the interaction between agents.

Modelling systems with multiple agents is not a trivial task as increase in

the number of agents exponentially increases the number of possible ways
in which they can interact with each other. Therefore, modelling multi-
agents reinforcement learning models can be done using mean field
scenarios (MFS) to reduce its complexity by making the assumption a priori
that all agents have similar reward functions.

Example: 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 if these cars think as a group. 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.

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

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.

Thus, a computer needs to be transformed into a cognition machine.

Did you know that our brain uses roughly 20 watts of power on average,
which is about half that of a standard laptop?
 How Does Neuromorphic Computing Work?

Traditional neural network and machine learning algorithms are best suited
for existing algorithms. They either provide fast computation or focus on
using low power and there is always a trade-off between the what to
achieve. Neuromorphic systems perform fast computation while consuming
less power consumption. Additionally, they also resemble human brain as
they have the following features:

1. They are massively parallel, to handle multiple tasks at once.

2. They are event-driven, as they respond to events dynamically based on
changing environmental conditions.
3. They are economical in terms of power used.
4. They are flexible as they are highly adaptable and are able to generalize.
5. They are strong and fault-tolerant, since they produce results even when
some component(s) have failed.

Neuromorphic computing achieves this brain-like function and efficiency

by building artificial neural systems that implement ‘neurons’ (nodes that
process information) and ‘synapses’ (the connections between those nodes)
to transfer electrical signals using analog circuitry. This structure controls
the amount of electricity being passed from one node to another to mimic
the varying degrees of strength that usually occurs in a human brain. Such a
system of neurons and synapses that transmit these electric pulses is known
as a spiking neural network (SNN). This facility was not available in
traditional neural networks (that uses digital signals).

Neuromorphic systems deviate from the Von Neumann architecture and

introduce a new chip architecture that collocates memory and processing
together on each individual neuron. Hence, there is no need to have separate
memory unit (MU), central processing unit (CPU) and data paths. Having
separate parts requires information to be repeatedly moved back and forth
between different components to complete a given task, creating a
bottleneck for time and energy efficiency. This is known as the von
Neumann bottleneck.

By bringing memory on the neuromorphic chip, information can be

processed much more efficiently, thereby making the chip more powerful
and efficient.

Neuromorphic Computing and Artificial General Intelligence (AGI)

We have seen that the term ‘artificial general intelligence’ (AGI) exhibits
intelligence equal to that of humans has not yet been achieved and may not
even be reached with the traditional computers. But with neuromorphic
computing, researchers are feeling optimistic about it.

Example: The Human Brain Project, featuring the neuromorphic

supercomputer SpiNNaker produces a functioning simulation of the human
brain and is one of the hottest research projects in AGI.
A machine is said to achieve AGI when the machine can do the following:

Reason and make judgments in an uncertain situation

Plan
Learn
Communicate using natural language
Represent knowledge
Integrate the above skills to fulfil a goal

An AGI may even have the ability to imagine, gain from experience and be
self-aware. To confirm whether a system has developed AGI or not, it can
be made to pass the Turing Test or the Robot College Student Test, in which
a machine enrolls in classes and obtains a degree like a human would.

Debatable Issue

Even if a machine develops human intelligence, then a question rises that

whether it should be handled ethically and legally. While some researchers
argue that it should be treated as a nonhuman animal in the eyes of the law,
others say that still a lot has to be done to inculcate consciousness in
machines.

Comparing Human Brain with Neuromorphic Computing

Brains need very less energy than most supercomputers. While a human
brain uses about 20 watts, the Fugaku supercomputer needs 28 megawatts
(about 0.00007% of Fugaku’s) power supply.

Supercomputers need elaborate cooling systems but our brain sits at 37°C in
between bones.

Supercomputers can make complex millions and billions of complex

calculations in second(s) but the human brain is very adaptable. A single
brain can write poetry, identify a familiar face out of a crowd, drive a car,
learn a new language, instantly take decisions, respond to events and do
much more.

While traditional computers designed on von Neumann architecture are

largely serial, brains use massively parallel computing. Brains are also more
fault-tolerant than computers. Therefore, harnessing techniques used by a
human brain will be the key to making more powerful neuromorphic
computers in the future.

So How Can You Make a Computer That Works Like the Human
Brain?

To understand how neuromorphic computers work, let us first recapitulate

how the brain works.

When we feel hot or cold, neurons (a type of nerve cell) carry messages to
the brain. To transfer messages, several neurons release chemicals across a
gap called a synapse. This happens until the message reaches the brain.
Once the brain identifies the sensation, it sends an appropriate message to
the point which felt the sensation using neurons.

An action potential is triggered either through sending multiple inputs at

once (spatial), or by building multiple inputs over time (temporal). The
brain transfers information quickly and efficiently through the huge
interconnectivity of synapses where one synapse might be connected to
10,000 other synapses.

Neuromorphic computers work like human brains using spiking neural

networks (SNN). While transistors used in a conventional computing can be
either on (1) or off (0), SNN can produce more than one outputs and
transfers information in temporal and spatial way as the human brain does.

Neuromorphic systems can be either digital or analogue. Synapses are

controlled either by software or memristors. Memristors can store a range
of values instead of storing just one and zero. Like the human brain, they
can vary the strength of a connection between two synapses to allow the
brain-based systems to learn.

Uses of Neuromorphic Systems

As of now, to perform heavy computational tasks, devices like smartphones

hand off processing to a cloud-based system. The cloud processes the query
and sends the output to the device. This scenario will change with
neuromorphic systems. In such a system, instead of passing queries and
responses back and forth, computations would be done within the device
itself.

The current generation AI is rules-based. It is extensively trained on

datasets to generate a particular outcome. However, the human brain does
not work like this. Researchers are trying hard to make the next generation
of artificial intelligence deal with more brain-like problems. For example,
constraint satisfaction, in which a system has to find the optimum solution
to a problem with a lot of restrictions.

Neuromorphic systems also perform well with noisy and uncertain data.

Examples of Neuromorphic Computer Systems Used Today

64 Intel’s neuromorphic chips named Loihi are used to make an 8 million

synapse system called Pohoiki Beach, consisting of 8 million neurons (may
soon reach 100 million neurons in the near future). They are being used to
create artificial skin and also develop powered prosthetic limbs.

IBM’s neuromorphic system, TrueNorth, launched in 2014, consists of 64

million neurons and 16 billion synapses. IBM in a partnership with the US
Air Force Research Laboratory is trying to create a ‘neuromorphic
supercomputer’ known as Blue Raven for creating smarter, lighter, less
energy-demanding drones.
The EU-funded Human Brain Project (HBP), a 10-year project that started
in 2013, has led to two major neuromorphic initiatives, SpiNNaker and
BrainScaleS. In 2018, the largest neuromorphic supercomputer of that time,
a million-core SpiNNaker system went live. It is hoped that this model
would soon scale up to one million neurons. BrainScaleS also has a similar
goal and its architecture is now on its second generation, BrainScaleS-2.

Challenges to Using Neuromorphic Systems

Shifting from Von Neumann to neuromorphic computing is not that easy

and straight-forward. For example, when analysing visual input, traditional
systems treat them as a series of individual frames, but a neuromorphic
processor would encode this information as changes in a visual field over
time.

New programming languages need to be developed to utilize the change in

the underlying architecture. New kind of memory, storage and sensor
technology need to be created to take full advantage of neuromorphic
devices. This may even change the entire process of creating and integrating
hardware and software components.
 9

Where AI Is Heading Today?

9.1 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. This system makes
decisions for complex problems using both facts and heuristics like a
human expert. 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. The
suggestion of spelling errors while typing in the Google search box is a
perfect example of expert systems used these days by all users.

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

FIGURE 9.1 Working of an expert system

9.1.1 Popular Examples of the Expert System

1. DENDRAL: It is a chemical analysis expert system. DENDRAL was

used in organic chemistry to detect unknown organic molecules with the
help of their mass spectra and knowledge base of chemistry. The expert
system studied a substance’s spectrographic data to predict its molecular
structure.
2. MYCIN: It is one of the earliest backward chaining expert systems that
was designed to diagnose blood clotting diseases and find the bacteria
causing infections like bacteraemia and meningitis. MYCIN could also
recommend antibiotics and drugs based on the person’s physical
conditions.
3. PXDES: It is an expert system that determines the type and level of lung
cancer by analysing picture(s) from the upper body.
4. CaDeT: CaDeT is an expert system that works like a diagnostic support
system to detect cancer at early stages.
5. R1/XCON: The expert system could select specific software to generate
a customized computer system based on user’s requirements.
6. DXplain: It is a clinical support system that could suggest the presence
of one or more diseases based on the findings of the doctor.

9.1.2 Characteristics of an Expert System

Expert systems are among the first truly successful forms of artificial
intelligence (AI) software. Let us read about their distinguishing
characteristics (refer Fig. 9.2).
FIGURE 9.2

1. High performance: An expert system provides high performance for

solving any type of complex problem of a specific domain with high
efficiency and accuracy. The success of an expert system depends on its
knowledge base. So, to get accurate knowledge continuously, the
knowledge base is updated regularly.
2. Understandable: The results of an expert system are easily understood
by users. The system takes input in human language and provides the
output in the same language.
3. Reliable: An expert system generates efficient and accurate output. This
makes it a reliable system.
4. Highly responsive: An expert system generates results for any complex
query within a very short period of time, making it highly responsive.
 The total time by an expert system should be less than that taken by a
human expert.
5. No memory limitations: An expert system can store and exploit huge
amounts of data.
6. Expertise in a domain: There are many human experts in a particular
domain and each of them specialize in a different skill and has different
experiences. When we put the knowledge gained from these human
experts into the expert system, the results are provided by utilizing all
the facts and knowledge
7. Not affected by emotions: An expert system is not affected by human
emotions such as fatigue, anger, depression, anxiety, etc. This helps them
to give a consistent performance every time they are used.
8. Not biased: To respond to any query, the expert systems check and
consider all the available facts.
9. Reduce cost: Human experts charge a heavy amount for examining,
diagnosing and treating a patient. But with an expert system, multiple
experts can be consulted at much cheaper rates.
10. Non-perishable: Human experts are perishable as they have a limited
life span but an expert system is permanent. Once developed, it can be
used without any limitations of expiry.
11. Intelligent: These systems use knowledge base and inference engine to
solve complex problems. This is done by deducing new facts from
existing facts that are represented as if-then rules.
 9.1.3 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

User Interface

The user interacts with expert system through the user interface. It is
through this interface that the user enters queries in a readable format. The
interface then passes it to the inference engine. After getting response from
the inference engine, it displays the output to the user. The user interface is
designed to be very simple and intuitive so that even a non-expert user can
communicate with the expert system to find solution to a complex
problem.
The user interface also explains how the expert system derived a particular
conclusion. This explanation may be text displayed on screen, verbal
narrations in natural language or a listing of rules on the screen. Through
this, the user interface makes it easy to trace the credibility of the
conclusions. An effective user interface must have the following features:

1. Help users to accomplish their goals in shortest possible way.

2. Be easy to customize as per user’s requirements.
3. Make efficient use of user input.

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. There are
two types of inference engine:

Deterministic Inference Engine

This engine uses facts and rules to draw conclusions. These conclusions are
thus assumed to be true.

Probabilistic Inference Engine

This inference engine contains uncertainty in conclusions as it is based on

the probability.
Efficient procedures and rules used by the inference engine helps to
formulate correct, flawless solutions. In case there is a conflict when
multiple rules are applicable to a particular case, the inference engine
resolves the problem by selecting the best rule that is applicable in the
current context. Inference engines can also include explanation and
debugging abilities. The explanation module, if incorporated, can be used
by the expert system to give the user an explanation about how it reached a
particular conclusion.

The inference engine generally uses two strategies for acquiring knowledge
from the knowledge base and deriving solutions. These are as follows:

1. Forward Chaining: As discussed earlier, forward chaining starts from

the known facts and rules (refer Fig. 9.4). After applying the inference
rules, conclusions are drawn and then added to the known facts.
Forward chaining strategy facilitates an expert system to answer the
question, ‘What can happen next?’ For example, forward chaining can
be used to predict stock prices after changes in interest rates.

FIGURE 9.4 Forward Chaining

2. 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). This strategy enables an expert system to answer the
question, ‘Why this happened?’
If an event has already taken place, then the inference engine can be used
to find out which conditions occurred in the past for this result. Thus,
backward chaining strategy is used for discovering cause or reason. For
example, diagnosis of blood cancer in humans.

FIGURE 9.5 Backward Chaining

Knowledge Base

Before going to knowledge base, let us first understand what knowledge is.
From junior classes, we are taught that data is a collection of facts. Data,
when processed, gives us information. Data, information and past
experience combined together is known as knowledge.

We have read earlier that the knowledge base stores knowledge acquired
from different experts of a particular domain. More the knowledge in the
knowledge base, more precise is the result from the expert system.

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.

Factual knowledge is based on facts and accepted by knowledge engineers

in that domain.

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.

FIGURE 9.6 Creation of Knowledge Base

CASE STUDY: Working of MYCIN as an expert system.

Step 1: Knowledge base is created with expert knowledge about bacterial
infection—their causes, symptoms, treatment, etc. (refer Fig. 9.6).

Step 2: The knowledge base is regularly updated. In order to test it, the
doctor gives the expert system a new problem of identifying the presence of
the bacteria by inputting the details of a patient, including the symptoms,
current condition, and medical history.

Step 3: The system prompts the patient to provide general information like
gender, age, etc.

Step 4: The system applies if-then rules using the inference engine and the
facts stored within the knowledge base.

Step 5: The output is displayed to the patient through the user interface.

9.1.4 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

 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.

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

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.

9.1.5 Capabilities of the Expert System

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

9.1.6 Advantages of Expert Systems

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.
9. They improve the quality of decisions.
10. They gather scarce expertise knowledge in good amount and use it
efficiently.

9.1.7 Limitations of Expert Systems

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.

9.1.8 Applications of Expert Systems

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.
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.

9.1.9 Expert System Technology

Expert systems technologies include the following.

1. Expert System Development Environment: The development

environment of an expert system includes hardware and tools like
1. Workstations, minicomputers, mainframes.
2. High level symbolic programming languages (like LISt Programming
(LISP) and PROgrammation en LOGique (PROLOG)).
3. Large databases.
2. Tools: The cost of developing an expert system gets reduced to a large
extent with the help of tools that help developers by providing,
1. Powerful editors
2. Debugging capabilities with multi-windows
3. Rapid prototyping
 4. Having in-built definitions of model, knowledge representation and
inference design.
3. Shells: A shell is an expert system without knowledge base. It facilitates
developers by providing modules for knowledge acquisition, inference
engine, user interface and explanation facility. Two popularly used shells
are JESS and Vidwan. While Java Expert System Shell (JESS) provides
Java API for creating an expert system, Vidwan, enables knowledge
encoding in the form of IF-THEN rules.

9.1.10 Development of Expert Systems

The process of developing an expert system is iterative. The steps to

develop such a system are described below.

Identify Problem Domain

The first step to develop any system is to identify the problem that it intends
to solve. Ensure that this problem is solvable by an expert system. If the
problem can be solved, identify the human experts in the task domain and
analyse the cost-effectiveness of the system. Knowledge engineer(s) and
domain experts work in coherence to define the problem and list the
characteristics of the system.

Design the System

To design the expert system, first identify the technology (hardware, tools
and shells) that will be used. Understand how well the system will be
integrated with the other systems and databases. Prepare a framework for
representing concepts in domain knowledge.

Develop the Prototype

In this step, the knowledge engineer acquires domain knowledge from the
human experts and represent it in the form of If-THEN-ELSE rules. The
knowledge expert also determines how heuristic knowledge could be
integrated in the reasoning process and what type of explanation would be
useful.

Test and Refine the Prototype

The knowledge engineer tests the prototype to identify any deficiencies in

performance or results by using sample cases. Testing is also done along
with end-users to unleash additional errors or discrepancies. The results of
testing are then used to refine the system.

Develop and Complete the System

Once the expert system is developed and tested, it is then tested along with
other elements of its environment, including end users, databases and other
information systems to ensure that it interacts in the prescribed manner and
smoothly integrates with them to render the desired functionality. At this
step, the expert system is clearly documented and users are trained for its
usage.
An expert system is also used for information management and help desks
management.

Maintain the System

Maintenance is an ongoing process. In this phase, the knowledge base is

regularly reviewed and updated. New requirements from the users or
blending with new interfaces of other information systems as those systems
evolve are steps that need to be done time to time.
TABLE 9.1 Traditional System vs Expert System
TABLE 9.2 Human expert vs Expert System

9.2 Internet of Things

Although the concept was not named until 1999, the Internet of Things
(IoT) has been in development for decades. 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. 9.8).
 Credit: 363868 / Shutterstock

FIGURE 9.8 Interconnected Things

If the thing, in the IoT, is a person, then he can have heart monitor implant;
if it is an animal, it can have a biochip transponder, if it is an automobile, it
can have sensors to generate alert alarms (e.g., when the pressure in a tyre is
too low) and if it is any other natural or man-made object, then it is
assigned an IP address and provided with the ability to transfer data over a
network.

The huge availability of address space in IPv6 has led to the development of
the IoT. With more IP addresses available, more devices can be connected
to exchange information with each other.

Today, IoT is being widely used for precision agriculture, building

management, healthcare, energy and transportation. The first Internet
appliance was a Coke machine at Carnegie Melon University in the early
1980s when the programmers succeeded in connecting the machine over the
Internet to check the status of the machine and determine the availability of
a cold drink. This check helped them to save their time as they did not have
to go to the machine to buy Coke when it was not available in the machine.

Later, other devices like cell phones, coffee makers, washing machines,
headphones, lamps, wearable devices and almost anything you can think of
could be connected. It is interesting to know that in 2022, the market for the
IoT is expected to grow 18% to 14.4 billion active connections and by
2025, there will be approximately 27 billion connected IoT devices. So, we
can conclude that, IoT is a giant network of connected ‘things’ (including
people) which will have relationships between people-people, people-things
and things-things.

In simple terminology, IoT is the concept of connecting any devices over

the Internet.

9.2.1 Examples of Applications of IoT

Whenever we read about a new technology, the first question that comes to
our mind is how is this technology going to help us? While with IoT,
anything that can be connected, will be connected, but the question is why
should we want these devices to be connected? There are many examples to
justify the answers to these questions.
Case 1: Imagine you are going for meeting and your car tells you the best
route to take. If there is a traffic jam, then your car sends a text message to
the other party, notifying them that you will be late.

Case 2: Imagine that the moment your alarm rings to wake you, a message
is sent to the coffee machine to start brewing coffee for you.

Case 3: The smart watch that your wear in office tells you when and where
you were most active and productive.

Case 4: You have a self-driving car with complex sensors to detect objects
in their path.

In 2019, IoT devices generated an estimated 18.3 zettabytes of data, which

is expected to grow to 73.1 ZB by 2025.

Case 5: A smart football that can track how far and fast it is thrown and
record those statistics via an app for future training purposes.

Case 6: You have a smart refrigerator that can text message you that milk
packets are over in the fridge and you need to buy them before you come
home. Or, your fridge checking the expiry date of an ice-cream and
notifying you that you should not eat it now.
Case 7: Imagine that you have an AC installed in your home which is
connected with your smartphone. On a very hot day, you can instruct your
AC to start and set the room temperature to 18 degrees before you finally
reach the house.

In an article, Ashton wrote that if computers knew everything about things

(using data gathered by them) without any help from us, then we would be
able to track and count everything. It would also reduce waste, loss and
cost. We would know exactly when things need to be replaced, repaired or
recalled, and whether they were fresh or past their best.
 Credit: monicaodo. Shutterstock

FIGURE 9.9 A Smart City

In this way, IoT can help organizations save a lot of money through
improved process efficiency, asset utilization and enhanced productivity.
With improved tracking of objects using sensors and connectivity,
companies can better analyse them and make smart decisions. For example,
if you own a car manufacturing company, then you can know which
accessories are the most popular by using sensors to detect which areas in
the showroom are the most popular, and where customers linger longest.
You can even use the available sales data to identify which components are
selling fastest and then automatically align sales data with supply, so that
popular items never go out of stock.

The information collected by IoT devices can be used to detect patterns,

make recommendations, and detect possible problems before they occur.

We have heard that the Indian government is working hard to develop smart
cities. Do you know that smart cities use IoT for efficient utilization of
resources? Figure 9.9 gives an overview of such a city. IoT has the potential
to transform entire cities by solving real problems citizens which are faced
by the people every day. With the proper connections and data, the IoT can
solve traffic congestion issues and reduce noise, crime and pollution.

9.2.2 IoT Products

Big companies like Honeywell, Hitachi, GE, Cisco, AT&T, Apple, Google,
IBM, Microsoft, Skyworks, Iridium Communications, Red Hat, Zebra
Technologies, InterDigital are already playing in the market to realize the
benefits of IoT. Given below are some examples.

Amazon Echo for Smart Home works through its voice assistant, Alexa.
Users can talk to Alexa and give order to perform a variety of functions. For
example, users can tell Alexa to play music, provide a weather report, get
sports scores, order an Uber and do much more.
Fitbit One – Wearables tracks your steps, floors climbed, calories burned,
and quality of sleep. The device wirelessly connects with computers and
smartphones to transmit your fitness data in understandable charts to
monitor your progress.

Barcelona - Smart Cities: The Barcelona city in Spain is one of the

foremost smart cities in the world. It has implemented several IoT
initiatives that have helped to enhance smart parking and the environment.

Amazon Web Services, Microsoft Azure, IBM’s Watson, Cisco IoT Cloud
Connect, Salesforce IoT Cloud, Oracle Integrated Cloud and GE Predix are
popular IoT platforms.

AT&T - Connected Car: AT&T added 1.3 million cars to its network in
the second quarter of 2016. With this, the count of total number of
connected cars rose to 9.5 million.

Thus, we see that IoT allows for virtually endless opportunities and
challenges thereby making it a hot topic for research. According to a report,
nearly $6 trillion will be spent on IoT solutions over the next five years.
FIGURE 9.10 Benefits of IoT

9.2.3 Challenges

Security is the biggest issue that we often come across while talking about
IoT. With billions of devices connected together, concerns of privacy and
data sharing always haunt our minds.

In fact, the protection of sensitive data was ranked as the top concern
among enterprises according to the 2016 Vormetric Data Threat Report.
Hackers try to penetrate connected cars, critical infrastructure and even
people’s homes. Therefore, the main focus of companies is to ensure
security of all the data generated by these devices. Most IoT devices do not
encrypt communications if the data is transferred over a local Wi-Fi
network. If the Wi-Fi network is unsecured, then it opens the gates for
security threats wide open.

IoT devices used on patients as wearable devices, if left unattended (with

technical errors), can be life-threatening for patient.

Another issue with IoT is that massive amounts of data is being generated
by these devices. So, companies need to figure out how they would store,
monitor, analyse and deduce results from this vast amount of data that is
continuously being generated.

FIGURE 9.11 Key challenges in IOT project development

The graph in Fig. 9.11 shows top challenges in the world of IoT. Besides
security, other key concerns are as follows.

1. Scalability: When billions of Internet-enabled devices are connected,

large volumes of data are generated that need to be processed. To process
this huge data, devices need big data analytics and cloud storage for
interpretation of useful data. Moreover, these systems need to be scalable
to accommodate ever expanding data.
2. Interoperability: Technological standards in most areas are still
fragmented and need to be converged to establish a common framework
and standard for the IoT devices.
3. Lack of standardization: Since standardization process is still lacking,
interoperability of IoT with legacy devices is critical which, in turn,
prevent us to move towards the vision of truly connected interoperable
smart objects.
4. Lack of government support: Government and regulatory bodies like
the FDA should set up regulations by forming a standard committee for
safety and security of devices and people.
5. IoT Malware and ransomware: Ransomware can lock out users from
various devices and platforms and still use a user’s valuable data and
information. For example, a hacker can hijack a computer camera and
take pictures. He may also lock a device and then demand ransom to
unlock it and return the data.
6. Connectivity: Poor Internet connectivity is a challenge where IoT
sensors are required to monitor process data and supply information.
 7. Lack of effective and informed government regulations: IoT is a fast-
developing area of technology and many legislators are either not aware
or do not fully understand the workings of IoT, so they are reluctant to
create or enforce regulations.
8. No privacy policies: There are no clear regulations for IoT devices on
which information is private or public. For example, an IoT toy could
collect information from a child without taking consent from the parents.
The toy can then share that data with the manufacturer who could then
share it with others.

IoT devices should be updated regularly. Updates should be small. Also,

IoT devices should be backward compatible with older devices.
9. Bandwidth strain: As the number of IoT devices connected in a
geographical area grows, more text/audio/video data will flow through
those devices, thereby requiring higher bandwidth and placing strain on
the server farms that process all that information.
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. As of
now, a few IoT experts are working in the industry, training new people
even more difficult. In a survey, it was found that 38% of respondents
reported that they lack in-house expertise in IoT network management,
while 27% were facing a dearth of in-house experience to deploy IoT.

CASE STUDY EXAMPLE

Several high-profile IoT security breaches have been reported till date. One
example is the 2015 Jeep Cherokee hack, when two hackers remotely took
control of a car using vulnerabilities in the entertainment system to access
its dashboard functions. The hacker initiated a series of unexpected
disturbances and finally disabled the brakes, causing the driver to swerve
into a ditch.

Wireless IoT SIM cards should automatically update the system, removing
the possibility of users not updating and leaving holes in system security.

CASE STUDY EXAMPLE

Inter-operability is a serious issue. Once it was seen that there was a

problem when an insulin pump communicated only with a particular
operating system. It makes the functionality unavailable for patients who
may have a computer with a different operating system installed in it.

There are issues around the device’s wireless standard. For example, if it
uses Bluetooth, then the device could inadvertently connect to the wrong
nearby device.

CASE STUDY EXAMPLE

A manufacturing plant where a large number of precise, high-resolution
pictures are sent to 3-D printers could slow down the servers or even cause
them to crash if the demands are too great.

Similarly, in a hospital, connected video cameras transfer of thousands of

high-resolution radiology images and hundreds of telehealth video calls
going on in a single day could significantly overload a server farm.

However, despite all odds, IoT is being used by several industries including
manufacturing, defence, transportation, banks, retail, oil & gas mining,
health care, connected home, banks to name a few.

9.2.4 Sensors

A sensor is a device that detects and responds to some type of input that it
receives from the physical environment. This input could be heat, light,
motion, pressure, moisture or any other environmental phenomena. As the
output, sensors usually generate a signal that is converted to human-
readable form and then displayed at the sensor location or transmitted
electronically over a network for reading or further processing.

For example, an oxygen sensor in a car’s emission control system detects

the gasoline/oxygen ratio. If the mixture is not optimal, the balance is
readjusted.
Another example is motion sensors in home security lights, automatic doors
and bathroom fixtures that send out microwaves, ultrasonic waves or light
waves and detect when the flow of energy is interrupted by something
entering its path.

Vision and imaging sensors detect the presence of objects or colours

within their fields of view. They display a visual image of whatever it
detects.

Temperature sensors detect thermal parameters and provide signals to the

inputs of control and display devices. They are used to measure the thermal
characteristics of gases, liquids and solids in many industrial processes.

Radiation sensors sense the presence of alpha, beta or gamma particles and
provide signals to counters and display devices. They are usually used for
surveys and sample counting.

Proximity sensors detect the presence of nearby objects through non-

contacting means within a range of up to several millimetres. They are used
in manufacturing operations to detect the presence of parts and machine
components.

Pressure sensors detect forces per unit area in gases or liquids and provide
signals to the inputs of control and display devices.
Position sensors sense the positions of valves, doors, throttles, etc. and
supply signals to the inputs of control or display devices.

Photoelectric sensors sense objects passing within their field of detection.

They can even detect colour, cleanliness and location, if required. Photo
sensors are commonly used in manufacturing and material handling
automation. For example, they are used for counting, robotic picking and
automatic doors and gates.

Particle sensors sense dust and other airborne particulates and supply
signals to the inputs of control or display devices. Particle sensors are
common in bin and baghouse monitoring.

Motion sensors sense the movement or stoppage of parts, people, etc. and
supply signals to the inputs of control or display devices. They are usually
used for detecting the stalling of conveyors or the seizing of bearings.

Metal detectors sense the presence of metal in a variety of situations

ranging from packages to people.

Level sensors are used to determine the height of gases, liquids or solids in
tanks or bins. The information is then passed as signals to the inputs of
control or display devices.

Leak sensors are used to identify or monitor the unwanted discharge of

liquids or gases. Some leak detectors are used to measure the effectiveness
of the seals in vacuum packages.

Humidity sensors measure the amount of water in the air and convert these
measurements into signals that can be used as inputs to control or display
devices.

Gas and chemical sensors sense the presence and properties of various
gases or chemicals and relay signals to the inputs of controllers or visual
displays.

Force sensors measure various parameters related to forces such as weight,

torque, load, etc. and provide signals to the inputs of control or display
devices.

Flow sensors sense the movement of gases, liquids or solids and provide
signals to the inputs of control or display devices. They are used extensively
in the processing industries.

Flaw sensors detect inconsistencies on surfaces or in underlying materials

such as welds. They are used in a variety of manufacturing processes.

Flame detectors sense the presence and quality of fire and provide signals
to the inputs of control devices. They are used in many combustion control
applications like burners.
Electrical sensors sense current, voltage, etc. and provide signals to the
inputs of control devices or visual displays.

Contact sensors detect physical touch or contact between the sensor and
the object being observed or monitored. They are often used in alarm
systems to monitor doors, windows and other access points. For example,
when a door or window is opened/closed, a magnetic switch provides an
indication to the alarm control unit so that the status of that entry point is
known. They are also used as proximity sensors in robotics applications and
automated machinery.

Non-contact sensors do not require a physical touch between the sensor

and the object being monitored in order to function. Radar guns used by law
enforcement to monitor the speed of vehicles is an example of a non-
contact sensor.

Speed sensors detect the speed of an object or a vehicle.

Ultrasonic sensors detect noise and measure the distance between two
objects. For this, high-frequency sound waves generated by active
ultrasonic sensors are received back by the ultrasonic sensor for evaluating
the echo. Time delay between transmitting and receiving the echo is used to
calculate the distance to an object. Passive ultrasonic sensors are used for
detecting ultrasonic noise present under specific conditions.
 9.3 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.

We now clearly know that while AI simulates human intelligence processes

by machines, IoT, on the other hand, is a system of interrelated computing
devices with unique identifiers and the ability to transfer data over a
network without requiring human-to-human or human-to-computer
interaction. A thing or a computing device in IoT can be a person’s heart
monitor implant, a self-driving car with built-in sensors, a door mat or any
other object that can be assigned an Internet Protocol address and transfer
data over a network.

AIoT enhances the capability of IoT by adding the power of machine

learning algorithms to improve decision-making processes.

9.3.1 How Does AIoT Work?

In AIoT devices, AI is embedded into programs and infrastructure

components, like chips, which are all connected using IoT networks.
Efficient exchange of data amongst all hardware, software and platform
components without human intervention is ensured using pre-written APIs.
 Credit: monicaodo / Shutterstock

FIGURE 9.12 Autonomous Vehicle

IoT devices generate as well as collect data, that is then analysed 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.

AIoT data when processed using Edge AI (discussed later in the chapter)
minimizes the bandwidth needed to move data while avoiding possible
delays to data analysis. This further adds to efficiency of IoT.

9.3.2 Where Does AI Unlock IoT?

IoT is all about implanting sensors into machines. These sensors collect
streams of data through Internet connectivity. All IoT-related services
follow five basic steps to perform the assigned task—Create,
Communicate, Aggregate, Analyse and Act. Since Act depends on Analyse,
the efficiency of any IoT system depends on the Analysis step. AI
technology when used as the analysis stage, improves both efficiency and
productivity manifold.

FIGURE 9.13 AI and IOT functional view

AIoT helps in achieving the following agile solutions:

1. Manage, analyse and obtain meaningful insights from data (refer Fig.
9.13)
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
 9.3.3 Applications and Examples of AIoT

AIoT is extensively used in the following domains.

1. Smart cities: Smart technology such as sensors, lights and meters

collect data that is processed using AI to improve operational efficiency,
drive economic growth and enhance quality of life for residents.
2. Smart retail: Smart cameras are now being used at retail outlets by
managers to identify faces of customers who have scanned their items at
the self-checkout counter before leaving the store.
Moreover, cameras and sensors installed at multiple places to observe
customers’ movement and predict when they will reach the checkout
line. This helps in maintaining dynamic staffing levels to enhance the
productivity of the cashiers and reduce the customer’s checkout time.
Major retailers can use AIoT solutions to grow sales through customer
insights. Data such as mobile-based user behaviour and proximity
detection offer valuable insights to deliver personalized marketing
campaigns to customers while they shop, increasing traffic in brick-and-
mortar locations.
3. Smart home: In smart appliances installed at homes, data collected from
human interaction with devices is analysed to understand user habits to
provide customized support.
4. Manufacturing: These days, manufacturers use smart chips to detect
when a particular equipment is not functioning properly, or when it is the
right time to replace.
5. Hiring: AIoT technology when used to integrate data from social media
and HR-related platforms can help to identify right talent without any
bias.
6. Autonomous vehicles: These vehicles collect real-time data about
nearby vehicles, pedestrians and other objects using multiple video
cameras and sensors. Data is also analysed to monitor driving conditions
and take appropriate decision instantly.
7. Robots: Autonomous robots have multiple sensors installed to gather
data about the environment. AI is then used to make decisions that help
robot to make smart and optimized moves.
8. Healthcare: Medical devices and wearables collect and monitor real-
time health data, such as heart rate. Moreover, healthcare facilities
produce high volumes of data including patient information, imaging and
test results. This information is valuable and necessary to take good care
of the patient care. So, this data needs to accessed and processed quickly
to suggest decision regarding patient’s diagnostic and treatment.
IoT combined with AI not only improves diagnostic accuracy, enable
telemedicine and remote patient care but also reduces the administrative
burden of tracking patient health in the facility.
9. Smart Thermostat Solution: The smartphone integration with smart
thermostat solution can be used by users to check and manage the
temperature from anywhere. Users can set the temperature based on their
work schedule and temperature preferences. For example, Nest’s smart
thermostat solution uses AI-powered IoT.
10. Drone Traffic Monitoring: AIoT can be used by drones to monitor real-
time traffic and make adjustments to the traffic flow to reduce
congestion. While drones can collect and transmit traffic data, AI can be
used to analyse that data and make decisions about how traffic can be
regulated- make diversions, adjustments to speed limits and timing of
traffic lights without human intervention.
For example, the ET City Brain, a product of Alibaba Cloud, uses AIoT
to optimize the use of urban resources. It can detect accidents, illegal
parking and can alter traffic lights to help ambulances reach patients/
hospitals faster.
11. Office Buildings: AIoT is used in smart office buildings. In such
buildings, usually, a network of smart environmental sensors is installed
that detect number of people present in the office and adjust
temperatures and lighting accordingly to improve energy efficiency.
These smart buildings are often accessed through facial recognition
technology using a combination of connected cameras and AI to
compare images taken in real-time against a database to determine who
should be granted access. this technique is also used by employers to
maintain attendance of employees for mandatory meetings.
12. Fleet Management: AIoT is used in fleet management to monitor a
fleet’s vehicles, reduce fuel costs, track vehicle maintenance and identify
unsafe driver behaviour. Through IoT devices such as GPS and other
sensors and an artificial intelligence system, companies are able to
manage their fleet better thanks to AIoT.
 9.3.4 Benefits and Challenges of AIoT

The benefits of AIoT include the following:

1. With AI, operational efficiency of IoT is increased. AIoT devices

analyse data to reveal patterns and insights and automatically adjusts
system operations.
2. On the fly decision-making: With AIoT and edge computing, vital
decisions can be taken instantly. Therefore, data generated by the
devices are analysed to identify points of failure/concerns to make
appropriate adjustments as and when required.
3. Reduced workload: Since data is analysed by AI and not humans, lot
of time and money spent for this time is saved.
4. Scalability: The number of devices connected to an IoT system can be
increased anytime as per requirements to optimize existing processes or
introduce new features. IoT devices range from mobile devices and high-
end computers to low-end sensors. Low-end sensors offer most of the
data. An AI-powered IoT ecosystem analyses and summarizes the data
from one device before transferring it to other devices. This not only
reduces large volumes of data to a handy level but also enables
connecting a large number of IoT devices.
5. Boosting Operational Efficiency: AIoT instantly analyses constant
streams of data to predict the operational conditions and identify the
parameters to be modified to obtain better results. Thus, intelligent IoT
can deduce which processes are redundant and time-consuming, and
 which tasks can be fine-tuned to enhance efficiency. For example,
Google uses AIoT to reduce its data centre cooling costs.
6. Better Risks Management: Pairing AI with IoT facilitates businesses to
understand as well as predict a wide variety of risks and automate
process to provide a prompt response. This helps them to handle
financial loss, ensure employee safety and prevent potential cyber
threats. For example, Fujitsu ensures safety of the employees by using
AI for analysing data coming from connected wearable devices.
7. Triggering New and Enhanced Products and Services: With NLP,
users can easily communicate with devices. With this human-machine
interaction, IoT and AI can together be used to create new products and
services. Even existing products and services can be enhanced by
quickly processing and analyzing real-time data. For example, Rolls
Royce uses AI technologies in the implementation of IoT-enabled
airplane engine maintenance to identify patterns and gain useful
operational insights.
8. Eliminates Costly Unplanned Downtime: In industrial manufacturing
and in some fields like offshore oil and gas exploration, equipment
breakdown can result in costly unplanned downtime. Predictive
maintenance with AI enabled IoT predicts equipment failure well in
advance and schedule orderly maintenance procedures. This reduces the
side effects of downtime. For example, Deloitte, integrates AI and IoT
for the following:
1. Reducing 20%–50% of their time spent in maintenance planning
 2. Increasing 10%–20% of equipment availability and uptime
3. Reducing 5%–10% amount spent as maintenance costs

All this portrays a rosy picture but do not forget that there are always two
faces of the same coin. AIoT may fail. For example, autonomous delivery
robots that fail might cause delays in product delivery; smart retail stores
may lead to accidently stealing of a product if it fails to read a customer’s
face, a road accident can be caused if an autonomous vehicle fail to read its
surroundings, like a red light on traffic signal.

9.3.5 Future of AIoT

Integration of AI and IoT (refer Fig. 9.14) creates a much smarter system
that makes accurate judgments without human intervention. AIoT will get a
tremendous boost with 5G networks that is designed to enable faster
transfer of large data files through higher bandwidth and lower latency.
AIoT can revolutionize supply chains and delivery models and solve
existing operational problems very efficiently. Several businesses have
already adopted AI and IoT as part of their processes and products. In fact,
AI and IoT are the top technologies in which companies are investing
money to increase their operational efficiency and provide a competitive
advantage. The graph in Fig. 9.15 depicts the details.
FIGURE 9.14 An AIoT Machine

FIGURE 9.15 Visualization of emerging trends

It is surprising to note that every day, around one billion gigabytes of data is
being generate by IoT devices. By 2025, it is estimated that 42 billion IoT-
connected devices will be present globally and with growth in number of
devices and network, the data generated is set to rise exponentially. With
more data, there will be more challenges than opportunities.
With AI, IoT networks and devices can learn from past decisions, predict
future activity, and continuously improve their decision-making capabilities
and thus performance. AI allows the devices to ‘think for themselves’ by
interpreting data without the delays and congestion that occur from data
transfers.

9.4 Edge Computing

Edge computing is a distributed information technology (IT) architecture in

which client data is processed at the periphery of the network, as close to
the originating source as possible.

We have seen that data is the lifeline of any business as it provides valuable
business insight and supports real-time control over critical business
operations. Today, every business generates oceans of data, most of it
coming from sensors and IoT devices operating in real time. This virtual
flood of data is also changing the way computations are done. Traditional
computing relies on centralized processing but bandwidth limitations,
latency issues and unpredictable network disruptions poses a big challenge
for hassle free movement of data and results. Edge computing offers a
solution to overcome these challenges.

Edge computing moves some portion of storage and computations out of

the central data centre and closer to the source of the data. This not only
avoids transmission of raw data to central server but also facilitates data
processing and analysis at the point where the data is actually generated—