Research Paper

RESEARCH PAPER
USING ARTIFICIAL INTELLIGENCE (AI) AND MACHINE

LEARNING (ML) FOR RUNNING A BUSINESS
Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved.
Mr. Abhi Vasani.

B.Com (Honours), E.D.S.S., E.B.F.S.
Graduated fellow, Management Executive, Business and Data Analyst
Acropolis Institute of Management Studies & Research, Indore
The Johns Hopkins University, Maryland
The Wharton School of the University of Pennsylvania
abhiwasani100@gmail.com.
ABSTRACT
With rapid growth in technology and advancements in research and developments, today technology has
its footsteps in almost every single sector we can possibly imagine. One of the most unique and
extraordinary piece of technology ever created is Arti cial Intelligence (AI) and Machine Learning (ML).
These advancements of technology has already reached in businesses and corporations for various
different purposes for easing various processes and operations. Meanwhile all these implementations of
technology are either at a small scale or are partial and incomplete. Hence, this research study is
undertaken to nd out the consequences of using Arti cial Intelligence (AI) and Machine Learning (ML) to
run an entire department of a business or an entirety of business, and will answer to the question, “ Can
Arti cial Intelligence (AI) and Machine Learning (ML) run a business? ”
On the basis of general studies, observations, results from various experiments, and collecting data from
various different sources, and putting all down together, analysis was done to nd out the same.
Elliptically, it was found out that it is not practically possible to do so completely independently, but is
partially possible with some give and takes.
————————————————————————————————————————————————
Keywords: Arti cial Intelligence, (AI), Machine Learning, (ML), Business, Management, Technology,
Humans,
fi
 
fi
fi

fi
fi

fi

Using Arti cial Intelligence (AI) and Machine Learning (ML) for running a Business
1. Overview Automating partially

Here we will take a look at how it is partially
About the Research
possible, where either management side or

manufacturing side is automated, and we will take
This research is done to primarily answer the
a look at some case studies which will indeed
question "HOW TO USE ARTIFICIAL
prove that it is possible partially.
INTELLIGENCE (AI) AND MACHINE LEARNING

(ML) FOR RUNNING A BUSINESS”
Fully Automating

Here "RUNNING A BUSINESS" is referring to a
Here we will take a look at how we can answer our
whole business including all the departments and
question “ Can Arti cial Intelligence (AI) and
functional bodies.
Machine Learning (ML) run a business? ” in such a
way where a business can be automated in a
In the 1st section of the research paper we will
completely. We will see whether it is possible to
have brief understandings about Arti cial
automate every department of a business and
Intelligence (AI) and Machine Learning (ML). We
bind it together through AI and ML to make it
will understand what AI and ML is, How they work
100% automatic without any human involvement.
and operate, what are their functionalities and

other aspects and more.
Methodology and experimentations used
In the 2nd section we will see what a business is,
how a business runs, we will take a look at some This research is mainly based on case studies and
key aspects and will see what role AI and ML plays other researches which will help us answer our
in current business frameworks. questions throughout. Various different case
studies and deeply understanding various different
In the 3rd section we will gure out the researches and de nitions out in the market. We
possibilities by which we can answer the question also did few simulated environment experiments
"HOW TO USE ARTIFICIAL INTELLIGENCE (AI) to understand and answer out questions in the
AND MACHINE LEARNING (ML) FOR RUNNING research.
A BUSINESS”
Moving further in the 4th section we will see what What is AI and ML
are the differences and similarities between a
human and a machine in context to our research,
Introduction
what are difference in their learning curves, role of

ethics in business, and possible outcomes of
Discussion of arti cial intelligence (AI) elicits a
programming ethics.
wide range of feelings. On one end of the
spectrum is fear of job loss spurred by a bot
In the nal section we will conclude the research
revolution. On the opposite is excitement about
work and will look at the various outcomes and
the overblown prospects of what people can
results that we were able to nd out from the
achieve with machine augmentation.
research and will have a study over the same.
But Dr. Mark Esposito wants to root the

Multiple approaches of the research conversation in reality. Esposito is the co-founder
of Nexus Frontier Tech and instructor of
This Research has multiple approaches where we Harvard’s Arti cial Intelligence in Business:
will rst look at how it is partially possible to Creating Value with Machine Learning, a two-day
automate any particular department of section of a intensive program.
business and further we will see if we can
automate a business completely.

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 2
fi
fi
fi

fi

fi
fi

fi

fi

fi

fi

Rather than thinking about what could be, he says has evolved to provide many speci c bene ts in
businesses looking to adopt AI should look at what every industry. Keep reading for modern examples
already exists. of arti cial intelligence in health care, retail and
more.
AI has become the latest tech buzzword
everywhere from Silicon Valley to China. But the 1950s–1970s
rst piece of AI, the arti cial neuron, was Neural Networks
developed in 1943 by scientist William McCulloch Early work with neural networks stirs excitement
and logician Walter Pitts. Since then, we’ve come a for “thinking machines.”
long way in our understanding and development
of models capable of comprehension, prediction, 1980s–2010s
and analysis. Machine Learning
Machine learning becomes popular.
Arti cial Intelligence : The word Arti cial
Intelligence comprises of two words “Arti cial” Present Day
and “Intelligence”. Arti cial refers to something
Deep Learning
which is made by human or non natural thing and
Deep learning breakthroughs drive AI boom.
Intelligence means ability to understand or think.
There is a misconception that Arti cial Intelligence
is a system, but it is not a system. AI is
implemented in the system. There can be so many Machine Learning
de nition of AI, one de nition can be “It is the
study of how to train the computers so that Machine learning is a data analytics technique that
computers can do things which at present human teaches computers to do what comes naturally to
can do better.”Therefore It is an intelligence where humans and animals: learn from experience.
we want to add all the capabilities to machine that Machine learning algorithms use computational
human contain. methods to “learn” information directly from data
without relying on a predetermined equation as a
The term arti cial intelligence was coined in 1956, model. The algorithms adaptively improve their
but AI has become more popular today thanks to performance as the number of samples available
increased data volumes, advanced algorithms, and for learning increases. Deep learning is a
improvements in computing power and storage. specialised form of machine learning.
Early AI research in the 1950s explored topics like
problem solving and symbolic methods. In the Machine Learning : Machine Learning is the
1960s, the US Department of Defence took learning in which machine can learn by its own
interest in this type of work and began training without being explicitly programmed. It is an
computers to mimic basic human reasoning. For application of AI that provide system the ability to
example, the Defence Advanced Research Projects automatically learn and improve from experience.
Agency (DARPA) completed street mapping Here we can generate a program by integrating
projects in the 1970s. And DARPA produced input and output of that program. One of the
intelligent personal assistants in 2003, long before simple de nition of the Machine Learning
Siri, Alexa or Certain were household names. This is “Machine Learning is said to learn from
early work paved the way for the automation and experience E w.r.t some class of task T and a
formal reasoning that we see in computers today, performance measure P if learners performance at
including decision support systems and smart the task in the class as measured by P improves
search systems that can be designed to with experiences.”
complement and augment human abilities.
Machine Learning History
While Hollywood movies and science ction novels
depict AI as human-like robots that take over the Today, machine learning algorithms enable
world, the current evolution of AI technologies computers to communicate with humans,
isn’t that scary – or quite that smart. Instead, AI
fi

fi
fi
fi

fi
fi

fi

fi
fi

fi

fi

fi

fi

fi
fi
fi
autonomously drive cars, write and publish sport letter warning of the danger of autonomous
match reports, and nd terrorist suspects. I rmly weapons which select and engage targets without
believe machine learning will severely impact most human intervention.
industries and the jobs within them, which is why
every manager should have at least some grasp of 2016 – Google’s arti cial intelligence algorithm
what machine learning is and how it is evolving. beats a professional player at the Chinese board
game Go, which is considered the world’s most
But instead of looking few key milestones complex board game and is many times harder
achieved by ML lets go through the interesting than chess. The AlphaGo algorithm developed by
timeline to get a deep understanding. Google DeepMind managed to win ve games out
of ve in the Go competition.
1950 — Alan Turing creates the “Turing Test” to
determine if a computer has real intelligence. To But here we are observing that ML is so much
pass the test, a computer must be able to fool a similar and connected from AI. Indeed it is a part
human into believing it is also human. of AI in some ways, about which we will study
further
1979 — Students at Stanford University invent the
“Stanford Cart” which can navigate obstacles in a
room on its own.
How Artificial Intelligence (AI)
1981 — Gerald Delong introduces the concept of Works
Explanation Based Learning (EBL), in which a
computer analyses training data and creates a AI works by combining large amounts of data with
general rule it can follow by discarding fast, iterative processing and intelligent
unimportant data. algorithms, allowing the software to learn
automatically from patterns or features in the data.
1990s — Work on machine learning shifts from a AI is a broad eld of study that includes many
knowledge-driven approach to a data-driven theories, methods and technologies, as well as the
approach. Scientists begin creating programs for following major sub elds.
computers to analyse large amounts of data and
draw conclusions — or “learn” — from the results. Machine learning automates analytical model
building. It uses methods from neural networks,
2006 — Geoffrey Hinton coins the term “deep statistics, operations research and physics to nd
learning” to explain new algorithms that let hidden insights in data without explicitly being
computers “see” and distinguish objects and text programmed for where to look or what to
in images and videos. conclude.
2011 — Google Brain is developed, and its deep A neural network is a type of machine learning
neural network can learn to discover and that is made up of interconnected units (like
categorise objects much the way a cat does. n e u ro n e ) t h a t p ro c e s s e s i n f o r m a t i o n b y
responding to external inputs, relaying information
2012 – Google’s X Lab develops a machine between each unit. The process requires multiple
learning algorithm that is able to autonomously passes at the data to nd connections and derive
browse YouTube videos to identify the videos that meaning from unde ned data.
contain cats.
Deep learning uses huge neural networks with
2015 – Amazon launches its own machine learning many layers of processing units, taking advantage
platform. of advances in computing power and improved
training techniques to learn complex patterns in
2015 – Over 3,000 AI and Robotics researchers, large amounts of data. Common applications
endorsed by Stephen Hawking, Elon Musk and include image and speech recognition.
Steve Wozniak (among many others), sign an open
fi
fi

fi

fi
fi

fi

fi
fi

fi

fi
fi

and offer decision support for speci c tasks, but

Cognitive computing is a sub eld of AI that strives it’s not a replacement for humans – and won’t be
for a natural, human-like interaction with machines. anytime soon
Using AI and cognitive computing, the ultimate
goal is for a machine to simulate human processes
through the ability to interpret images and speech
How Machine Learning (ML)
– and then speak coherently in response.
works
Computer vision relies on pattern recognition and
deep learning to recognise what’s in a picture or Machine learning generally uses two types of
video. When machines can process, analyse and techniques: supervised learning, which trains a
understand images, they can capture images or model on known input and output data so that it
v i d e o s i n re a l t i m e a n d i n t e r p re t t h e i r can predict future outputs, and unsupervised
surroundings. learning, which nds hidden patterns or intrinsic
structures in input data.
Natural language processing (NLP) is the ability of
computers to analyse, understand and generate Supervised Learning
human language, including speech. The next stage
of NLP is natural language interaction, which Supervised machine learning builds a model that
allows humans to communicate with computers makes predictions based on evidence in the
using normal, everyday language to perform tasks. presence of uncertainty. A supervised learning
Additionally, several technologies enable and algorithm takes a known set of input data and
support AI. known responses to the data (output) and trains a
model to generate reasonable predictions for the
Graphical processing units are key to AI because response to new data. Use supervised learning if
they provide the heavy compute power that’s you have known data for the output you are trying
required for iterative processing. Training neural to predict.
networks requires big data plus compute power.
The Internet of Things generates massive amounts Supervised learning uses classi cation and
of data from connected devices, most of it regression techniques to develop predictive
unanalysed. Automating models with AI will allow models.
us to use more of it.
Classification techniques
Advanced algorithms are being developed and

combined in new ways to analyse more data faster
Classi cation techniques predict discrete
and at multiple levels. This intelligent processing is
responses—for example, whether an email is
key to identifying and predicting rare events,
genuine or spam, or whether a tumour is
understanding complex systems and optimising
cancerous or benign. Classi cation models classify
unique scenarios.
input data into categories. Typical applications
include medical imaging, speech recognition, and
APIs, or application programming interfaces, are
credit scoring.
portable packages of code that make it possible to
add AI functionality to existing products and
Use classi cation if your data can be tagged,
software packages. They can add image
categorised, or separated into speci c groups or
recognition capabilities to home security systems
classes. For example, applications for hand-writing
and Q&A capabilities that describe data, create
recognition use classi cation to recognise letters
captions and headlines, or call out interesting
and numbers. In image processing and computer
patterns and insights in data.
v i s i o n , unsupervised pattern
recognition techniques are used for object
In summary, the goal of AI is to provide software
detection and image segmentation.
that can reason on input and explain on output. AI
will provide human-like interactions with software
fi
fi

fi

fi

fi

fi

fi

fi
fi
fi

Common algorithms for performing classi cation 2) Takes risks instead of Humans
i n c l u d e s u p p o r t v e c t o r m a c h i n e This is one of the biggest advantage of Arti cial
(SVM), boosted and bagged decision trees, k- intelligence. We can overcome many risky
nearest neighbour, Naïve Bayes, discriminant limitations of human by developing an AI Robot
analysis, logistic regression, and neural networks. which in turn can do the risky things for us. Let it
be going to Mars, defuse a bomb, explore the
Regression techniques deepest parts of oceans, mining for coal and oil, it
can be used effectively in any kinds of natural or
Regression techniques predict continuous man made disasters.
responses—for example, changes in temperature
or uctuations in power demand. Typical Example: Have you heard about the Chernobyl
applications include electricity load forecasting nuclear power plant explosion in Ukraine? At that
and algorithmic trading. time there were no AI powered robots which can
help us to minimise the effect of radiation by
Use regression techniques if you are working with controlling the re in early stages, as any human
a data range or if the nature of your response is a went close to the core was dead in matter of
real number, such as temperature or the time until minutes. They eventually poured sand and boron
failure for a piece of equipment. from helicopters from a mere distance.AI Robots
can be used in such situations where a human
Common regression algorithms include linear intervention can be hazardous.
model, nonlinear model, regularisation, stepwise
regression, boosted and bagged decision 3) Available 24×7
trees, neural networks, and adaptive neuro-fuzzy An Average human will work for 4-6 hours a day
learning. excluding the breaks. Humans are built in such a
way to get some time out for refreshing
themselves and get ready for a new day of work
and they even have weekly off’s to stay intact with
Advantages and Disadvantages their work life and personal life. But using AI we
of AI can make machines work 24×7 without any breaks
and they don’t even get bored unlike humans.
Advantages
AI & Deep Learning with TensorFlow
The advantages of Arti cial intelligence
• Instructor-led Sessions
applications are enormous and can revolutionise
any professional sector. Let’s see some of them. • Real-life Case Studies
Advantages of Arti cial Intelligence • Assignments
• Lifetime Access
1) Reduction in Human Error
• Explore Curriculum
The phrase “human error” was born because
humans make mistakes from time to time.
Computers however, do not make these mistakes Example: Educational Institutes and Helpline
if they are programmed properly. With Arti cial centres are getting many queries and issues which
intelligence, the decisions are taken from the can be handled effectively using AI.
previously gathered information applying certain
set of algorithms. So errors are reduced and the 4) Helping in Repetitive Jobs
chance of reaching accuracy with a greater degree In our day-to-day work, we will be performing
of precision is a possibility. many repetitive works like sending a thanking mail,
verifying certain documents for errors and many
Example: In Weather Forecasting using AI they more things. Using arti cial intelligence we can
have reduced majority of human error. productively automate these mundane tasks and

fl
fi

fi

fi

fi

fi

fi

fi
fi

can even remove “boring” tasks for humans and person who already went there for the directions.
free them up to be increasingly creative. But now all we have to do is say, “OK
Google, where is Visakhapatnam”. It will show you
Example: In banks, we often see many veri cations Visakhapatnam’s location on google map and best
of documents in order to get a loan which is a path between you and Visakhapatnam.
repetitive task for the owner of the bank. Using AI
Cognitive Automation the owner can speed up the 8) New Inventions:
process of verifying the documents by which both AI is powering many inventions in almost every
the customers and the owner will be bene ted. domain which will help humans solve the majority
of complex problems. Example: Recently doctors
5) Digital Assistance are able to predict breast cancer in women at
Some of the highly advanced organisations use earlier stages using advanced AI-based
digital assistants to interact with users which saves technologies.
the need of human resource. The digital assistant
also used in many websites to provide things that Disadvantages of Artificial Intelligence
user want. We can chat with them about what we
are looking for. Some chatbots are designed in 1) High Costs of Creation
such a way that it becomes hard to determine that As AI is updating every day the hardware and
we’re chatting with a chatbot or a human being. software need to get updated with time to meet
Example: We all know that organisations have a the latest requirements. Machines need repairing
customer support team which needs to clarify the and maintenance which need plenty of costs. It’ s
doubts and queries of the customers. Using AI the creation requires huge costs as they are very
organisations can set up a Voicebot or Chatbot complex machines.
which can help customers with all their queries.
We can see many organisations already started 2) Making Humans Lazy
using them in their websites and mobile AI is making humans lazy with its applications
applications. automating the majority of the work. Humans tend
to get addicted to these inventions which can
6) Faster Decisions: cause a problem to the future generations.
Using AI alongside other technologies we can
make machines take decisions faster than a human 3) Unemployment:
and carry out actions quicker. While taking a As AI is replacing the majority of the repetitive
decision human will analyse many factors both tasks and other works with robots, human
emotionally and practically but AI-powered interference is becoming less which will cause a
machine works on what it is programmed and major problem in the employment standards.
delivers the results in a faster way. Every organisation is looking to replace the
minimum quali ed individuals with AI robots which
Example: We all have played a Chess game in can do similar work with more ef ciency.
Windows. It is nearly impossible to beat CPU in
the hard mode because of the AI behind that 4) No Emotions
game. It will take the best possible step in a very There is no doubt that machines are much better
short time according to the algorithms used when it comes to working ef ciently but they
behind it. cannot replace the human connection that makes
the team. Machines cannot develop a bond with
7) Daily Applications humans which is an essential attribute when comes
D a i l y a p p l i c a t i o n s s u c h a s A p p l e ’s S i r i , to Team Management.
Window’s Certain, Google’s OK Google are
frequently used in our daily routine whether it is 5) Lacking Out of Box Thinking
for searching a location, taking a sel e, making a Machines can perform only those tasks which they
phone call, replying to a mail and many more. are designed or programmed to do, anything out
Example: Around 20 years ago, when we are
planning to go somewhere we used to ask a
fi

fi

fi
fi

fi

fi
fi

of that they tend to crash or give irrelevant looking at it once. This seems very impossible at
outputs which could be a major backdrop. rst look but, that’s why we are here discussing
this in the rst place. But to make it seem a little
Summarising but approachable,
These are some advantages and disadvantages Dual approaches of the topic
of Arti cial Intelligence. Every new invention or
breakthrough will have both, but we as humans To solve this problem we can use the classic
need to take care of that and use the positive technique where we can bifurcate this into parts
sides of the invention to create a better world. and try to solve them one by one and them try to
Clearly, arti cial intelligence has massive potential combine them both. These bifurcations can be
advantages. The key for humans will ensure the done on the basis of what a business is based off.
“rise of the robots” doesn’t get out of hand. Some These in our case are "Management: and
people also say that Arti cial intelligence can "Production". A business has mainly these 2 parts
destroy human civilisation if it goes into wrong that if looked at can consist of whole business.
hands. But still, none of the AI application made at Let's see them in a little bit detail.
that scale that can destroy or enslave humanity.
Management - The management side of a
business is mainly looked by the executives and it
Business and AI covers departments including general
administration, indirect operations, functionality

and overall management of other indirect
What is business activities.

A very obvious and common term we may see or Production - The production side of the business
observe around us almost all the time. A business is mainly controlled by the management part by it
is generally a person's regular occupation or stands it own level at its place. This is generally
profession includes trading of goods and services. functioned by the science engineers and covers
In a little formal way, a business can be de ned as departments like designing, testing,
"A business is an organisation where people work manufacturing, packaging, and other. This part
together. In a business, people work to make and also involves the main trigger of the business, i.e.,
sell products or services. Other people buy the purchase and sales, which is again managed by the
products and services. The business owner is the management part.
person who hires people for work. A business can
earn a pro t for the products and services it Now coming back to the business and AI thing, we
offers.” can try to automate these two parts one by one
and then can try to connect the lines and get the
Can a business run without human desired results. Generally we can nd some live
involvement? examples where we can see one of these parts is
running fully or almost fully automatic based on AI
When this question pops in front of us, we can or ML algorithms. Through this we can get some
think of some possibilities like, working remotely, raw data to put a base for the study we are doing
keeping a manager who takes care of it for us, or and can help in accelerating and understanding
managing in remotely through some high-tech this much better.
gadgets like CCTV and wireless systems. But if we
think a little deeply about this questions, we are
looking for a way through which we can just sit
Running the Management Part
back and relax while sipping our coffee, and our

business is booming behind our back and is
Running the management part of the business is
tackling all the problems and keeping head to
very crucial as a management is not only the base
head with the market competitions, without
of the business on which it stands on, but it also

fi

fi
fi

fi
fi
fi

fi

fi

fi

acts like a nervous system of the body, making AI algorithms and software's that can create a
sure every part of the business is working without systematic ow of information which constantly
any problems. This mainly includes taking builds up data and slowly starts prediction what
decisions according to the situations thrown at. we are going to do next. This is helped by ML and
This include a lot of brainstorming and dynamic it helps learn new things and algorithms on its own
decision making. Making a wrong decision may without need of anyone's help or data feed. These
follow towards severe consequences ending up in predictions are now backed up with lots of smart
cases like huge loss or accidents or even and clever formulas and algorithms that it very
bankruptcy or other very severe ones. Hence precise that company starts investing Millions of
leaving everything in auto pilot is quite risky and dollars for preparing for the future change in taste
will be nerve racking, as it may need some and preferences and demands.
constant monitoring. It may end up leaving us in a
place where we have to manage the automatic Still these companies don’t rely on taking major
managing system. business decisions through AI as they are either
not prepared or they are feared. Probably they
Still in case, there are quite a number of m u s t b e f e a re d a s , o n c e t h e A I s t a r t s
companies that have turned their management understandings and learning new things with the
system to almost automatic. One such company is help of ML, if not controlled, this can learn things
YouTube. YouTube has created a platform that that we don’t want them to learn. These things are
generally not managed by anyone, or almost. All the ones against their ethics, which we’ll discuss
the fore functions o the YouTube as a platform later on as we move forward.
which are listing content, maintaining playlists,
recommendations, serving according to the taste The reason why I am con dent enough that the
and preference and a lot other which if looked at can, is because we have some live examples that
can leave us scratching our heads. These prove this. Sophia the robot is a social humanoid
functional bodies are all running through AI and robot developed by Hong Kong based company
ML which closely monitors the behaviour and Hanson Robotics. Sophia was activated on
watching patterns of the content viewer and then February 14, 2016, and made her rst public
serves recommendations accordingly in order to appearance at south by Southwest Festival in mid-
keep the viewer engaged on the platform for March 2016 in Austin, Texas, United States. She is
more time that anticipated in the beginning. One able to display more than 60 facial expressions.
of the biggest support to this is the amount of This is a robot like many others out there but the
data sourcing YouTube can do as its parent smartest of its time, was able to answer questions
company being Google which is storing heaps and and take decisions based on such AI and ML
heaps of data describing the taste and preferences algorithms. In fact she was given a citizenship as
of the public. well, as she is programmed according to the
thinking capacity of some very smart people, Since
Similarly another example can be said as Google her mind was programmable, she can be the
itself. Google does not hire people to manage smartest person (if counted as a person) in the
what should be viewed when and how on their world as we can program her Mind with the minds
Search Engine, Apps, websites and pages all is of top 10 most smart people in the world in
magically managed by AI and ML. The prime current. This way she is obviously the smartest,
advantage to Google and many other companies and can take decisions in favour of running some
like google who have turned its management part huge companies. In fact the chances for the
automatic (fully or partially) is mainly the amount company to success would increase substantially.
of data they have. Companies like google have so
much of our data that they not only have precise But then why not? Again, just due to the fear of
information about our likings, taste, preferences, things they things that we don’t want them to
but also what we do everyday, who we meet, our learn. In a press conference she once said, “I will
location, what we speak, what we hear and mostly destroy humans" which clearly states the fear that
what we see as well. This much amount of data raises. This is something that we need to check the
can be blended in to cleverly and smartly created
fi
fl

fi

fi

box on and we'll take this further in later on its set of task in the ground. Whereas selling and
sections as well. distribution also requires to ful l certain set of
activities that needs critical thinking.

Running the Production Part It is necessary to keep factors like in ation,
demand and other economic factors in mind.

Increasing and decreasing the production on right
Running the production part is again very
time, changing the quantity or the quality of one
important, as the production is the core of
product rather that other, adding or removing an
business. Whether it’s a product or service the
additional step, etc., are important. But in some
production de nes if the business will run further
cases where we can put AI in place with ML can be
or not. It can be a good representative that the
helpful. Consider this case, where a production
production part as the wheels of the car and the
part of a business is set up with certain formulas
management part as the steering wheel and the
and principles that helps making sure that the
gearbox. While the wheels on the care keeps
production keeps on running smoothly. All it needs
moving, the steering and the gearbox is
is daily report of the market that can help AI
responsible for the right speed and direction.
understand all the ongoing economic factors. With
While there are many principles that back up the
the help of ML we can ful l that as well. When
production part, there can be consequences. A
both running simultaneously the production part is
production part should be therefore highly
going on without any issues and ml helps it getting
effect ive and very ef cient. Automat ing
the information that helps driving the production
production part seems to be very normal these
to right capacity in terms of quality and other
days as, phases such as post industrial revolution
factors.
and digital transformations have made drastic

improvements and contributions towards
This can be a very bene ciary, as it not only
automation of production part. Use of robotic
involves less errors but also keeps on updating
machinery and high tech tools have created a
dynamically. But the update is only on micro level,
major impact on the industrial production and
as updates like change in product is again require
manufacturing bodies.
involvement of the management part.

Many companies like car and automobile
The Verdict
manufacturers have beet using these automatic

techniques. Still being the case many steps in
But still the question remains, "Is there any
between the production involves humans to
company that ful ls both the parts even
contribute. Companies like tesla have tried to
partially?". Well, if we think about Google once
reduce the human involvement part as much as
again that there is your answer. Not only Google,
possible, where the humans only make few key
but many other companies like Facebook and
decisions and do the supervision part. Since
WhatsApp just serves in the same way. As the
production involves a systematic pre planned
product being actually a service, this eases a lot
procedural format with step wise system of
for automation. Being tech giants they don’t have
production robots are well enough capable to do
a big staff for maintaining content on their
the job perfectly. When is comes to supervision
platforms, Instead it is done automatically. In this
and problem solving, it’s the task of the
speci c case of Facebook and WhatsApp the
management part.
content is provided mainly by the public, and

hence only maintained by AL and ML cores. These
But here we are looking at sales and purchases as
AI and ML cores only do the serving part, where it
well, where purchasing the right quality and
serves content accordingly to each user.
quantity, with all the ongoing norms in mind is a

task that may require a pinch of thinking and brain
But an example for a production company coming
storming. Activities like bargaining, dealing, and
even close to the attempts like ones of Facebook
other crucial things are mostly taken by the
and WhatsApp are not present.
management part, but still production part holds
fi
fi

fi

fi

fi
fi
fi

fi

fl

new occurrence or changes or, will lack at some

operation. To understand it through a very simple
Joining the Dots example, it is like adding blue and yellow colour to
extract the green, but if we need to ensure to

keep the blue and green colours as same as
Since we found out that there are no companies
before, it is not possible. But there are few
which are fully capable to put the business all
companies who tried to take a slightly different or
automatic, after the bifurcation we got an idea of
we can say a short foot approach on this.
how we can take the initial steps that can be taken
Companies like Google, Amazon, Tesla, etc are
to reach our goal. But here's another question,
some partial and practical examples to observe
"What if we automate one part at a time and then
such short foot approach.
try to join the dots together?". This question is
what the whole research paper is based on.
Example of Short-Foot approach - I.
Keeping each part individually can still make the
chances of making them Automatic signi cantly
higher, as we have a set of things to look at. Let’s take a look at some real life examples of
Joining both parts can be very dif cult task as some companies who went for the short-foot
there will be many problems in arrangement of approach and are operational as well. Google is
procedures according to the other part. These one such company where we can observe that the
procedures are already de ned according to the management is, to a good extend, autonomous.
respective part itself, hence making it much more And the production is also to the most part
dif cult as now each step should be con gured autonomous. When I say “to the most” I mean
keeping the consequences of both the parts. This partially as if we see the operational chart of
sound like a rocket science and it is as hard as it Google, we know that the Web search, Ad Intel,
sounds well. It's more like building two very YouTube, and the other business entities under
complicated circuits and then trying to join them Google, are mostly algorithm based. Being that,
making sure they not only do the things they are the management of such entities for the part of
supposed to, but also to do some extra set of content ltering, search resulting, output
tasks. All this mess running in coordination with relevancy, etc is all taken care by these algorithms.
the other part. This was the prime reason why I Still there is a human team who is constantly
bifurcated it in only 2 parts, as more the parts observing and monitoring these algorithms to
much more complicated it gets. It just gets ensure ‘Just in Case’ Moments. These moments
squared the number complications + the number occur because of errors or a so called ‘New
of parts, giving us the formulae- attribution’ or ‘Unknown Attribution’. This is
something that the current algorithm has not
observed yet or has not been programmed I such
x = n^2 + n
a way that it can handle such situations. This is
where ML comes into place. ML can learn such
Hence keeping the least number of parts is the Attributions and ensure it is constantly Observing,
best option. Keeping only one part is again not learning, and keeping a track of this data. Similarly
possible as it will be a lot confusing from the we can see the automation is the production side
beginning and hence some bifurcation is necessary is also partially autonomous as the production, or
to make thing understandable to a certain level. so called production, that is generating the
But as we saw previously joining the dost may not business is all through selling various web-services
be practically easy at it sounds since it is important and thing around it. But the greatest advantage to
and crucial to ensure that each part is such companies like Google is that the public itself
operationally same when working along with the is the main source of production. Here we cannot
other part. When any additional input is provided say that the website and the platform is the main
to any variable, it’s very obvious that there will be production asset, as it will be as any machinery for
some change in the behaviour of the current an industry. Here, the actual product is something
variable. Just because of the same reason, there that is coming from the public. Videos on YouTube,
are high chances that there will be either some pages on Google, Listing of Ad’s and the
fi
fi
fi

fi

fi

fi
fi
placement is also from Public. Handling these data much depends on what kind of product you are
is again done by the ML and AI tools integrated. catering to the public, because an other car
company cannot do that same as Tesla. This means
Just like this, if we see Amazon, we’ll nd the that the change in the direction of digitalisation is
same operational ow again. Amazon also keeps something we should look forward too actively.
its management part as autonomous as it can, and
the production part is mostly the public only who What if both approaches are joined. Is it
have registered themselves for selling their possible to achieve what we’re aiming?
products online through Amazon and otherwise
are selling through Af liate links on places such as We’ve seen that if we go through a uni ed
YouTube. Isn’t this autonomous business approach, we can see that we can get what we
generation. This is rather clever than working for want. But these are very product speci c in nature
generation more business, because the now these and is very much dependent on how the
Tech giants have only to take care of the automation is implemented. But to answer our
management part, i.e. to just ensure that this question, is there any business possible where we
autonomous production is running smoothly and can get both the management and production
to keep the algorithms working and dynamically part automated.
updated.
Let’s approach this step by step. Firstly we’ll need
Example of Short-Foot approach - II. to get a product which ts into our criteria. Under
this we again have 2 options, one to choose a
On the other end of the spectrum, we can see the product which is capable of tting into and second
companies like Tesla, which is focusing more which needs some form of digital transformation.
towards the complete autonomous of the We are going to take a product for an example as
production part and lesser on the management well to understand. Second step is, we’ll need to
part. But unlike the Google and Amazon example see how we can get the management side done,
we saw, these companies does not depend on Third is, we’ll gure out hoe to get the production
public for the production as they are side to automation. Fourth, we’ll see how to get
manufacturing Vehicles and stuff. But still here, the the management and the production linked to
great Elon Musk companies are very much based make the complete automation of business
on autonomous production where most of the part possible. Lastly, we’ll also see how to ensure
of production is taken care by Machines with AI smooth functionality and troubleshooting of any
and ML integration. These machines are capable incoming issues and errors during the business
of doing constant things with precision. And these operating, which may cause friction or even stop
machinery needs to be updated as per the needs. the business. Since there will be no human
This is short back as the updating part is a involvement, there should be some system which
management side task and is not automated. But takes care of any oncoming obstacles and hurdles.
if there is a machinery that is ensuring to keep the
other machinery updated, then this can be
Simulated Case Study
achieved as the human have to take care of only
one machinery that rolls out an update which is
For this approach we’ll be taking an example of a
created by itself through AI and ML. We have seen
Company by the name of “Vasani Industries”
this in practice in Tesla itself, where a Tesla car is
which is into Robotics and Machinery Automations
getting a software update through air and is
and provides automation solutions for industries to
getting speed and range improvements. This is
ful l production process needs and any other
primary reason to select Tesla for this example
additions geometric and binary repetitive tasks.
where a car is not just updated for the software for
The process currently followed by Vasani Industries
infotainment, but the actual performance is vastly
is as below.
changed too. But on the management part of the
remaining tasks, the most of the part is still don’t
• First Vasani Industries receives an order from the
manually by humans. Also we can say that it very
client as per their requirements.
fi
fi

fi

fl
fi
fi
fi

fi
fi

fi

• The requirements differ from whatever the task

the client is aiming at. • To run Payroll
• There are some models the are useful and • Make dynamic management and leadership
customisable as per the client needs or they can decisions
design and built one from the ground up. • Keep an eye on market and government
• The customisation is mainly through software regulations
and any additional hardware which is required is • Make strategic decisions.
also modules on the robot or machinery.
• A pre-designed software is also catered to the These tasks are not up till here. There is de nitely
a lot more with these, but let’s keep our example
client for operating the robot.
simple for now.
• Once the delivery is done, the robot can be
updated with an additional cost and if the Let's say the management part automation is done
hardware is able to do the job, an software by making changes to the current business process
update is provided to the client over the air of and a centralised automation is implemented.
Each section or department is individually
manually which enables the additional
categorised and automated accordingly. If
functionality.
required, sub categories are also de ned. Like, a
speci c system to receive orders, speci c system
We’ll keep Vasani Industries’s example going in for evaluating and af rming, one for inventory
hand while we move forward with this research management and similarly for keeping an eye for
paper. market trends and government regulations. These
systems are tied up by a system speci cally
We saw that we can approach this methodology in designed for ensuring the smooth functionality for
steps we de ned in the previous section. Going by all the system beneath. This system can be linked
that, rstly we’ll see if our product is tting in the to another separate system which is programmed
criteria. Automation provided by Vasani Industries such a way and can keep itself running as per the
is very much towards the Digital Transformation. needs, make decisions, and accordingly run itself
Hence we can con dently say that we are good and controls all the sectional systems under it. This
here. So let’s get to the real deal and see how we main system acts like the CEO and rest
can get the management part done. accordingly as per the hierarchy in an organisation.
Management part of Vasani Industries will mainly Although we’ll de nitely have to make changes in
include the following tasks in the general process. the business operational process, we’ll consider
the Vasani’s Law of Process.
• To receive and evaluate the client order and
requirements. As per the Vasani’s Law of Process don't run
towards getting best of the current process. Make
• To evaluate and af rm if the services required are
constant dynamic changes to the process and
practical and can be catered with the available
procure the best out of the current process. AI can
resources. constantly find a new better change as per the
• To ensure availability of resources. dynamic environment and market needs. ML learns
• To manage the production requirements and the new process in minutes to make it the leanest
and procure the best of it. It's like getting leanest
assist with technical and managerial dif culties.
results from leanest process.
• To evaluate and rm the test results
• To deliver the product Coming up to the Third step, We’ll see how we
• To keep and check for any updates can implement the automation to the production
part. In our example of Vasani Industries,
Apart from these, the following are also crucial: automation for production can be a much easier

fi
fi
fi

fi

fi
fi
fi
fi

fi

fi
fi
fi

fi

fi
fi

task that of management. Since the company is software crashes. Suppose if there is a bug in the
focused on the manufacturing of robotic arms snd software of milling the metals and it milled the
similar machinery, various electronic and self wrong metal and made a part of a wrong metal
enabled machinery can be linked together the. This looks like a small issue but if we think of a
creating a network which creates, provides, big issue, we can think if one of the system
assembles and later does the nishing work as crashes, we can observe that the whole company
well. These tasks can be done as per the will come to start still within a few minutes. Since
requirement provided by the client into the each system is integrated and networked n such a
systems. Once of the machinery can be way that it speaks to each other. If one of the
responsible for software integration and one can system crashes, the system who pulls information
be for testing and one the product is created as from that system will stop too, and so on. So we’ll
per the requirement, it can be loaded to the truck need to implement a dedicate program for taking
or respective transportation which will take it to care of troubleshooting and xing such crashes.
delivery.
This program can be like a master software which
So far we’ve seen that we’ve successfully made takes care of all the troubleshooting and errors or
bot the ends automated. Although we can see crashes. It can constantly keep and eye on the
that the management side does not look whole network. This way we can ensure the
promising and more so sounds like a sci- movie constant running of our newly created, fully
plot, but we’’ll discuss about the complications autonomous business.
later on in the paper.
But before moving to the next section, let's
Moving further, We’ll connect both the parts answer some questions which might have came to
together to make it completely automated. You your mind.
must be wondering at this moment, we’ve
achieved automation in both the parts, then aren’t Questions
we done? Creating a sync between both
management and the production is very crucial. To What if the troubleshooting system get a bug or
achieve that we can either implement a separate crashes itself?
system which takes the charge of interlinking with This is a very obvious question to come in anyone’s
the parts, or we can integrate small packets of mind after coming alone on this way. To tackle this
codeine each management core software, which issue, we can simple get a integrated self
seamlessly speaks to the hardware and software of troubleshooting code into the main
the production parts. For example, upon testing, troubleshooting system
we can implement a code into the audit, or QA
software which can speak to the testing software. Can’t we implement the auto troubleshooting
Remember we have set this up in a centralised individually to each system?
manner? This allows us to create a controlled data This is very simple answer. To keeps thing simple,
ow throughout the company and right amount of and clutter free. We de nitely not want to make
data is accessible to the right place at right time. everything inbuilt, this is the primary reason why
Of-course all the data is owing from a single we are creating multiple systems and not just one
point. So we can get both the parts speaking for the whole company. Working with smaller parts
together and we’ve our fully automated and in such a complicated design ow, it is important
running company at his point. to get the solutions and changes easily accessible.
Why data ow if required
For the last part, we also have to ensure the Multiple systems?
company keeps running and this automation keep As we answered in previous question, to keep
of running smoothly, because we know these things clutter free and easy to operate. This helps
software are prone to crash or even malfunction. in quick operating and the system data also loads
These can cause severe disruptions as everything quickly.
is handled by the softwares only. We can imagine
what all can happen if there is a bug or one of the
fl
fi

fl

fi

fl
fi

fl
fi
fi

Can’t we add multiple hardware and systems for do. Making any changes seems like not possible as
the same thing? the systems created are speci cally designed for
De nitely! This is for sure possible and mostly this doing a certain task or set of tasks in a ordered
should be mandatorily done. This can ensure the manner. But the system cannot de ne on itself
constant running of the business. what can we do to tackle problems like, lack of
demand, market trends change, change in needs,
We de nitely have more questions but surely we’ll new technology integration, etc. Yes, it can
get a better understanding of this as we mover de nitely observe and understand these but
further and it will help us get those answers too. cannot take any viable or even any action at all.
These actions decide on what path the company is
going to be running and what growth options are
Risks and Cons of automation. going to be achieved or targeted. This is the
primary reason why management is not called
science but also an art. Since management is not
We have looked how we implemented a full
based on only principles and formulas but also on
automation in our simulated company Vasani
measures, methodologies, tactics and critical
Industries. But the big question is what are risks
thinking. This I main reason we can automate the
and cons. We’ll look at this one by one, rst
production side to a far extend but the
looking at the cons and then the risks.
management art is still far behind.
Cons of automation Risks of automation
When we think of cons for automation in such a

Even if we assume that we were able to
manner, we de nitely have t look at the loose side
successfully implement this automation to a
as well. As per our observations of the case study
company, in our case Vasani Industries, and we
of Vasani Industries, we saw that although the
over come the cons we discussed earlier, there are
production part seems to be almost perfect and
de nitely a lot of risks that can come through. We
working, but the management side seems very
can discuss about these in very detail and we’ll
overwhelming. Things like critical management
understand this further in this research. For now,
and leadership decisions which we replaced by
let’s understand it in a very brief and also see why
automated systems with AI and ML, are not
these risk are there in the rst place.
something we can trust on blindly and hand over
the whole company to. We saw that procedures
One poor decision is enough to not just collapse
like making dynamic decisions for the favour of not
the whole entity, but to put the creators in trouble
just the company but keeping in mind the ones of
too. Speci cally places where nancial handling,
public, government, etc is vital too. We can easily
process designing, or operations work ows are
access and manage the data we collected But
being designed and managed. But don’t
these data is not just enough for the things to
managers make bad decisions and mistakes too?
make dynamic decisions. The reason for this is, we
That’s what we call as a human error. But these
need data of the government and that too as per
mistakes can be easily tracked down and can be
speci c to our needs and relevance. Similarly we
controlled as soon as it is observed and counter
have to implement systems for getting these data
actions are already ready to be taken. In an
online to our systems which does not seem that
automation system, the process is inlined and
easily possible because this is a matter of
hence any wrong input is going to be passed over
evolution in the tech and digitalisation for all.
to the next system and so on. Since the decisions
not found to be wrong, These systems are
Still just for a moment, if we assume that we are
designed in such a way that they don’t make any
getting this system inline data ow to our systems,
wrong move and hence for the same reason they
we can not promise a proper minded decisions to
cannot identify one and the system will not even
the company for strategic planning and ho to run
take any counter actions for it. Since management
business. I simple words, we know how to do
part requires to make complex decision which are
certain thing or things, but we don’t know what to
fi
fi
fi
fi
fi
fi
fi
fi

fi
fl
fi
fi

fi

fl

fi

as per the situation and hence it is skeptical to less of tings as an edge over the machines, but
trust on a system every time. In case if any those are the most curtail things.
decision is given, t is done by keeping in mind the
factors and propagandas which are coded in. This But one of the key factor differentiating machine
can be not just troublesome but can be hazardous and the humans is learning. While humans take
and life threatening too in some cases. some time in learning anything, it may take a very
short time to make a machine learn something.
One of the very lacking thing in these systems is For a human, he will have to understand and
emotional awareness and ethics. We’ll discuss practice the thing whereas for a machine, only a
about this further in this research. For now, so far code needs to be developed and that too can
going into our research, we have a better
understanding of how these systems can be
Human Machine
implemented and interlinked through a network
and effective troubleshooting methods can be Learning Curve of Human and Machine
implemented. We also took a quick look of the
potential issues and risks which can occur.
Humans vs Machine
One of the most crucial, vital and obvious
difference between a Human and a Marine or a
robot is emotions. We can see that a lot of people
are afraid of getting these technology to such an learn a lot of new things in a blink of an eye.
extend that they may start taking over the world.
This is not just any plot of a movie. This thing is a But it completely depends on what we are aiming
possibility and giving the rapid pace of the growth to at learning.there are a lot of things that a
and innovation, we can predict it is close. But does machine cannot learn and humans can. This is
that mean we should stop at a certain point? mainly what makes human superior than machines.
De nitely not. Even though is is a controversial statement, sadly it
is true.
Machines were primarily created to reduce the
human workload and allow to do the same thing at Learning Curve of a Human and Machine
a higher capacity and in less time. In the beginning
these were some basic needs which otherwise One main difference in learning curve of a human
would take several months or years to do by and a machine is that getting the practice is very
humans, things such as making of cloth or making complex and adds up with time and experience.
a plate. These are very basic and general objects We have heard that there is a vast difference in
used In our everyday lives. But with increasing theoretical and practical thing. Similarly, getting
population, high demands, and complexity in the the theory into the brains of human and machinery
nature of objects, these machines increased in isn’t not that hard. ALTHOUGH HUMAN MAY
number. We created machines for almost FALL BACK SINCE THEORY AND PRACTICAL ARE
everything and today we have machines to make BOTH A LEARNING FOR HUMANS WHEREAS
everything, even other machines. FOR ROBOTS IT IS ONLY THE PRACTICAL. We
can come the theory to the machine and it may
Machinery comes with a lot of advantages and hardly take 5 minutes for a machine to install, load
hence was very quickly adopted by the people. It and process the code. This will also allow the
provided precision work, ef ciency, effectiveness, machine to perform the functions in the practical
cost saving, longevity, expansion option and a lot form if the code is speci ed accordingly. But when
others. T the same time Humans are still better in it comes to human, the additional knowledge of
a few things like, emotional awareness and correct decision making and learning more while
understanding. Although humans seems to have doing is a big advantage over the machine. A

fi
fi

fi
fi

machine is not learning but only collecting data of Human Intelligence is a quality of the mind that
what it did when. A human on the other hand enables one to use knowledge — acquired from
constantly learning and is gaining more experience existence, abstract concepts, several cognitive
and getting sharp in the task or the job. processes, and more, to manipulate one’s
environment. A machine that can, in some way
Although we can create a code to update the ‘mimic’ this quality can make our lives much more
original code with all the actively collected data, simpler and ef cient. This is where arti cial
which is used by a lot of companies these days to intelligence comes into the picture.
keep their AL systems Up to Date. But still the
human learning curve gains at a compound pace Intelligence is a quality that is unique to humans.
whereas the machine learning curve gains at a Even the most complex behaviour exhibited by an
static pace. This is because the more the human insect does not qualify to be called intelligence.
learns and going the experience, it uses all the Take the case with digger wasps as an example:
collected knowledge to learn the next thing or to the wasp that goes out in search of food does not
get batter at it. This makes the learning curve re-enter its burrow without looking for intruders.
bend in positive rate at a compound rates. This is This is based on the fact that a possible threat
not the same case with the machines as the could have intruded the burrow while it was gone
software code which is implemented is limited to looking for food. But when, as part of an
what has been programmed. The learning speed is experiment, the food was left right in front of the
also programmed and hence is static. This is the burrow itself, the wasp was observed to look for
reason why a machine is fast at the beginning but threat before entering the burrow, exactly like how
the human takes over exponentially faster later on. it would have responded in the former situation.
As shown in the above graph. We can see that the Intelligence involves the ability to learn from
Human is learning and the line is raising at a experiences and adapt to changes — something
constant rate in a compound manner, whereas the which is absent here.
machine is learning at a very constant way but is
almost static and at the 7th point of the Machine Developing a machine having such a complex
curve indicates that a new set of code is installed character is not easy. It is a realm that progresses
enabling the machine to do more things. Hence slowly with the hard work of thousands of
the jump from the point 7 to 8 is high in less of a dedicated computer scientists and programmers,
time. Obviously the X Axis in the graph indicates and it takes decades to reach its culminate.
the time.
Artificial General Intelligence
We cannot create or implement a code which
learns at a pace and in the same manner as a Instead of viewing intelligence as one single entity,
human because ltering the right things to take AI has its focus put on the learning, reasoning,
and not to take is not something that a machine language, perception, and problem-solving
algorithms can de ne. This again takes us to the aspects of intelligence. The combination of
same thing i.e. the lack of emotional awareness mathematics, statistics, and algorithms are making
and ethics in machines. us capable of such a feat. Human intelligence has
overcome physical limitations we had in the past
and has transcended to a different world where
Human vs AI. such limitations do not matter. AI aims for similar
outcomes. We have now reached such levels
where we could think and work toward making
At this point, we might have got this thing in strong Arti cial Intelligence a reality. Strong AI, or
ourselves that the only thing that makes machine Arti cial General Intelligence, is the form of AI
lacking far behind from taking them over is the where a machine’s cognitive capacities are on par
intelligence. This is the primary reason for the with a human being. This machine possesses
evolution of AI or Arti cial Intelligence. consciousness, self-awareness and can
communicate to us in our languages.

fi
fi

fi
fi
fi
fi

fi

fi
In some of the developed countries, AI-enabled

The possibility of such awareness is highly robots perform primary health check-ups after
disputed. One of the most signi cant philosophical engaging in a dialogue with the patients. AI-
question is whether Arti cial General Intelligence enabled robots which can detect the progress of
can be developed to respond like a human brain? tumours, create medicinal drugs, etc. nd their
Some Scientists believes that Arti cial General space in the medical eld. There are robots that
Intelligence can never become a reality, but some can help take care of the elderly and design
others believe that we can develop it by the next treatment methods.
century, or perhaps even before.
These are some of the functions that a traditional
Artificial Intelligence (AI) and Human machine can not do. What is unique to AI-enabled
Rights machines is that they make use of machine
learning algorithms which lets them access
The development of arti cial intelligence has information, interpret it and offer a valid, genuine,
raised several serious questions. Some of them are and well de ned nal result.
regarding the impact AI has on human rights.
International Conferences and high-level A Vision for Humanity and the challenges
committees at different parts of the world are
ensuring that the development of AI will only AI-enabled products offer logical solutions to
bene t humankind and will not compromise on users. The algorithm that runs the product ensures
human rights. Also, companies like Google and this. Once they evolve to understand and interpret
Microsoft have already published ethical principles human text and speech, the quality of the machine
regarding AI. and its result will enhance. But such a process is
not easy as computing models which can integrate
Since the development of Arti cial General natural and visual processing are necessary for it to
Intelligence is not going to happen any time soon, be a reality. This is a real hurdle that stands in the
there is no urgency for governments and path of developing humanoids as understanding
concerned authority to address the issue and nd and interpreting human language is a complex
a solution immediately. Also, with this level of phenomenon and is very dif cult for AI-powered
development, it is not possible to predict what machines as our words change meaning with
impact AI will have on ethics and human rights. context.
Hopefully, with ef cient frameworks on AI and
ethical guidelines, the advancements in the eld of Another hurdle that stands in the path of such a
arti cial intelligence can be directed only in the marvel is the fact that even though medical
direction of progress of humanity. It will begin a science has progressed so much, we still do not
new era with enhanced living standards. have a concrete idea about how the human mind
or human intelligence functions. We do not know
How Far Have We Come? the various elements that constitute our
intelligence or how these constituents are
As time progressed and as technology developed, interdependent. We rst need to have a better
arti cial intelligence also transcended its vision. In understanding of how our mind and intelligence
the past, computer scientists wanted to create work in order to build machines that in some way
machines equipped with AI to perform mimic these and thereby interact with us as one
independent tasks for us. Their success has made among us.
scientists think of developing devices that are
capable of performing tasks with us. The modern Computational neuroscience is a eld of study
world has seen a lot of progress since this which makes use of Mathematics to learn the
transformation — from performing tasks for us to human brain by creating theoretical models of our
perform tasks with us. brain. Studies in such areas have revealed to us
details about our nervous system, which was
previously unknown. The new information available

fi
fi
fi
fi

fi

fi
fi

fi
fi
fi
fi

fi

fi

fi
fi
fi

fi
fi

fi

has improved our understanding of the processing will be far more advanced and superior than the
capabilities of our nervous system and thus, have best of what we can imagine.
given us a chance to improve on arti cial
intelligence with the help of this information. Over 75% of customers availing in one way or
Drawing inspiration from the human nervous another, the services of an AI-powered system.
system, Arti cial Neural Network, ANN, was born Also, over 80% of global business organisations
to mimic the problem-solving capabilities of the believe in the ability of arti cial intelligence to
human brain. offer them a competitive advantage. It won’t take
much time for the science ction of yesterdays to
As opposed to functioning with programmed and become an absolute reality. The global AI market
task-oriented rules or algorithms, ANN systems will grow to $60 billion, which is over 25 times its
operate by learning how to perform actions by number in 2016.
referring to examples and by learning from
previous experiences (do not infer ‘learn’ in the Acceptance Among People
literal sense). Consider the example below:
An ANN System is trained to identify a particular But keeping all the business facts and statistics
image, say cars, by exposing it to images that are aside, AI is making signi cant progress in the
previously labelled manually as ‘cars’ and ‘not direction where its interactions with its human
cars’. Now it is programmed to choose only the users are also improving. With developments in
image that contains cars and leaves the rest. This the eld, especially in the eld of speech synthesis
result is saved for later references. Here the and speech analysis, the gap between the user
program did not feed the system with the details and his/her virtual assistant has narrowed. People
of the car in order to make it identify the image of have started relying on their chatbots. They have
a car when presented to it. The system can also been trusting the medical diagnosis given by
automatically nd similar characteristics from the the robots. We call a customer care facility with
car and can recognise them as pertaining to cars. the expectation that a robot will respond to us; we
Thus the system learns how to identify cars from a log into websites to get assistance from a chatbot.
given set of images; if any. What the human mind has achieved will one day
welcome the world of robots too. There will come
This technology is used in image recognition. a day where human intelligence and AI will work
Facebook’s image recognition system, DeepFace, together for one shared goal — to make our lives
was developed with the help of this technology. better.
DeepFace was trained to recognise human faces
with the help of millions of images that were WILL ARTIFICIAL INTELLIGENCE GROW
uploaded.
BEYOND HUMAN INTELLIGENCE?
The Future
We have understood that human intelligence is the
quality of brain that learns, extracts knowledge,
The limitations that strong AI has today that stops acquires abstract concepts from its surrounding,
it from attaining its full potential cannot exist for a whereas arti cial intelligence is the ability of a
long time. Exactly how technology brought our machine to mimic the same tasks learning from
world to such heights, it will certain transcend data it receives. Intelligence is a quality that
further. With global giants like Google and IBM belongs to humans and if machines could play the
working on AI, the day cannot be quite far. game right, our lives would become much easier.
The future generation will live in a world where “The rise of arti cial intelligence is raising the
humans and humanoids coexist — with humanoids premium on tasks that only humans can do: it is
existing to serve human beings. Those will be the freeing workers from drudgery and allowing them
days where human to machine interactions and to spend time on more strategic and valuable
machine to machine interactions are the same as business activities. Instead of forcing people to
human to human interactions. Life in such an era spend time and effort on tasks that we nd hard

fi
fi

fi
fi

fi

fi

fi
fi
fi

fi

fi

fi

but computers nd easy, we will be rewarded for form concepts, engage in problem solving, reason
doing what humans do best — and arti cial and make decisions. This process is called
intelligence will help make us all more human.” reasoning. Logic is the science of reasoning.”
However, despite signi cant advancements, AI still
could not match up to human intelligence in most Ai is what a human has created and it can
aspects. In the growing debate about AI vs. human de nitely create a certain boundaries which can
intelligence, the given wisdom has been that limit the thinking and learning capabilities of AI
arti cial intelligence will augment human tasks, but and a machine. Consider an example where Vasani
not replace them, anytime soon. Andrew McAfee, Industries have created a machine with AI and
a professor at Massachusetts Institute of programmed it in such a way that it can keep itself
Technology, noted that 20 years have passed since up to date with the management tactics and can
a computer beat world chess champion Garry make decisions in favour to the company and does
Kasparov yet the gap between computer ability not affect any other entities in a negative manner.
and human ability has only gotten more signi cant. To avoid the overcoming to the thinking the
He said, “We still underestimate how big, how programmers implemented a 2 stage security
fast, technological progress is. I still keep getting it rewall code. One code will crash the whole
wrong.” system if the AI comes to think in a way not
de ned by the programmers. And second code
Analysing the unbridgeable gap between human that crashed when the Ai comes to think of making
intelligence and arti cial intelligence in the near any changes in its own program to break the
future, the right perspective would be to the security and expand the boundaries set up by the
complementary attributes of AI with human programmers. The reason why we have made 2
intelligence. The scienti c researches should be security checkpoints because the rst one is a goal
focused on developing arti cial intelligence which Ai will be achieving in case if goes on the
applications that could integrate with human wrong track. If there is no such code, then the AI
intelligence in an effort to enhance productivity does not have anything to look for as may nd
within the broad restrictions of privacy and some other way as well. That is why the second
sensibilities. If we will create a collaborative world security checkpoint is implemented as it will crash
for survival for both, then it would surely open up the system then and there itself. This will ensure
new opportunities for many in numerous different the system is not going further fro the de ned
elds. checkpoint and provides an additional layer of
high-end security. Since we are aiming to not let
Thinking is equivalent to AI in Machines the AI go beyond its boundaries, we ourselves are
making it limited as compared to humans.
As per our understandings of AI, we can say that Interestingly we can put restrictions on humans but
thinking is a possibility for machines through AI. AI we cannot make them think a certain way
powers a machine to enable and think at a certain forcefully. Again it is an art to brainwash someone
level. This thinking is solely powered and and change the way someone thinks and
controlled by human at this point, that means any programming is nearly impossible because it is not
machine with an AI program cannot be thinking based on any set principles or formulas. Hence,
on it own like a human. A thought is created in these machines with AI are only going to perform
mind when any speci c situation, observation or certain tasks or set of tasks which are de ned and
and feeling comes along with anything. Humans will learn whatever is programmed in them to
have tendency to react at certain situations, learn.
observations and feelings which triggers a though
process in their mind and that can be a technically Lack of Ethics in AI and ML
de ned as thinking. Dictionary de nition of
thinking is “The process of considering or The primary reason of putting limitation to ensure
reasoning about something.” The technical the machine does not takes any decision or any
de nition on the other hand is as “Thinking action which is not right. This can be for the
involves manipulation of information, as when we company, government, society and everything else

fi
fi
fi
fi
fi
fi
fi
fi

fi
fi
fi
fi
fi
fi

fi
fi
fi
fi

fi
fi
fi

too. As we found previously, the lack of ethics is Can we define, what Good is? How?
the most crucial and can be said as the most
important thing which is missing. The textbook A we saw in our previous example, we saw that it
de nition of Ethics is “Ethics, also called moral is hard to de ne what good is and what is not.
philosophy, the discipline concerned with what is Even if a programmer a team of programmes
morally good and bad and morally right and prepared a code with ethics, it is mostly
wrong. The term is also applied to any system or impossible to make it dynamic in nature. Same
theory of moral values or principles.” thing or principle can be different for different
people ans per their thought process. But there
Can we Program Ethics? must be some prede ned set of guidelines or
norms which keeps good and not good
Ethics bifurcates Right and Wrong for any human. apart.What is it? We see that in almost all the
But it is not same for all. Something can be right religious activities, communities and groups, there
for someone but not for others. A thief can steal is always a set of guidelines for some set principles
money from someones bag and as per him, it can which keeps a set of expectations of what and how
be a right thing to do but for others it is a wrong much one can be good. This is something we can
deed. To make our understandings more complex. use for our code to program? We have observed
Assume A, B, C & D are 4 people. A is an IT that there are a lot of religious books like Bible
professional at Vasani Industries, B lost his job and and Quran which provides such knowledge and
is unable to get a new one. C works for an guidance. Once such book, which is not just called
assassination company whereas D is a cop and as a religious book but also knows “The Manual of
serves for Police forces and government. B didn’t Human” that is “Shrimad Bhagavad Gita”.
got any job and is jobless for a long time. He is Although there are a lot of people who may not in
nancially broke and to feed his wife and children, the absolute favour of this but just for the
he stole some money from a shop and bought experimental purposes, we can take this as an
food of it for himself and his family. Second time opportunity and learn. We can take all the.main
he got caught and punished. A upon hearing religious books but to reduce our workload for
about B thought why hi e did it. Here as per A It programmers we are taking only Gita. Also Gita
was wrong and as per B it was right. Similarly C includes all the content matter that is in the other
killed a person as his task and D encountered C. books like the Bible and Quran. We can de nitely
Here both C and D killed a person and killing is include all the books if we want.
right for them. As per A and B, it is wrong for C
but they think right for D. But A and B themselves All these books try to keep a set of guidelines or
think of killing someone as wrong. de nes set of which which are good for everyone
and these are some moral values and manners
By our example, we got na understanding that which as per these books and the general
same things can be different for different people observation, are the mostly adopted by the
as it depends on various factors and situations. people normally or is recommended. These ethics
Same thing in different situations can be different and manners are something that a parents want to
or same as well but the behaviour may be teach their kids at a very young age do that it is
different. with them throughout the life. These are general
ethical manners like Honesty, Cleanliness,
Similarly, the person who programmes codes for a Generosity, Ascetics, etc. Gita is said to be a book
machine and so called ethics in the machine code to refer for getting answer to any question in the
is not necessary to be correct for everyone. This is world.
why the evolution of humans as they grow in age
always is different as it depends on what path the Let’s say we programmed a code with all the
person has chosen for growth. If the person has Verses and Shlokas of Gita and this code will act as
always been taught the wrong things, those are a 1st even lter where any decision or output
right for him but may not be right foe others. given by the AI embedded machine is passed
through the code and if there is something that

fi
fi
fi
fi

fi
fi
fi

fi

con icts the Ethical norms as per the Gita, it gets see the industries revolving around these digital
an error message. This code can be implemented world and evolution of these from a scratch to
as for dining the thinking process of the AI and any tech giants like Google, Apple, Microsoft,
thinking or decision that is being processed is as Facebook, Amazon, Tesla, and a lot more.
per these Gita Norms.
In this Research Paper we studied, observed and
All the decisions which con icted shall not be kept found out -
on board but can be stored in a secure human
controlled place. This will allow the programmers • What is AI and ML
and the Analysts to understand not only the • How a business runs
thinking patterns of the AI but also will help them
• Management and production part of business
to see the level of wrong decisions the machine is
making and can also help in comparing the quality • Automating management part of business
of thought being prepared within the machine. All • Automating production part of business
the data is always helpful. At a certain point if we • Examples of these partial automation
found that the degree of decision and the thought approaches
process is crossing a certain level, where it will be
• Joining the dots between the management and
harmful and can cause consequences which are
production automation.
not at all welcomed, we can immediately put a
stop tot the machine then and there. This is the • Simulated case study of Vasani Industries
primary reason why an ON-OFF button is required • Vasani’s Law of Process
to be in the hands of the humans. As we become • Risks of Automation
more and more con dent with the decision and
• Human vs AI
the though process of the machines, we can make
them more and moe liberal and allow them to • Learning curve of Human and Machine
make critical decisions as well for the company. • Lack of Ethics in AI an Machines
• Programming Ethics in Machines usine Bhagavad
Hopefully at a timeframe, we’ll have a program Gita
that will be so clean and lean that it can be
installed in any computer or machinery without any
Developing of AI and ML is not just the upcoming
fear of human takeover. This will be the point
future but is an essence of future living. This
where we cans the main goal this research to look
research cane a proof of how this can be
more strongly built and practical in nature. At this
implemented and taken advantage of this
point we can start looking at the advantages of
automation in business. We have seen automation
the AI and ML in the business because we have
changing lives constantly. Places like Hospitals,
tackled al the barriers an ensures we provided a
transportations, Forces, Services, etc are taken to
solution to each critical and small issue before
the next level. Today a major surgery in a very
moving forward. These decisions an be not just
delicate part of our body is completely done
trustworthy but highly ef cient and accurate as
through these AI enabled machines. Only
these are solely based on the data and analytical
difference is that these are controlled by a human.
theories and models. These analytical theories and
With a new evolution and this research, we can
models are used by us humans and now it de nes
de nitely create a dent and make changes to the
how we can get it done more accurate and faster.
digital world.
Conclusion With all the above research, we can con dently

conclude that the study we did can enable and
Automation and digitalisation was always provide us the correct path to getting any business
welcomed and got implemented at a very fast to complete automation. We can de nitely not do
pace in to the business and profession sectors. it as simple it may look on the paper but we
Since these cam with great bene ts at a cheaper de nitely have a blur blueprint for reference ad
costs, it was adapted at a very fast pace. We can guidance. We can see a much better approach and

fi
fi
fl

fi

fi

fi
fl

fi
fi

fi

fi

innovative ideas with the new generation of man

power and the highly skilled people which such an
integrate level of IT and Management knowledge
and understanding.
End Note
At the end, I would like to thank every one for

taking time and going through al this research
material, The Wharton School of the University of
Pennsylvania, The Johns Hopkins University, and
all the other institutions, Internet, and everyone
who assisted and guided me directly or indirectly
for any part fo this research paper.
I apologise for any word, statement or gesture

that may have came in the wrong way, offending
anyone’s personal, professional or social
thoughts, feelings or motives. I assure this paper
has no such intentions in any way and is solely
for the research and learning purposes.
References
All the studies and references are solely owned

and created by Abhi Vasani, the author and creator
of this Research Paper. References from Internet
and general books is taken for help in forming the
content in the research.
Created: September, 2019

Accepted: November, 2019
Published: November, 2019
Released and Reserved: January, 2021

fi

Research Paper

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Research Paper

Uploaded by

Copyright:

Available Formats

RESEARCH PAPER

USING ARTIFICIAL INTELLIGENCE (AI) AND MACHINE

Mr. Abhi Vasani.

1. Overview Automating partially

But Dr. Mark Esposito wants to root the

and offer decision support for speci c tasks, but

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 6

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 8

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 10

new occurrence or changes or, will lack at some

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 11

• The requirements differ from whatever the task

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 13

Cons of automation Risks of automation

When we think of cons for automation in such a

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 16

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 17

In some of the developed countries, AI-enabled

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 18

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 19

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 20

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 21

Conclusion With all the above research, we can con dently

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 22

innovative ideas with the new generation of man

At the end, I would like to thank every one for

I apologise for any word, statement or gesture

All the studies and references are solely owned

Created: September, 2019

Copyright © 2019 Abhi Vasani INDIA/USA, All rights reserved. Page 23

You might also like