Professional Documents
Culture Documents
Bits - Bytes - Data Digest - January - Editio - 2023 Edition
Bits - Bytes - Data Digest - January - Editio - 2023 Edition
manish.kakumani22@gmail.com
WIREJ2S1OC
JANUARY
2023 EDITION
This file is meant for personal use by manish.kakumani22@gmail.com only.
Sharing or publishing the contents in part or full is liable for legal action.
January 2023 Edition
WHAT’S INSIDE?
Leadership Speaks 03
Discover 07
What’s New? 12
Industry Trends
manish.kakumani22@gmail.com
13
WIREJ2S1OC
AI at Work 15
Mentor Speaks 16
Crossword Solution 18
LEADERSHIP SPEAKS
your team (juniors). Loyalty to the organisation
does not simply imply a long tenure; it also entails
making honest and continuous efforts to achieve
the organization’s goals. On the other hand, you
must be loyal to your junior team members who
look to you for guidance and inspiration. Leaders
and organisations can demonstrate loyalty to their
teams by not overburdening them, maintaining
respectful interactions, identifying developmental
areas, and providing the necessary support.
RISHABH GUPTA
Associate Director, Sales, Great Learning c) Meritocracy – Meritocracy is not diametrically
1. What are your core values and how do you opposed to loyalty. A meritocratic workplace
ensure that the organization and its activities are promotes high performers rather than rewarding
aligned with them? someone based on tenure, relationship, or any
other factor. These are also the places where high
For me, the values of any individual/ performers feel valued, are not dragged down by
organization should be easily understandable bean counters, and can give their all.
and implementable rather than lofty concepts.
I conduct my business keeping in mind the 2. Talk about a leader that inspires you and why?
manish.kakumani22@gmail.com
following:
WIREJ2S1OC
I admire all entrepreneurs, but especially those who
a) Integrity – Being true to your mission and build long-term businesses rather than those that
maintaining Integrity in your communication and rely on borrowed funds. Mr Sridhar Vembu was
delivery will provide a long-term advantage in born in a small village in India but went on to build
a world where sales and marketing are taking a hugely profitable business (Zoho Corporation)
over and enabling the distribution of low-quality with no outside funding. Despite being a billionaire,
products and services. False promises are strictly he has remained true to his roots and is extremely
prohibited. humble in his demeanour. He operates in rural
Tamil Nadu and has established schools and
b) Loyalty – loyalty is a two-way street, and you training centres for rural people to learn software
should be loyal to both your organisation and development, making a significant impact on the
ground.
DISCOVER
manish.kakumani22@gmail.com
WIREJ2S1OC
https://www.mygreatlearning.com/blog/expert-
https://www.mygreatlearning.com/blog/difference- systems-in-artificial-intelligence/
data-science-machine-learning-ai/
A Complete understanding of
LASSO Regression
In this blog, we will see the techniques used to
overcome overfitting for a lasso regression model.
Regularization is one of the methods widely used
to make your model more generalized.
Lasso regression is a regularization technique. It is
used over regression methods for a more accurate
prediction. This model uses shrinkage. Shrinkage
is where data values are shrunk towards a central
point as the mean. The lasso procedure encourages
simple, sparse models (i.e. models with fewer
parameters). This particular type of regression
is well-suited for models showing high levels of
multicollinearity or when you want to automate
certain parts of model selection, like variable
selection/parameter elimination.
https://www.mygreatlearning.com/blog/
understanding-of-lasso-regression/
manish.kakumani22@gmail.com
WIREJ2S1OC
manish.kakumani22@gmail.com
6. max_delta_step: Acceptable range is 0 to ∞.
WIREJ2S1OC
Ideal range is 1 to 10.
WHAT’S NEW
What is ChatGPT and generative AI? NXP Protects Machine Learning IP with
eIQ® Model Watermarking
ChatGPT is a free chatbot that can generate
responses to almost any question. It was developed NXP® Semiconductors has added the eIQ Model
by OpenAI and will be available for public testing Watermarking tool to its eIQ Toolkit for machine
in November 2022. It is already widely regarded learning development. eIQ Model Watermarking
as the best AI chatbot ever created. According is the market’s first practical tool for protecting
to ecstatic fans who posted examples online, the developers’ machine learning investments.
chatbot has been known to generate computer Developers can use the tool to demonstrate that
code, college-level essays, poems, and even half- a machine learning model is a replica or clone of
decent jokes. Others, from tenured professors to their intellectual property without having access to
advertising copywriters, among the diverse range the model’s source code, and the model is granted
of professionals who make a living by creating copyright ownership.
content, are trembling. Why is it significant? The adage “data is the
new gold” has never been more true than in
Despite the reservations that many people have the field of machine learning, where developing
expressed about ChatGPT, machine learning has highly effective models is critically dependent
undeniably positive potential (and AI and machine on domain expertise and good training data.
learning more generally). Since its widespread Despite the fact that machine learning models
adoption, machine learning has had an impact are a significant and differentiating asset to a
manish.kakumani22@gmail.com
on a variety of industries, enabling tasks such as
WIREJ2S1OC firm, they typically lack the copyright protection
high-resolution weather forecasting and medical that prevents unauthorised copying or cloning
imaging analysis. It is obvious that generative AI of regular software. Developers can use eIQ
tools like ChatGPT and DALL-E (an AI-generated Model Watermarking to protect their proprietary
art tool) have the potential to change the way intellectual property (IP) and copyright their
many professions are carried out. However, the full machine learning models while also detecting
scope of that impact and its consequences remain illegal use.
unknown.
INDUSTRY TRENDS
Anticipating the potential of such Large Language
In the last edition, we read about the
Models few pertinent questions come into mind:
importance of Large Language Models. Let us
read about the challenges faced in developing
1. How much is it going to impact human lives
Large Language Models
due to its enormous and unknown possible
The major challenges can be divided into three uses (or misuses)?
parts:
2. Impact on the labour market (what should be
1. Proper understanding of the limitations of the automated vs what should not be?)
model developed:
The presence of the statistical relationships in 3. Misinformation or Disinformation can be a
the dataset used for training the models can be real concern. Incorrect narratives may get
biased as they itself may have discriminatory generated which are much cheaper and easily
texts, historical bias etc. This will lead to incorrect be used for false propaganda compared to
underlying characteristics and formulate proper restricting or for that matter making correct
methods to curate the datasets used. Full-fledged use of such enormously potential models?
2. Model fine-tuning:
A. Choice of the right dataset (domain-specific) for
fine-tuning
B. Sufficient amount of data to fine-tune the model
C. Formulating fine-tuning guidelines to get the
best performance of the model
AI AT WORK
Even though we started this to try out the
concepts of AIML into our application just for
curiosity. However, after implementing this solution
to one of our modules, we were blown away by the
model’s efficiency and accuracy. We reduced the
use of the Shipping Rates API in the Awards feed
module. Shipping Rates API is now only used for
Real-Time Checkout processes.
Having said that, what began as a curiosity has led
to the implementation of AIML in our application;
we have listed a few modules in our application
where AIML can assist in increasing performance
VISHNU KP
and producing better results.
PGPAIML ALUMNUS
MENTOR SPEAKS
Q4. How did you get your first job and describe In terms of ML algorithms, If you are learning
your journey (difficulties that you faced and how one algorithm learn it completely. I have used
did you overcome) Linear regression to find if a laptop is costly or
not, if a product is a seasonal product or not
My first ever DS role was within the organization, and forecasted stock market price. While Linear
where I found some key areas where DS can help regression is dedicated for regression problems, it
and came out with proper solution architecture to can also be used for classification and forecasting
achieve the same. I would always encourage DS problems. Rather than learning n number of
aspirants to at least consider this route. Domain algorithms, what matters is how well we learn the
knowledge (Subject Matter Expert) is one of the algorithm.
most important skill sets for a Data Scientist and
if someone has worked in an industry then they
must possess this already. Why this is important?
While Data scientist is the hottest job, it is also the
role where many attritions happen. A very recent
survey found 80%+ DS projects are failing and
this is mainly because these models are not able
to make any business impact nor are explainable.
For eg: We may build a highly accurate model that
would predict the customers who would churn
out of our company, but what good in having this
model if it cannot say why these customers would
churn out and what measures the company has to
manish.kakumani22@gmail.com
take to prevent the same? Data Scientists without
WIREJ2S1OC
domain knowledge can never answer this question.
CROSSWORD SOLUTION
manish.kakumani22@gmail.com
WIREJ2S1OC
ACROSS DOWN
1. The _______ keyword designates a function with 1. In Python, ________ can be string, numeric, or
no name or with several statements that return its Boolean- LITERALS
results- LAMBDA
3. A _______ serves as a blueprint or template
2. With the use of______, programmers can divide from which objects can be built- CLASS
or decompose a problem into smaller parts, each
of which can carry out a specific task- FUNCTIONS 4. A _______ is a sequence of characters and is
written within single or double quotes- STRING
6. PEP 8 is a manual for python code that contains
______ outlining how to write better-looking 5. A _______ is an ordered immutable collection of
python code- RULES data- TUPLE
7. An ______ is a container that can hold multiple 9. A ________ is a selection of observations from a
items simultaneously in Python- ARRAY population- SAMPLE