Introduction to AI/ML

Data Science
Why, What, Where and How?

Disclaimer:
• Only working definitions
• No clear agreeable-to-all definitions and subject
classification
• No one right answer, and there are no 100% wrong
answers
• Everything depends on the context
 Data Science
Are DS, AI & ML same?

Disclaimer:
• Only working definitions
• No clear agreeable-to-all definitions and subject
classification
• No one right answer, and there are no 100% wrong
answers
• Everything depends on the context
What is Data Science?

2022 2023
Understanding ML, AI & DS
“ Data
Data Everywhere

Data
• something raw,
meaningless,
• structured, unstructured
• numbers, text, symbols,
pixels, voxels, audio, video,
etc
Data Everywhere

Data
• something raw,
meaningless,
• structured, unstructured
• numbers, text, symbols,
pixels, voxels, audio, video,
etc
Data Everywhere

Data
• something raw,
meaningless,
• structured, unstructured
• numbers, text, symbols,
pixels, voxels, audio, video,
etc
Data Everywhere
Data Everywhere

Internet connections in
India (in millions)

Data to grow …
Data Science

Data
• becomes information when converted into
a useful form (pre-processing, cleaning,
transformation, reduction, discretization,
storage: csv, tsv, XML, RSS, JSON),
• Eg.: 1964, 2056, 5898, MFF, 36, 40, 23,
8575, 10500, 2520

Cust. Id Gender age Monthly Income ($)

1964 M 36 8575
2056 F 40 10500
5898 F 23 2520
Data Everywhere

Analysis Analytics
• analysis refers to hands-on data • explores potential future events, predictions
exploration and evaluation
• qualitative analytics = analysis + intuition
• looks backward, providing a historical view
• quantitative analytics = models build from
• Explains how? why? in the past analysis + algorithms

past present future

Analysis Analytics
Data Analysis Data Analytics
Business Analysis Business Analytics
“ Machine Learning
Machine Learning

• The ability of machine to make predictions without being explicitly programmed

• Creating and implementing algorithms to receive data and process it to
• make predictions
• analyse patterns
• provide recommendations

Y (=mx+c)
Simple Linear
Regression c m*x

x
Machine Learning

• The ability of machine to make predictions without being explicitly programmed

• Creating and implementing algorithms to receive data and process it to
• make predictions
• analyse patterns
• provide recommendations

(℉ − 32)5/9
℉ ℃
How to find this relation?
=> ML Training
Machine Learning Machine Learning
“ Artificial Intelligence
 A Brief History of AI
More mathematical models,
Claim: In 20yrs, AI can do Revived by expert
verifiable results
whatever a human can do systems

1956 1960 1970 1980 1990 2000 21stC

LISP system
Field of AI AI-winter DS revived AI:
collapse. Longer
was born Advances in ML,
AI-winter
supercomputers, improved
algorithmic, data access,
open-source tools, cloud
computing etc
A Brief History of AI
From Logic to Learning

• The evolution of AI: From rule-based systems to learning algorithms.

• How AI has transformed industries and daily life.
• The rise of Machine Learning: AI that improves with data.
Artificial Intelligence

• Simulating human knowledge and decision-making with computers, e.g, self-driving cars.

• AI but not ML?

• E.g., line follower robot, symbolic reasoning: high-level human-readable representation of problems and
logic (rarely used?)
• Today, ML is the only form of general AI that is being applied.

Simple AI:
Data Science Machine Learning Artificial Intelligence

Data Science
Data
Stories Problem
Formulation

IoT

Artificial
Data
Intelligence Engineering
Machine DL Big Data
Learning Data
Visualization
MLOps
Federated Edge
Learning Computing
Career Paths

AI, ML and Data Science

 AI Systems

Subject Matter Experts

AI Systems

Data Engineers
Career Paths
Data Engineer

Data Capture Software Big Data Skills

solutions
advanced creating software MySQL, MongoDB, Scala,
transforms data programming and solutions around Dask, Apache Spark, Hadoop
into a useful format system creation big data
for analysis Hadoop, MapReduce, Hive,
creating software create data Pig, Data streaming, NoSQL,
Collection, cleaning solutions around pipelines SQL,
big data understand the
right tool
Career Paths
Data Engineer
AI Systems

Data Scientists
Career Paths
Data Scientist
Data analyst who applies ML/AI
ML Expert picked up programming out of
necessity

Interact with business side

Domain Expert understand the domain enough to
make insights

Verbally and visually communicate

complex results in a way that the
Data Story business can understand and act on
them

Math, statistics and basic

programming to analyse data
Skills R, create ML models
Jupyter, TensorFlow, PyTorch,
RStudio. etc
Career Paths
Data Scientist
AI Systems

ML Engineers
 ML
Engg. ML Engineer
ML/AI
• Cloud Native ML Platform : AWS SageMaker, Azure ML
Studio, GCP AI Platform

Career
• Containerized Workflow: Docker, Kubernetes, public
and private containers

Paths
• Serverless technology: AWS Lambda, AWS Athena,
Google Cloud Foundation, Azure Functions
• Special hardware for ML: GPUs, TPUs, A14, AWS
Inferentia Elastic inference
• Big data platforms: Databriks, Hadoop/Spark,
snowflake, Amazon EMR Google Big Query, Google big
Query
Career Paths
ML Engineer
Computational Science
Vs.
Data Science

I am not a Software Engineer;

Can I become an ML Engineer?

Computational Science

Deploy
• Static package
Simulation • Cloud service
• Find solution (SaaS)
Modelling • Evaluation
• Define rules • Verify and
• Describe the Validate
dynamics
Application
• Define
Problem
Computational Science

Dynamics
process

Validation Assumptions,
Verification Simplifications

Computational
Science

Extreme-scale Modelling
computing

Numerical
Schemes
AI & Data Science ML Engineer

Experiments
Data collection

AI, Data
Visualization, cleaning,
Data story Data
Data munging
Science

AI
Extreme- Analytics,
Scale Machine
computing Learning
Generative AI

Beyond Prediction to Creation

Generative AI
• Generative AI is not about predicting the future but creating something new.
• It learns from data to generate original content - images, music, text, and more.
• The magic of Generative AI lies in its ability to surprise us with its creations.
Why Generative AI Matters
• A tool for innovation in art, music science, and business.
• Breaks creative boundaries, enabling new forms of expression.
• Powers personalized and dynamic content, transforming engagement.
What is Generative AI?
• Generative AI refers to algorithms that can create new data resembling the data
they were trained on.
• Unlike traditional AI that interprets or classifies data, Generative AI produces new,
original outputs.
• It can generate text, images, music, and more, bridging the gap between machine
understanding and human-like creativity.
Generative AI vs. Traditional AI
• Traditional AI: Focuses on understanding, analyzing, and responding to input data
(e.g., spam filters, recommendation systems).

• Generative AI: Aims to create new content that is indistinguishable from its
training data (e.g., creating new artworks, composing music).
• This shift from interpretation to creation marks a significant leap in AI
capabilities.
How Does Generative AI Work?
• Imagine teaching a child to draw animals.
• You show them many pictures, and over time, they start drawing animals on their own.
• Generative AI operates similarly: It 'learns' from vast amounts of data, understanding
patterns, styles, and structures.
• Once trained, it can generate new data—like drawing a new animal no one has seen
before
Generative AI
Generative AI

Image of a mythical creature with the

body of a white elephant and the head
of a lion playing joyously in the ocean
Welcome to AI/ML!