Philosophically Fundamental Questions
Can a machine act
intelligently? Can it
solve any problem that
a person would solve
by thinking?
01
Are human intelligence
and machine
intelligence the same?
Is the human
brain essentially a
computer?
02
Can a machine have
a mind, mental states,
and consciousness in
the same way that a
human being can? Can
it feel how things are?
03
AI is the next digital frontier
In 2016 Companies invested
In artificial Intelligence
Tech Giants Startups
AI Adopters - 20%
in multiple technology areas
AI Partial Adopters - 40%
skeptical about Business Cases and ROI
Laggards - 40%
contemplators
AI Adoption Story
Digitally
Mature
Large
Businesses
AI adoption in
Core Activities
Focus in
Growth over
Savings
C Level
Executives
support AI
EarlyAIAdopters
High Tech
Companies
Automotive &
Assembly
Financial
Services
MediumAIAdopters
Retail
Media &
Entertainment LaggardAIAdopters
Education
Travel &
Tourism
HealthCare
Four Areas where AI creates significant
Value
Smarter R&D
and
Forecasting
1
Optimized
Production and
Maintenance
2
Targeted Sales
and Marketing
3
Enhanced User
Experience
4
AI, Machine Learning, Deep Learning…
ARTIFICIAL
INTELLIGENCE
Approaches
that help enable
machines to mimic
humans
intelligence
MACHINE
LEARNING
Specific Algorithms
that help machines
learns patterns from
data
DEEP
LEARNING
Algorithms based on artificial
neural networks that help machines
to learn patterns from data
The AI Landscape
Machine Learning
(Non Specific)
Computer
Vision
Natural
Languages
Autonomous
Vehicles
Virtual
Agents
Smart
Robotics
What
does AI help
us to do ?
Human Beings can write a lot of algorithms but some
algorithms are intrinsically difficult and hard..
• Write an algorithm that can looking at Medical Scans
identify diseases
• Write an algorithm to drive a Car
• Write an Algorithm that helps to identify the speakers
of different voice samples
• Write an Algorithm that determines the emotions
implicit in a piece of text
• Write an algorithm that recommends new products to
customers depending on what they have been
purchasing for the last few years.
And so on….these are hard problems to write algorithms
for, for even the “smartest Practioners of that Domain”
We want Machines to help us write these difficult programs
Enter Machine Learning!
Output
Input data Programmer writes a program
based on rules
Result is produced by the program
written by programmers
Input
Human Written
Algorithm
A model
Is discovered that allows
For Input Output
Set of training data Inputs and
Outputs
A machine learning algorithm Output is an automatically learned
algorithm which is called as model
Prepared Input
Machine Learning
Algorithm
Expected Output
Traditional Programming Paradigm
Machine Learning Paradigm
Even preparing Inputs is hard …
Enter Deep Learning!!
Model / Program
Is discovered
Set of training data Inputs and
Outputs
A Deep learning algorithm Output is a automatically learned
program which is called as model
Raw Input
Deep
Learning Algorithm
Expected Output
Deep Learning Paradigm
Model / Program
Is discovered
A machine learning algorithm Output is a automatically learned
program which is called as model
Prepared Input
Machine Learning
Algorithm
Expected Output
Machine Learning Paradigm
How does it do that?
Mathematical
Modelling
Probability,
Statistics,
Linear Algebra
And Calculus
Computational
Programming
Computational
Libraries
Visualization
Libraries
Powerful
Platforms
GPU Hardware
Parallel/Multi
Processing
The Essential Idea
Problem
Representation
Model the
problem using
Mathematics
Transform
Representation
Transform the
representation
In mathematical
space
Iterative
Experimentation
Use fast
hardware to
Facilitate this
transformation
A simple example
Problem
Representation
Model a picture as a
vector of pixel values and
outcome as
0 if cat and 1 if dog
Transforming
Representation
Transform the input
representation and
optimally find the
parameters that give you
the best representation
Iterative
Experimentation
Try different approaches
and
Deploy the code/model
that satisfies the
expectations of the Client
Two important Problem Modelling
Problem
Representation Transforming
Representation
Iterative
Experimentation
Supervised
Modelling
Un-Supervised
Modelling
• Predictive Modelling (Classification
and Regression)
• Discover Groups in Data
• Reduce Complexity of Data
• Visualization of Data
• Anomaly Identification
What is happening behind the scenes
Input Data/Output
Labels
ML
Algorithm
Use the ML
Algorithm
Represents the Input and Output in
An appropriate Mathematical Space
Starts making informed guesses
Of the Output given the
Input
Calculate how far is the guess
from the actual outcome in that
Mathematical space
Adjusts its parameters/knobs
To minimize the distances
Between its guesses and
The actual output
What is the AI workflow?
Build
Model
Optimize
Model
Predict &
Deploy
Validation Data
Training Data
Test Data
Acquire Data
Images
Photographs, Scans, Graphic Art, Hand writing, Paintings
Time Series Data
Financial Data, Transactional Data, Sensor Data
Video
Surveillance Videos, Motion Picture Videos, Streaming
Video
Sound/Speech Recordings
Voice Samples, Conversations
Natural Language Text
Reports, SMS, Tweets, Web Content, Newspaper,
Books..
Structured data
Databases, CSV Data, JSON Data, Excel Data
Different forms of Data –Possible AI use cases exist..
‘Data-Specific’ Use Cases
Train a ML model with Disease Scan Images like X-Ray, MRI, ultrasound and make it diagnose diseases
Train a ML model to convert speech to Text
Train a ML model to convert evaluate Stock Movement and predict Stock price into the future
Train a ML model to monitor security surveillance cameras and alarm when there is a security breach
Train a ML model to understand the User preferences when they come shopping.
Train a ML model to understand the Research Reports and send you a summary of the important ones.
Some more ideas in AI
Anomaly Detection/Fraud Detection
Items, events or observations which do not conform to an expected pattern or other items in the dataset
Cluster Analysis
Find Natural groups in the Dataset based on certain attributes
Dimensionality Reduction
Reduce the complexity without compromising on information loss of the data
Visualization
Take Complex Data and make it humanly visual
The Sciences behind AI
Linear Algebra and
Matrix Mechanics
1
Statistical and
Probability Based
Modeling
2
Machine Learning
Algorithms
3
Neural Net
Architectures
4
The Scientific Method of
iterating - Hypothesis,
Experimentation and
Conclusion Building
5