CS158-1 LECTURE

CHAPTER 1 Can Machines Really Learn and Why Does It

Matter?
INTRODUCTION TO ARTIFICIAL
INTELLIGENCE 1996: IBM’s Deep Blue won against the world’s
best chess player, Garry Kasparov.
2011: IBM’s Watson won against the best
What is AI?
Jeopardy! Players
“The theory and development of computer
2016: Google’s DeepMind won against the
systems able to perform tasks normally requiring
world’s best Go player
human intelligence, such as visual perception,
speech recognition, decision making, and 2017: Libratus, an artificial intelligence program
translation between languages. “ – Google developed by Carnegie Mellon University, won
against the world’s best poker players.
“A branch of computer science dealing with the
simulation of intelligent behavior in computers.” AI Experts
– Merriam-webster
Elon Musk: This CEO and co-founder of Tesla
is one of the most prominent speakers on the topic
of the potential pitfalls of AI. He founded Open
What Makes Artificial Intelligence So
AI Research, a non-profit company committed to
Important Right Now?
discovering and enacting a path to safe artificial
“AI is the new ELECTRICITY” intelligence technologies.

- About 100 years ago, electricity Demis Hassabis: Founder and CEO of
transformed every major industry. AI has DeepMind, a leading artificial intelligence
advanced to the point where it has the research firm purchased by Google in 2014, this
power to transform every major sector in British neuroscientist are one of the tops AI
the coming years. researchers and experts in the world.
Max Tegmark: This Swedish-American
1. Speed of AI Implementation cosmologist and professor at the Massachusetts
Institute of Technology is also the president of the
2. Potential Impacts on Society Future of Life Institute. He has written over 200
3. Prioritizing AI by Every Large Tech Company technical papers on topics ranging from
cosmology to artificial intelligence.
4. Shortage of Knowledgeable Workers:
Kai-Fu Lee: This Taiwanese venture capitalist
5. Competitive Advantages for Companies whom and CEO of Sinovation Ventures developed the
First Apply AI Correctly: world’s first speaker-independent, continuous
speech recognition system as part of his Ph.D.
6. Legal Implications Worldwide:
thesis at Carnegie Mellon University. He is
7. Ethical Development: recognized as one of the leading experts on AI
technologies in China.
8. Communication of Advantages and
Opportunities: Fei-Fei Li, Ph.D.: As the Director of the Stanford
Artificial Intelligence Lab and the Stanford
9. Collaboration Between Private and Public Vision Lab, Fei-Fei Li is one of the leading
Sectors: experts in computer vision. She is also the Chief
CS158-1 LECTURE

Scientist of Artificial Intelligence and Machine COMPUTER VISION – Deals with how
Learning for Google Cloud. computers can be made to identify objects,
scenes, and activities from images, including
Jenn Wortman Vaughan: As a Senior
image processing, recognition, detection,
Researcher for Microsoft Research who
analysis and understanding.
specializes in machine learning and algorithmic
economics, Jenn Wortman Vaughan focuses on NLP (Natural Language Processing) - mainly
the ways that AI can augment human capabilities. involves knowledge acquisition and expression,
natural language understanding, and natural
language generation.
Disciplines that gave ideas, viewpoint,s and
strategies to AI
- Philosophy
- Mathematics
- Economics
- Neuroscience
- Psychology
- Information Technology
- Control theory and cybernetics
- Linguistics

MACHINE LEARNING
- The study and construction of programs
that are not explicitly programmed, but
learn patterns as they are exposed to more
data over time
Two main types of Machine Learning:
SUPERVISED LEARNING:
Dataset: Has a target column
Goal: Make Predictions
Example: Fraud Detection

UNSUPERVISED LEARNING
Dataset: Does not have a target column
Goal: Find Structure in the Data
Example: Customer Segmentation
CS158-1 LECTURE

CHAPTER 2 3D PRINTING
DISRUPTIVE TECHNOLOGY - printing creates three-dimensional
objects based on digital models by
- Is the technology that affects the normal
layering or “printing” successive layers
operation of a market or an industry.
of materials. The technology relies on
- Clayton M. Christensen of Harvard
innovative “inks,” including plastic,
invented the term
metal, glass and wood.
INTERNET OF THINGS
HOW DIFFERENT 3D TECH WORKS
- Is a network of physical objects –
1. Fused Deposition Modeling (FDM) -
devices, vehicles, appliances – embedded
The FDM printing process starts with a
with sensors, software, network
string of solid material called the
connectivity and computing capability
filament which in turn is guided from a
that enable them to collect, exchange and
reel attached to the 3D printer to a heated
act on data, usually without human
nozzle inside of the 3D printer that melts
intervention,
the material.
FOUR MAIN COMPONENTS USED IN 2. Stereolithography and Digital Light
IOT: Processing (SLA & DLP)- These
technologies create 3D printed objects
- Low power embedded systems from a liquid (photopolymer) resin by
- Cloud Computing using a light source to solidify the liquid
- Availability of big data material.
- Networking Connection 3. Selective Laser Sintering (SLS) -
IoT Enablers Selective Laser Sintering (SLS) uses a
laser to melt and solidify layers of
- RFIDs powdered material into finished objects.
- Sensors 4. Material Jetting (PolyJet and MultiJet
- Nanotechnology Modeling) - Material Jetting (Stratasys
- Smart networks PolyJet and 3D Systems MultiJet
Modeling) technologies are similar to
ADVANTAGES
inkjet printing but instead of jetting drops
- Machine to machine interaction of ink onto paper these 3D printers jet
- Good automation and control layers of liquid photopolymer onto a
- More technical information build tray and cure them instantly using
- Possess strong monitoring features UV light
- Saves times 5. Binder Jetting - The binder jetting
- Save money technology is similar to SLS in the way
that the printer uses thin layers of
DISADVANTAGE powdered material to build up an object
- No international compatibility standard but instead of using a laser that sinters the
- Highly complex resulting in failure layer together, these printers use a
- Security breach binding agent extruded from a nozzle to
- Reduction of employment bind the powder together.
- May take control of live overtime
CS158-1 LECTURE

DRONES AUGMENTED REALITY

- Drones or unmanned aerial vehicles - Augmented reality (AR) is a visual or
(UAVs) are flying vehicles that do not audio “overlay” on the physical world
have on board a human pilot. that uses a smart display to provide
contextualized digital information that
BENEFITS:
augments a user’s real-world view.
- Better Precision
BLOCKCHAIN
- Improved productivity
- Reduced Labor cost - A blockchain is a distributed digital
- Scalability and Adaptability database or, more broadly, a digital
- Increased Safety ledger that uses software algorithms to
record and confirm transactions with
LIMITATIONS
reliability and anonymity.
- Bandwidth availability
4 strategies for blockchain success
- Stored energy density
- Coverage potential 1.) Make the business case
- Lack of Maturity 2.) Build an Industry ecosystem
3.) Design Deliberately
4.) Navigate regulatory uncertainty
ROBOTICS
ARTIFICIAL INTELLIGENCE
- Robots are machines that offer enhanced
- Artificial intelligence (AI) is not a
sensing, control and intelligence to
monolithic technology. Software
automate, augment or assist human
algorithms automate complex decision-
activities.
making tasks to mimic human thought
CHARACTERISTICS processes and senses.

1.) Appearance – Robots have a physical NATURAL LANGUAGE PROCESSING

body
- Natural language processing (NLP) is a
2.) Brain – Onboard control unit (another
branch of AI that helps computers
name for robot)
understand, interpret and manipulate
3.) Sensors – gather info from outside then
human language. It deals with how to
send to the brain
program computers to process and
4.) Actuators – helps the robot move
analyze large amounts of natural
5.) Program – instructions for the robot
language data.
6.) Behavior – Decided by the program

VIRTUAL REALITY
“ For organizations to thrive, they must be able to
- A computer-generated simulated
create and implement a business strategy that
experience that creates an artificial
accounts for the major shifts caused by these
threedimensional image or environment.
technology “
Viewers can use special equipment to
interact with the simulation in realistic
ways.
CS158-1 LECTURE

CHAPTER 3 ITERATION KEYWORDS

PYTHON PROGRAMMING BASICS FOR: This keyword is used to control flow and
for looping.
WHILE: Has a similar working like “for”, used
Python - is one of the achievements of free
to control flow and for looping.
software and is purely free software and both its
source code and interpreter comply with the GNU BREAK: is used to control the flow of the loop.
General Public License (GPL) protocol. The statement is used to break out of the loop and
passes the control to the statement following
FOUNDER: Guido Van Rossum, Amsterdam,
immediately after the loop.
1989
CONTINUE: “continue” is also used to control
the flow of code. The keyword skips the current
PYTHON KEYWORDS iteration of the loop but does not end the loop

TRUE: This keyword is used to represent a

boolean true. If a statement is true, “True” is
CONDITIONAL KEY WORDS
printed.
IF: Truth expression forces control to go in the
FALSE: This keyword is used to represent a
“if” statement block.
boolean false. If a statement is false, “False” is
printed. ELSE - Truth expression forces control to go in
the “if” statement block.
NONE: denote to a null value or void.
ELIF: It is short for “else if”
(It’s important to remember, that 0, any empty
container (e.g empty list) does not compute to DEF: used to declare used-defined functions.
None. It is an object of its datatype – NoneType.
It is not possible to create multiple None objects
and can assign them to variables) RETURN KEYWORDS

AND: This a logical operator in python. “and” RETURN: return from the function
Return the first false value. If not found return
YIELD: return from a generator
last. The truth table for “and” is depicted below.
OR: This a logical operator in python. “or”
Return the first True value.if not found return last. CLASS: declare user defined class
The truth table for “or” is depicted below.
WITH: to wrap the execution of a block of code
NOT: This logical operator inverts the truth
value. The truth table for “not” is depicted below. AS: to create alias for the module imported

IN: This keyword is used to check if a container PASS: nothing happens, it is a null statement,
contains a value. This keyword is also used to used to prevent indentations.
loop through the container
IS: This keyword is used to test object identity, LAMBDA: used to make inline returning
i.e to check if both the objects take the same functions with no statements allowed internally
memory location or not.
CS158-1 LECTURE

IMPORT: used to include module to the DIVIDED INTO GROUPS:

program
- Arithmetic operators
FROM: used with import, to import functionality - Assignment operators
from the module imported. - Comparison operators
- Logical operators
- Identity operators
EXCEPTION HANDLING KEYWORDS - Membership operators
- Bitwise operators
TRY: used to catch errors in the code
LISTS – used to store multiple items in a single
EXCEPT: used with “try” variable [ ]
FINALLY: whatever the result of try, “finally” - ORDERED AND CHANGABLE
is always executed.
TUPLE used to store multiple items in a single
RAISE: raise an exception explicitly variable ( )
ASSERT: for debugging purposes - ORDERED AND UNCHANGABLE
DEL: used to delete reference to an object. SETS - used to store multiple items in a single
variable { }

PYTHON COMMENTS - UNORDERED AND UNINDEXED

Single Line Comments- starts with # no spaces DICTIONARIES – used to store data values in
and last till the end of the line. key:value pairs

Multi-line strin as a comment- enclosed with a FUNCTIONS – A block of code which only runs
delimiter (“””) on each end of the comment. when it is called

Input (): takes the input from the user and

converts it into a string
Raw_input() – takes exactly what is typed and
converts it to string

PYTHON VARIABLES
Variables – containers for storing data
PYTHON NUMBERS – INT, FLOAT,
COMPLEX
Strings – surrounded by single of double
quotation marks
Booleans – represent TRUE or FALSE
Operators – used to perform operations on
variable and values
CS158-1 LECTURE

extensively used for unit-testing and self-

verification to detect and diagnose
CHAPTER 4:
different types of errors.
PYTHON LIBRARIES FOR MACHINE - Theano is a very powerful library that has
LEARNING been used in large-scale computationally
intensive scientific projects for a long
time but is approachable and
NumPy straightforward enough to be used by
individuals for their projects.
- used for working with arrays.
- It also has functions for working in TensorFlow
domain of linear algebra, Fourier - TensorFlow is a popular open-source
transform, and matrices. NumPy was library for high-performance numerical
created in 2005 by Travis Oliphant. It is computation developed by the Google
an open-source project and you can use it Brain team in Google.
freely. - Tensorflow is a framework that involves
- NumPy stands for Numerical Python. defining and running computations
SciPy involving tensors.
- It can train and run deep neural networks
- SciPy is a scientific computation library that can be used to develop several AI
that uses NumPy underneath. applications. TensorFlow is widely used
- SciPy stands for Scientific Python. in the field of deep learning research and
- It provides more utility functions for application.
optimization, stats and signal processing.
- Like NumPy, SciPy is open source so we Keras
can use it freely. SciPy was created by - Keras is a very popular Machine
NumPy's creator Travis Olliphant. Learning library for Python. It is a high-
Scikit Learn level neural networks API capable of
running on top of TensorFlow, CNTK, or
- Scikit-learn is one of the most popular Theano.
ML libraries for classical ML algorithms. - It can run seamlessly on both CPU and
It is built on two basic Python libraries, GPU. Keras makes it really for ML
viz., NumPy and SciPy. beginners to build and design a Neural
- Scikit-learn supports most of the Network. One of the best things about
supervised and unsupervised learning Keras is that it allows for easy and fast
algorithms. Scikit-learn can also be used prototyping
for data-mining and data analysis,
making it an excellent tool for starting PyTorch
with ML. - PyTorch is a popular open-source
Theano Machine Learning library for Python
based on Torch, an open-source Machine
- Theano is a popular python library is Learning library implemented in C with
used to define, evaluate and optimize a wrapper in Lua.
mathematical expressions efficiently - It has an extensive choice of tools and
involving multi-dimensional arrays. libraries that support Computer Vision,
- It is achieved by optimizing the Natural Language Processing(NLP), and
utilization of CPU and GPU. It is many more ML programs. It allows
CS158-1 LECTURE

developers to perform computations on

Tensors with GPU acceleration and also
helps in creating computational graphs.
Pandas
- Pandas is a popular Python library for
data analysis. It is not directly related to
Machine Learning as we know that the
dataset must be prepared before training.
- In this case, Pandas comes handy as it
was developed specifically for data
extraction and preparation. It provides
high-level data structures and a wide
variety of tools for data analysis. It
provides many inbuilt methods for
groping, combining, and filtering data.
Matplotlib
- Matplotlib is a very popular Python
library for data visualization. Like
Pandas, it is not directly related to
Machine Learning.
- It mainly comes in handy when a
programmer wants to visualize the
patterns in the data. It is a 2D plotting
library used for creating 2D graphs and
plots.
- A module named pyplot makes it easy for
programmers to plot as it provides
features to control line styles, font
properties, formatting axes, etc. It offers
various kinds of graphs and plots for data
visualization, viz., histogram, error
charts, bar charts, etc.
Seaborn
- Seaborn is a library mainly used for
statistical plotting in Python. It is built on
Matplotlib and provides beautiful default
styles and color palettes to make
statistical plots more attractive