Transcript Slide Deck

ARTIFICIAL INTELLIGENCE
JOURNEY TO A WORLD WHERE COMPUTERS DREAM AND DESIRE
HOW DO WE NORMALLY CODE?
•
Addition of two numbers
print("Enter two Numbers\n")
a = int(raw_input('A='))
b = int(raw_input('B='))
c=a+b
print ('C= %s' %c)
•
Date (The four pages program in OOPs lab)
#include<iostream.h>
#include<stdlib.h>
int a[13]={0,31,28,31,30,31,30,31,31,30,31,30,31};
int h[13]={0,31,29,31,30,31,30,31,31,30,31,30,31};
int flag1=0;
class date
{
int flag, day, month, year;
public:
date(int d, int m, int y)
{
day=d;
month=m;
year=y;
if(year%400==0||(year%4==0&&year%100!=0))
flag=1;
else
flag=0;
}
}
HOW DO WE NORMALLY CODE? (Contd..)
• We understand the given problem.(Sometimes at least)
• We code the program with a pre-existing algorithm or CCP :P
• We test it with some test cases that you can think about.
• And then shout! “Ma’am I got the output ma’am!”
My question is.. Is your program really
intelligent?
NO! It is an explicitly written program which has no intelligence involved.
WHEN IS YOUR PROGRAM INTELLIGENT?
• A program is said to be intelligent when the program can learn by itself to
improve its decision making skill.
• “A computer program is said to learn from experience E with respect to some
task T and some performance measure P, if its performance on T, as measured
by P, improves with experience E.”
Tom Mitchell
And that’s what Artificial Intelligence is all about.
What Is AI?
It is the science and engineering of making intelligent
machines, especially intelligent computer programs.
It is related to the similar task of using computers to
understand human intelligence.
MACHINE LEARNING
THE MACHINE LEARNING APPROACH
• Instead of writing a program by hand for each specific task, we collect lots of
•
•
examples that specify the correct output for a given input.
A machine learning algorithm then takes these examples and produces a
program that does the job.
Massive amounts of computation are now cheaper than paying someone to
write a task-specific program.
THREE TYPES OF LEARNING
• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning
Supervised Learning
• The computer is presented with example inputs and its target(output), given by
a "teacher", and the goal is to learn a general rule that maps inputs to
outputs.
• Categories:
• Regression
• Classification
Supervised Learning: Example
•
Suppose you have been given a basket filled with some fresh fruits and your task is
to arrange the same type fruits at one place.
•
•
The fruits are apple, banana, cherry, grape.
•
•
Here your previous work is called as training data in data mining.
•
•
•
This type of learning is called as Supervised Learning.
So you already know from your knowledge that, the shape of each and every fruit.
So it is an easy to arrange the same type of fruits at one place.
So you already learn the things from your training data, This is because of you have
a response variable which says you that if some fruit have so and so features it is
grape, like that for each and every fruit.
This type solving problem come under Classification.
So you already learnt the things so you can do you job confidently.
Supervised Learning: Real World Example
• Prediction of stock prices
• Predicting weather
• All examples with historical data
Unsupervised Learning
• The computer is presented with example inputs and it has “no teacher” and no
target, the goal is to learn a general rule that infers output from the input.
• Category
• Clustering
Unsupervised Learning: Example
• Suppose you have been given a basket filled with some fresh fruits and your
task is to arrange the same type fruits at one place.
• The fruits are apple, banana, cherry, grape.
• This time you don't know any thing about that fruits, you are seeing them for
the first time. So how will you arrange the same type of fruits?
• What you will do is select any physical character like color, size, shape and
try to arrange them according to the physical appearance.
• This type of learning is called as Unsupervised Learning.
Examples Of Unsupervised Learning
•
Clustering similar News articles under one relevant headline.(https://news.google.co.in)
Other Examples
• Social network analysis
• Image Processing
• Speech Recognition
Reinforcement Learning
• The computer is presented with a dynamic environment and the goal is to
perform a task with the continuously changing environment without a teacher
explicitly telling it whether it has come close to its goal or not.
Reinforcement Learning: Real World Example
• Autonomous cars(self-driving car)
• Google Self-driving cars
• Tesla Model S – Autopilot mode
Reinforcement Learning: Real World Example
• Gaming bots
• Deep blue : Chess playing computer developed by IBM. It the masterpiece of a truly
intelligent player. It won against Garry Kasparov(Chess grandmaster) in a match
becoming the first computer system to defeat a reigning world champion in a match
under standard chess tournament time controls.
APPLICATIONS OF ARTIFICIAL INTELLIGENCE
AND MACHINE LEARNING
Adaptive Websites
• Product recommendations
Computer Vision
• Understanding images
• Face recognition
Oh Mark!! :O
Natural Language Processing(NLP)
• Making computer systems understand natural human language.
• Extensively uses Formal Language(Automata) Theory
• Ex: Google Now by Google, Siri by Apple Inc, Cortana by Microsoft.
Poor chap, got rejected by a computer. :D
Well, can you believe that it is a machine that is talking to its user?
Speech Recognition
• Making computer systems understand the human voice and converting it into
text.
• Ex: Google Voice
Search Engines
•
Using Intelligent search algorithms to bring the best and relevant results to the users.
•
Ex: Google Search Engine, Bing etc.
•
Includes semantic search
And a lot more use cases
• Internet fraud detection
• Spam detection in E-mails
• Robotics
• Online Advertisements
• Computational Linguistics(Sentimental Analysis)
• Software Engineering too :P
IMPLEMENTATION
HOW IS IT IMPLEMENTED?
VARIOUS METHODS & TOOLS
• Linear Regression
• Logistic Regression
• Decision trees
• k-NN (nearest neighbors algorithm)
• Naive Bayes classifier
• Neural Networks
• .. And a lot more ..
NEURAL NETWORKS
NEURAL NETWORKS
• Algorithms that try to mimic the brain.
• Was very widely used in 80s and early 90s; popularity diminished in late
90s.
• Recent resurgence: State-of-the-art technique for many applications.
• The “one learning algorithm” hypothesis
• The field dealing with neural networks is called as Deep Learning
REPRESENTATION
Three major parts:
• Synapses(Nerve Ending)
• Axon
• Dendrite
REPRESENTATION
Three major components:
• Nodes
• Layers
• Edge Weights
Three layers:
• One Input layer
• N Hidden layers(processing part of the neural network)
• One Output layer
COMPLEXITY
This neural network can be used to
process a 28x28 pixel image.
Types
• Feedforward neural network (Unidirectional)
• Recurrent neural networks (Bidirectional)
Interesting Notes
• No one really knows how a Neural Network works/learns. No One!
• They can compute anything(at least theoretically)
• But Neural Networks are slow learners
• They can DREAM!
DREAMS
Leonardo DiCaprio

DREAMS: What Do They Signify?
• With simple words you give to an AI program a couple of images and let it
know what those images contain (what objects - dogs, cats, mountains, bicycles,
... ) and give it a random image and ask it what objects it can find in this
image.
• Then the program starts transforming the image till it can find something
similar to what it already knows and thus you see strange artifacts morphing
in the dreamed image (like eyes or human faces morphing in image of a
pyramid).
ADVANCEMENTS AND BREAKTHROUGHS
IN AI AND ML
ADVANCEMENTS
1. Cheap parallel computation.
2. Vast amount of unstructured and unmined data (It’s also cheap).
3. Advanced Learning algorithms.
4. Many Bots developed which defeated humans in a particular task.
5. Cloud computing has given us the power of utilizing high end servers.
6. Tesla successfully launched its Autopilot mode in cars.
CURRENT SCENARIO
1.
AI is changing the face of Indian IT services
• Automating repeated tasks
2. Driverless or autonomous cars by Google, Tesla, Uber.
3. Search by Images
4. Personalized content
5. Self aware Robots
6. Evolution of Emotional Intelligence
7. IoT + AI = Next Big Thing!
Why Learn AI now?
• Machine Learning and AI are being used extensively in the current scenario.
• Even a matrimonial/dating site uses Machine Learning. :/
• Every human likes his computer to be personalized for him.
• You can build products that can change the way humans live.
QUESTIONS?
FURTHER READING
• Online courses
• Coursera, Udacity and other MOOC.
• Stanford University
• Text Books
•
•
• T. Mitchell (1997). Machine Learning, McGraw-Hill Publishers.
• Peter, Norvig. Artificial Intelligence: A Modern Approach, Pearson.
Journals, e.g.
• Machine Learning, Kluwer Academic Publishers.
• Journal of Machine Learning Research, MIT Press.
Conferences, e.g.
• International Conference on Machine Learning (ICML)
• Neural Information Processing Systems (NIPS)
• AI Blogs and Communities
“ ANY A.I. SMART ENOUGH TO PASS A TURING
TEST IS SMART ENOUGH TO KNOW TO FAIL IT.
IAN MCDONALD
Thank you!
”