BREAKING THE

MISSION
1 Artificial Intelligence: Rationale and
Intro

2 AI Techniques: Supervised and

Unsupervised

3 Deep Learning
Artificial Intelligence: Rationale and
Intro
What is AI??
 AI Today
AI has potential to transform every dimension of
human life

Education IoT Health & Medicine Social Media

Economy Autonomous ShowBiz NLP Chatbots

Vehicles Comm
 Artificial Intelligence
Machine Learning (ML):
• Focuses on developing algorithms that enable computers to learn from
and make predictions or decisions based on data
• Techniques include supervised learning, unsupervised learning, and
reinforcement learning.
o Recommendation Systems: Netflix, Amazon
o Fraud Detection: identify and flag unusual transactions
o Predictive Maintenance: Manufacturing companies use to predict
equipment to fail
Neural Networks: Inspired by the human brain,
• series of algorithms that attempt to recognize patterns in data. They are
the foundation for many deep learning models, which consist of multiple
layers of neural networks.
o Speech Recognition: Google Voice, Apple's Siri
o Deep Learning for Image Classification: Google Photos to
classify and tag images based on their content.
o Autonomous Driving: Tesla use neural networks to process vast
amounts of data from sensors and cameras to make real-time
driving decisions.
 Artificial Intelligence
Natural Language Processing (NLP):
• NLP deals with the interaction between computers and human language.
• It enables machines to understand, interpret, and generate human language,
facilitating applications like chatbots, language translation, and sentiment
analysis.
o Language Translation: Google Translate
o Sentiment Analysis: analyze customer reviews and social media posts
to gauge public sentiment about their products or services
o Chatbots and Virtual Assistants: chatbots and virtual assistants like
Amazon's Alexa and Facebook's Messenger bots
Computer Vision:
• enables computers to interpret and make decisions based on visual data
• Applications include image and video recognition, object detection, and facial
recognition.
o Facial Recognition: Security systems and social media platforms like
Facebook use CV for facial recognition, identifying and tagging individuals
in photos.
o Medical Imaging: analyze medical images, such as X-rays and MRIs, to
detect abnormalities like tumors or fractures
o Autonomous Drones: Drones equipped with computer vision can
navigate complex environments, avoid obstacles, and perform tasks like
Artificial Intelligence

Figure- 1: Adaptive recommendations through a structured process

Artificial Intelligence

Figure- 2: Adaptive recommendations through a structured process

AI and Business: Consumer Marketing
• Have you ever used any kind of credit/ATM/store card while
shopping?
• if so, you have very likely been “input” to an AI algorithm
• All of this information is recorded digitally
• Companies like Nielsen gather this information weekly and
search for patterns
• General changes in consumer behavior
• Tracking responses to new products
• Identifying customer segments: targeted marketing, e.g., they find out that
consumers with sports
Cars who buy textbooks respond well to offers of new credit cards. (Association
Rules)
• Currently a very hot area in marketing

• How do they do this?

• Algorithms (“data mining”) search data for patterns
• Mathematical theories of learning
• Completely impractical to do manually
 AI : Predicting the Stock Market
Value of ?
the
Stock
?

time in days
• The Prediction Problem
• Given the past, predict the future
• Very difficult problem!
• Use learning algorithms to learn a predictive model from
historical data
• prob(increase at day t+1 | values at day t, t-1,t-2....,t-k)
• Models are routinely used by banks and financial traders to
manage portfolios worth millions of dollars
AI and Internet of Things
• IoT and Semantic Web
• IoT operates with object addressing, tracking, and discovery
• Information representation, storage, and exchange is equally
important
• Lack of explicit and formal representation in IoT knowledge may cause
• Ambiguity in terminology
• Hinder semantic interoperability of entities in the IoT world.
• Suite of Semantic Web technologies
• Ontologies, Semantic annotation, Linked Data and Semantic Web services
• Interoperability, Share-ability and Reusability of the same ontology with
heterogeneous IoT
entities.
• Ontology as Semantic Registry for IoT apps
IoT needs AI Backbone to
•
Typical Human Activities

Exercise

Travelling Laundry

Sleeping Cleaning

Eating Working
 Activity Recognition
• Activity recognition is the attempt to recognize actions
and motions of a user from observations using various
sensors.
• It has many different potential applications and the
possible connection to many different fields of study.
 Healthcare Applications
 Games and Robotics Industry
 Virtual Reality
 Home and Office Automation
 Security and Surveillance

Human Activity Recognition

Natural Language Processing (NLP)
NLP Operations
NLP Future
 Classifying AI Problems
• Supervised learning- Given a set of input/output
pairs, learn to predict the output if faced with a
new input e.g. classification
• Unsupervised Learning- Learning patterns in the
input when no specific output values are supplied
e.g. clustering
Classification
 Example: Spam Filter
• Input: an email Dear Sir.
• Output: spam/ham First, I must solicit your confidence in
this transaction, this is by virture of its
• nature as being utterly confidencial and
Setup: top secret. …
• Get a large collection of example emails,
each labeled “spam” or “ham” TO BE REMOVED FROM FUTURE
• Note: someone has to hand label all this MAILINGS, SIMPLY REPLY TO THIS
data! MESSAGE AND PUT "REMOVE" IN THE
• SUBJECT.
Want to learn to predict labels of new,
future emails 99 MILLION EMAIL ADDRESSES
FOR ONLY $99
• Features: The attributes used to
make the ham / spam decision Ok, I know this is blatantly OT but I'm
• Words: FREE! beginning to go insane. Had an old Dell
• Text Patterns: $dd, CAPS Dimension XPS sitting in the corner and
decided to put it to use, I know it was
• Non-text: SenderInContacts working pre being stuck in the corner,
• … but when I plugged it in, hit the power
nothing happened.
 Model-Based Classification
• Model-based approach
• Build a model (e.g. Bayes’
net) where both the label
and features are random
variables
• Instantiate any observed
features
• Query for the distribution of
the label conditioned on the
features

• Challenges
• What structure should the
BN have?
• How should we learn its
parameters?
 Artificial Neural Networks
ANN for Classification

Input Layer Hidden Layer Output Layer

I -1

I -2

Output
Input

I -3

I -n
MLP
• Synonym for Single-
Layer, Feed-
Forward Network
• Other networks were
known about but the
perceptron was the only
one capable of learning
• Notation can be simpler,
i.e.

O  Step0 j

WjIj
Recurrent Networks
• Feed forward networks:
• Information only flows one way
• One input pattern produces one output
• No sense of time (or memory of previous state)
• Recurrency
• Nodes connect back to other nodes or themselves
• Information flow is multidirectional
• Sense of time and memory of previous state(s)

• Biological nervous systems show high levels

of recurrence (but feed-forward structures
exists too)
Hidden Markov Models
• Markov chains not so useful for most agents
• Need observations to update your beliefs

• Hidden Markov models (HMMs)

• Underlying Markov chain over states X
• You observe outputs (effects) at each time step

X1 X2 X3 X4

E1 E2 E3 E4
Linear Support Vector Machines - Tennis
example

Temperatu
re

Humidi = play tennis

ty = do not play
 What is clustering?
• Organizing data into classes such that there
is
• high intra-class similarity
• low inter-class similarity
• Finding the class labels and the number of
classes
directly from the data (in contrast to
classification).
• More informally, finding natural groupings
 Defining distance measures
Definition: Let O1 and O2 be two objects from the
universe of possible objects. The distance
(dissimilarity) between O1 and O2 is a real number
denoted by D(O1,O2)
Peter Piotr

0.2 3 2 342.
AI: Important Concepts
• Data: labeled instances, e.g. emails marked spam/ham
• Training set
• Held out set
• Test set
• Features: attribute-value pairs which characterize each x
• Experimentation cycle
• Learn parameters (e.g. model probabilities) on training set
• (Tune hyper-parameters on held-out set)
• Compute accuracy of test set
• Very important: never “peek” at the test set!
• Evaluation
• Accuracy: fraction of instances predicted correctly

• Overfitting and generalization

• Want a classifier which does well on test data
• Overfitting: fitting the training data very closely, but not
generalizing well
• We’ll investigate overfitting and generalization formally in a few
lectures
Deep Learning

• Deep Learning
• For machine learning for higher prediction
accuracy
• A powerful class of machine learning model
• Modern reincarnation of artificial neural
network
• Collection of simple, trainable mathematical
functions
Deep Learning
Rules Based System
Hand
Input designe Output
d
Progra
m
Classic Machine Learning
Hand Mapping
Input designe from Output
d Feature
Feature s
s

Deep Learning
Additional Layers Mapping
Sample of more Abstract
Input Features from Output
Features
Feature
s
Deep Learning Software
 AI Problems
• AI systems greatly rely on quality of data
• Data pre processing technique selection is subjective
• AI systems still lack: broad understanding of the
world, common sense, ability to learn from very few
examples, truly out-of-the-box creativity…
• We don’t understand consciousness. (Does it
matter for AI?)
 Where to go next?
• How to continue:
• Data Science
• Probability
• Optimization
• Cognitive modeling
• Machine learning
theory
• Vision
• Robotics
• NLP
Conclusion: Adapt the Processes to AI

3
References
• Role of science and technology for future development. http://unu.edu
• BROWN, DAVID G. 2000. "The Jury Is In!" In Teaching with Technology,
• David G. Brown. Bolton, MA: Anker. CONNICK, GEORGE P. 1997. "Issues and Trends to
Take Us
into the Twenty-First Century." In Teaching and Learning at a Distance,
• Thomas E. Cyrs. San Francisco: Jossey-Bass. Innovations and new technology - what
is the role of research?Implications for public policy, vinnova ANA LYSis VA 2014:05
• Walker, W. (2000) “Entrapment in Large Technology Systems: Institutional
Commitments and Power Relations”, Research Policy, Vol. 29, Nos 7-8, pp. 833-
846
• VINNOVA (2011) “Utveckling av Sveriges kunskapsintensiva innovationssystem –
Bilagor.
• Underlag till forsknings- & innovationsproposition” VINNOVA Policy VP 2011:05
• Whitehead, A.N. (1925) “Science and the Modern World.” New York, Cambridge
University
• Press
• Teece, David J.; Pisano, Gary 1994: “The dynamic capabilities of Firms: an
Introduction”,
References
• Sörlin, S. and Vessuri, H. (eds) (2007) Knowledge Society vs. Knowledge Economy:
Knowledge, Power, and Politics. Knowledge, International Association of
Universities/New York: Palgrave Macmillan.
• Ravetz, J. (2006) When communication fails: a study of failures of global systems.
• Guimarães Pereira, A., Guedes Vas, S. and Tognetti, S. (eds) Interfaces between
Science and Society.
• “Technology and National Security: Risks and Responsibilities” By William J. Perry
Stanford
University
• Science, technology, and economic growth ARIEL PAKES* AND KENNETH L.
SOKOLOFF
• Science, Technology, and the Federal Government: National Goals for a New Era,
Committee on Science, Engineering, and Public Policy, http://www.nap.edu/9481
• High Level Group on the Modernization of Higher Education, New modes of learning
and
• teaching in higher education