Technologies of AI

Professor Ravee Choudhury

3rd May 2019

Second Talk: Webinar on “Artificial Intelligence- demystified”
Underpinning technologies of AI

• Technologies that power AI systems are

Expert systems
Machine learning
Natural language processing
Computer vision
Automated speech recognition
AI Planning
Expert systems

• Symbolic means : information is stored and transferred in symbolic

containers, or entities, rather than as raw data.
• Imagine a human anatomy database including these symbolic
entities: head, eyes and brain. Alone, these symbols are meaningless.
However, we add considerable meaning to our database with a
couple of associations.
Head contains(“brain”)
Head contains(“eyes”)
Eyes push_data(“brain”)
• Expert systems work though lists of ifs, thens and associations
written in human-like language.
•
Expert systems

• If we asked questions like “how many eyes do heads contain”, our

database would be no help.
• Expert system containing enough data points and data associations
has uses, especially in fields where other forms of AI are unavailable,
where use cases are simple or when human experts are readily
available for clarification.
Machine learning

• Machine learning covers a range of statistical techniques giving computers

the ability to learn. The machine progressively enhances its capacity to
execute a task.
• There are more than a dozen of these statistical techniques, one of which is
deep learning.
• Deep learning allows associations and characteristics to emerge from data
, whereas humans write expert systems using strings with conscious and
subconscious real-world meanings.
• These associations are possible, because deep learning uses artificial
neural networks (sets of nodes in a series of layers). Basically, deep learning
is nodes, layers and connections between those nodes and layers
Machine learning(contd…)

• Whereas an expert system could only answer questions if the

answers were given to it by a human, a machine learning could
answer them despite never being told the actual answers (depending
on its training data).
• If the training data was a series of photos of humans and animals with
species, heads, and eyes labeled, then the system could answer
questions like: what species is in this photo? And, for this species,
how many eyes do heads usually contain?
Natural-language processing

• Natural-language processing (NLP) is concerned with the interactions

between computers and human (natural) languages.
• NLP separates into 2 sub-categories: natural-language understanding
and natural-language generation, and is closely related to automated
speech recognition (ASR).
• NLP is the cornerstone of chatbots and virtual assistants (such as
Siri).
Computer vision

• Computer vision is a field concerning how computers see and

understand digital images and videos.
• Example of computer vision is Facebook’s facial recognition AI, which
is now human-like in recognising faces.
• Facebook has a big edge in the computer vision game , because (in
an age of machine learning) data is digital gold. They are now training
deep learning models using 3.5 billion Instagram photos and 17,000
hashtags.
Automated speech recognition

• Automated speech recognition (ASR) and Natural Language Processing

(NLP) are linked, but each is a unique category of AI..
• Imagine you are late for a meeting and you need to send a message to your
friend, so you open your phone, click Dictation and say “Hi John, I’m on my
way” and click send.
In this scenario, does your phone need to know ”Hi John” is a greeting?
Does your phone need to know “I’m on my way” means you’re moving
from one location to another location? No. All your phone needs to know is
the sounds you made correspond with particular words.
• In a nutshell, whereas NLP is concerned with the meaning of words, and
computer vision is concerned with recognising images and videos, ASR is
concerned with the meaning of sounds.
AI Planning

• Automated planning and scheduling (AI Planning) is a branch of AI

concerning strategies and action sequences.
• Self-driving cars and other autonomous robots need AI Planning to
operate.
What is Machine Learning ?

• Machine Learning (ML) is the science of getting computers to learn and act
like humans do, and improve their learning over time in autonomous
fashion, by feeding them data and information in form of observations and
real-world interactions.”
• The above definition encapsulates the ideal objective or ultimate aim of
machine learning.
• Machine Learning at its most basic is the practice of using algorithms to
parse data, learn from it, and apply what it has learnt to , make a
determination , informed decision or prediction about something in the
world.” – Nvidia
• “Machine learning is the science of getting computers to act without being
explicitly programmed.” – Stanford
What is Machine Learning ?(contd..)

• “Machine learning is based on algorithms that can learn from data

without relying on rules-based programming.”- McKinsey & Co.
• “Machine learning algorithms can figure out how to perform
important tasks by generalizing from examples.” – University of
Washington
• “The field of Machine Learning seeks to answer the question “How
can we build computer systems that automatically improve with
experience, and what are the fundamental laws that govern all
learning processes?” – Carnegie Mellon University
What is Machine Learning ? (contd..)

• The science of algorithms and statistical models that computer

systems use to effectively perform a specific task, without using
explicit instructions and relying on patterns in the data and relying on
inferences drawn from those data.
• Machine learning is learning in which machines learn on their own. It
provides the system, the ability to automatically learn and improve
from experience.
• Machine learning can be explained through another concept called
Data Mining . Data Mining is technique of examining a large pre-
existing data base and extracting new info from that database. ML
does the same.
Some examples of use of machine learning.

• Amazon uses ML to give better product choice recommendations to

Customers based on their preferences
• Netflix uses ML to give better suggestions to their users on what TV
programs or movies they may like based on their past choices.
Machine Learning Basic Concepts

• There are many different types of machine learning

algorithms, typically grouped by either learning style (i.e. supervised
learning, unsupervised learning, semi-supervised learning) or
by similarity in form or function (i.e. classification, regression,
decision tree, clustering, deep learning, etc.)
• Regardless of learning style or function, all ML algorithms consists of
the following:
Representation (a set of classifiers or a language that a
computer understands)
Evaluation (aka objective/scoring function)
Machine Learning Basic Concepts (contd…)

Optimization (search method; often the highest-scoring

classifier, for example; both off-the-shelf and custom
optimization methods are used)
• The fundamental goal of machine learning algorithms is
to generalize beyond the training samples i.e. successfully interpret
data that it has never ‘seen’ before.
How we get machines to learn ?

• There are different approaches for machines to learn, such as using

basic decision trees , clustering or layers of artificial neural
networks (the latter of which leads to deep learning). The learning
approach depends on what task you’re trying to accomplish and the
type and amount of data that you have.
• Emphasis is often placed on choosing the best learning algorithm.
Researchers have found that some of the most interesting questions
arise because none of the available machine learning algorithms
perform to par.
• Most times this is a problem with training data, but this also occurs
when working with machine learning in new domains.
How we get machines to learn (contd…)

• The research done when working on a real application often drives

progress in this field ( 1. Researchers discover the limitations of
existing methods 2. Researchers and Developers work with domain
experts to improve performance)
• Sometimes combinations of many learning algorithms are used to
improve accuracy
• Machine learning is not just, or even about, automation. If you think
this way, you’re bound to miss the valuable insights that machines
can provide and miss the resulting opportunities.(example: rethinking
an entire business model, for example, in industries like
manufacturing and agriculture).
How we get machines to learn (contd…)

• Machines that learn are useful to humans because, with all of their
processing power, they’re able to quickly highlight or find patterns in
big (or other) data, that would have otherwise been missed by
human beings.
• Machine learning is a tool that can be used to enhance humans’
abilities to solve problems and make informed inferences on a wide
range of problems, from helping diagnose diseases to coming up with
solutions for global climate change.
Challenges and limitations of machine learning

• Machine learning can’t get something from nothing…what it does is “get

more from less.” – Dr. Pedro Domingo, University of Washington
• The two biggest problems in machine learning is overfitting (in which the
learning model exhibits bias towards the training data and does not
generalize to new data, and variance i.e. the machine learns random
things , when trained using new data.
• Not having access to a large enough data set is a problem in some cases.
• When a learning algorithm is not working well , often the quicker path to
success is to feed the machine more data, the availability of which is a
primary driver of progress in machine and deep learning algorithms.
Challenges and limitations(contd..)

• However, this can lead to issues with scalability, in which we have

more data but time required to learn from that data remains an
issue.
Key Takeaways in Applying Machine Learning

• Here is a selection of best-practices and concepts of applying machine

learning. Some of these principles clarify how ML is used, and how to
avoid some common pitfalls that companies and researchers might
be vulnerable to in starting an ML-related project.
• The most important factor in successful machine learning projects is
the features used to describe the data (which are domain-specific),
and having adequate data to train the models in the first place.
• Most of the time when algorithms don’t perform well, it’s due a to a
problem with the training data (i.e. insufficient amounts/skewed data;
noisy data; or insufficient features describing the data for making
decisions.
Key Takeaways in Applying Machine Learning (contd…)

• Obtaining experimental data (as opposed to observational data, over

which we have no control) should be done if possible (for example,
data gleaned from sending different variations of an email to a
random audience sampling).
• Always set aside a portion of your training data set for cross
validation; you want your chosen classifier or learning algorithm to
perform well on fresh data.