workshop understanding artificial intelligence · to learn about its applications. artificial...

UNDERSTANDINGARTIFICIALINTELLIGENCE

WORKSHOP

INTRODUCTION: ARTIFICIAL INTELLIGENCE

WHAT IS AI AND WHY IT IS IMPORTANT TO LEARN ABOUT ITS APPLICATIONS

ARTIFICIAL INTELLIGENCEIS IT ANALYSIS?

ARTIFICIAL INTELLIGENCEIS IT ANALYSIS?

ARTIFICIAL INTELLIGENCEIS IT PREDICTION BASED ON ANALYSIS?

ARTIFICIAL INTELLIGENCEIS IT PREDICTION BASED ON ANALYSIS?

ARTIFICIAL INTELLIGENCELEARNINGREASONINGSELF-CORRECTION

ARTIFICIAL INTELLIGENCEINTELLIGENT SYSTEMS

WHAT IS DATA?DATA IS THE DIGITAL REPRESENTATION OF THE WORLD

WHAT IS DATA?THE DIGITAL REPRESENTATION OF THE WORLD

TEAM JEROME TEAM ANANDAN

WHAT IS AN ALGORITHM?AN ALGORITHM IS A PROCEDURE OR FORMULA FOR SOLVING A PROBLEM, BASED ON CONDUCTING A SEQUENCE OF SPECIFIED ACTIONS.

WHAT IS AN ALGORITHM?THINK OF IT AS FOLLOWING A RECIPE

WHAT IS MACHINE LEARNING?AN APPLICATION OF AI THAT PROVIDES SYSTEMS THE ABILITY TO AUTOMATICALLY LEARN AND IMPROVE FROM EXPERIENCE WITHOUT BEING EXPLICITLY PROGRAMMED.

MACHINE LEARNINGCAPABILITIES

MACHINE LEARNING IN USEEMAIL SPAM FILTERING

MACHINE LEARNING IN USEFRAUD DETECTION

MACHINE LEARNING IN USEFACIAL RECOGNITION

WHY USE MACHINE LEARNING?

Problems for which existing solutionsrequire a lot of hand-tuning or long lists of rules

Complex problems for which there is no good solution at all using a traditional approach

Fluctuating environments: a Machine Learning system can adapt to new data

Getting insights about complex problems and large amounts of data

TYPES OF MACHINE LEARNING SYSTEMSMACHINE LEARNING SYSTEMS CAN BE CLASSIFIED ACCORDING TO THE AMOUNT AND TYPE OF SUPERVISION THEY GET DURING TRAINING.

MACHINE LEARNING SYSTEMSFOUR MAJOR CATEGORIES

1. SUPERVISED LEARNING

MACHINE LEARNING SYSTEMS

In supervised learning, the training data you feed to the algorithm includes the desired solutions, called labels.

A typical supervised learning task is classification. The spam filter is a good example of this: it is trained with many example emails along with their class (spam or ham), and it must learn how to classify new emails.

SUPERVISED LEARNINGEXAMPLE: CLASSIFICATION

2. UNSUPERVISED LEARNING

MACHINE LEARNING SYSTEMS

In unsupervised learning, as you might guess, the training data is unlabeled. The system tries to learn without a teacher.

A typical supervised learning task is clustering. For example, say you have a lot of data about your blog’s visitors. You may want to run a clustering algorithm to try to detect groups of similar visitors. At no point do you tell the algorithm which group a visitor belongs to: it finds those connections without your help.

UNSUPERVISED LEARNINGEXAMPLE: CLUSTERING

3. SEMISUPERVISED LEARNING

MACHINE LEARNING SYSTEMS

Some algorithms can deal with partially labeled training data, usually a lot of unlabeled data and a little bit of labeled data. This is called semisupervised learning.

Some cloud hosting providers, such as Google Photos, are good examples of this. Once you upload all your family photos to the service, it automaticallyrecognizes that the same person A shows up in photos 1, 5, and 11, while another person B shows up in photos 2, 5, and 7.

SEMISUPERVISED LEARNINGEXAMPLE: CLUSTERING + CLASSIFICATION

4. REINFORCEMENT LEARNING

MACHINE LEARNING SYSTEMS

Reinforcement Learning is a little different. The learning system, called an agent, can observe the environment, select and perform actions, and get rewards in return (or penalties in the form of negative rewards). It must then learn by itself what is the best strategy, called a policy, to getthe most reward over time.

A policy defines what action the agent should choose when it is in a given situation.

REINFORCEMENT LEARNINGEXAMPLE: POSITIVE AND NEGATIVE REWARDS

MACHINE LEARNING: WHAT TO WATCH OUT FOR

In Machine Learning projects, we usually select a learning algorithm and train it on some data, so the things that can go wrong are related to bad algorithmsand bad data. However, data often matters more than algorithms for complex problems.

Let’s look at some examples of what we mean by bad data.

INSUFFICIENT DATANOT ENOUGH DATA

NON-REPRESENTATIVE DATADOES NOT REPRESENT ALL THE CASES

OVER-FITTING DATAFORCE-FITTING DATA TO MATCH THE USE CASE

UNDER-FITTING DATADATA IS TOO SIMPLE TO ACCOUNT FOR VARIANCE

POOR QUALITY DATADATA IS FULL OF ERRORS, OUTLIERS, AND NOISE

APPROACHINGPROBLEMS THROUGH AN AI LENS

7 QUESTIONSTO CONSIDER

ARTIFICIAL INTELLIGENCETECHNICAL LIMITATIONS

• Obtaining sufficiently large data sets

• The need to label training data

• Difficulty explaining results from large, complex neural-network-based systems

• Difficulties with domain adaptation and generalising

• Risk of discrimination and bias

• Privacy concerns

• Data quality, quantity, completeness

DISCRIMINATION AND BIAS IN AIEXAMPLE

DISCRIMINATION AND BIAS IN AIEXAMPLE

QUESTIONS?LET’S TALK ABOUT AI :)

THANK YOU.

DR. JEROME WHITESENIOR RESEARCHERWADHWANI AI

jerome@wadhwaniai.org