This document provides an overview of artificial intelligence and machine learning. It begins by defining AI as computer systems that can perform tasks autonomously and adaptively. Machine learning is described as getting computers to learn without being explicitly programmed. Examples of machine learning in daily life are discussed. The basics of supervised and unsupervised learning are explained. Ethical issues around AI like bias, fairness, and determining appropriate use are then discussed. Options for addressing these issues like ensuring diversity of data and viewpoints are presented. The document concludes by providing recommendations for further learning.
1. Mark S. Steed,MA (Cambridge) MA (Nottingham), MSc (Ashridge-Hult Business School)
Director, JESS, Dubai
@JESSDubai
@JESS_Director
@IndependentHead
The Ethics ofAI
JESS DubaiToK Lecture
Monday4th March2019
2. The Rise ofAI
1. What is AI and
how does it work?
2. The Ethics of
Artificial
Intelligence
6. What isAI?
Artificial Intelligence refers to those
computer systems which are both
autonomous and adaptive.
Autonomy:
The ability to perform tasks in complex environments
without constant guidance by a user.
Adaptivity:
The ability to improve performance by learning from
experience.
7. What isAI?
Machine Learning is the science of
getting computers to learn, without being
explicitly programmed.
12. Machine
Learning
Machine Learning is the science of
getting computers to learn, without being
explicitly programmed.
Supervised
Learning
has a data
training set
Unsupervised
Learning
has no data
training set
13. Machine
Learning
Supervised
Learning:
We are given an input, for
example a photograph
with a traffic sign, and the
task is to predict the
correct output or label, for
example which traffic sign
is in the picture (speed
limit, stop sign, ...).
Machine Learning is the science of
getting computers to learn, without being
explicitly programmed.
15. Supervised
Machine
Learning
How Supervised Machine Learning works:
Nearest Neighbour to predict behaviour
Training Set Test Set Nearest Neighbours
Age
Blood Sugar Level
Source: University of Helsinki
Elements of AI course
16. Supervised
Machine
Learning
How Supervised Machine Learning works:
Regression to predict behaviour
Source: University of Helsinki
Elements of AI course
Student ID Hours Studied Pass/ Fail
A 15 Pass
B 9.5 Pass
C 2 Fail
D 5 Fail
E 6.5 Fail
F 6 Pass
…… …… …….
ZZ 10 Fail
17. Supervised
Machine
Learning
How Supervised Machine Learning works:
Regression to predict behaviour
Source: University of Helsinki
Elements of AI course
If you want to have an 80% chance of passing the
exam, how many hours should you study for?
18. Supervised
Machine
Learning
How Supervised Machine Learning works:
Classification – defining the Decision Boundary
Source: Stanford University Coursera
Machine Learning course
Definition of the Decision Boundary:
“Present” when X1 + X2 ≥ 3
“Not present” when X1 + X2 < 3
Machine Learning
works out the
Decision Boundary
formula by
examining the
training data and
deciding the best fit.
22. Machine
Learning
Unsupervised
Learning:
There are no labels or correct
outputs.The task is to
discover the structure of the
data: for example, grouping
similar items to form
“clusters”, or reducing the
data to a small number of
important “dimensions”.
Machine Learning is the science of
getting computers to learn, without being
explicitly programmed.
Supervised
Learning:
We are given an input, for
example a photograph
with a traffic sign, and the
task is to predict the
correct output or label, for
example which traffic sign
is in the picture (speed
limit, stop sign, ...).
27. The Ethics of
Artificial
Intelligence
Ethics of Classification Problems
1 = yes
0 = no
X XX X X X
X XX X XX
Hours of Study
Passing?
Likelihood of passing exam (pass, fail)
Source after : University of Helsinki
Elements of AI course
28. The Ethics of
Artificial
Intelligence
Ethics of Classification Problems
Breast Cancer (malignant, benign)
Source: Stanford University Coursera
Machine Learning course
1 = yes
0 = no
X XX X X X
X XX X XX
Tumour Size
Malignant?
32. The Ethics of
Artificial
Intelligence
Source: McKinzey Podcast on
Ethics of Artificial Intelligence
1. Bias
Dataset doesn’t reflect main population
Algorithmic Bias
Three levels of Concern
33. The Ethics of
Artificial
Intelligence
Source: McKinzey Podcast on
Ethics of Artificial Intelligence
2. Fairness
Training Dataset is based historical data
which reflects unfair practice.
Algorithmic Bias
Three levels of Concern
35. Policing and Recidivism
Recidivism: Likelihood of Reoffending:
Prisoner Questionnaire:
• How many prior convictions have
you had?
• What part did others pay in the
offence?
• What part did alcohol or drugs play?
36. The Ethics of
Artificial
Intelligence
Source: McKinzey Podcast on
Ethics of Artificial Intelligence
2. Fairness
Training Dataset is based historical data
which reflects unfair practice.
Algorithmic Bias
Three levels of Concern
37. The Ethics of
Artificial
Intelligence
Source: McKinzey Podcast on
Ethics of Artificial Intelligence
2. Fairness
Training Dataset is based historical data
which reflects unfair practice.
1. Bias
Dataset doesn’t reflect main population
3. Unethical
Data model is deliberately skewed or
behaves dishonourably.
Algorithmic Bias
Three levels of Concern
38. The Ethics of
Artificial
Intelligence
1. Diversity of Background of theTeam
(to avoid “group think”)
2. Diversity of Mindset
(personality testing)
3. Diversity of Data
4. Diversity of Algorithmic Models
Source: McKinzey Podcast on
Ethics of Artificial Intelligence
How do we avoid Algorithmic Bias?
Diversity is the key
42. The Ethics of
Artificial
Intelligence
Awad et al ‘The Moral Machine’
NatureVol 563 November 2018
Who decides? - We all do
The Moral Machine Experiment platform
gathered 39.61 million decisions in ten languages
from millions of people in 233 countries
43. The Ethics of
Artificial
Intelligence
1. sparing humans (versus pets),
2. staying on course (versus swerving),
3. sparing passengers (versus pedestrians),
4. sparing more lives (versus fewer lives),
5. sparing men (versus women),
6. sparing the young (versus the elderly),
7. sparing pedestrians who cross legally
(versus jaywalking),
8. sparing the fit (versus the less fit), and
9. sparing those with higher social status
(versus lower social status).
Source: Awad et al ‘The Moral Machine Experiment’
NatureVol 563 November 2018
A 13-accident session: 9 Core Scenarios
44. The Ethics of
Artificial
Intelligence
Awad et al ‘The Moral Machine’
NatureVol 563 November 2018
Who will you spare?
https://create.kahoot.it/share/the-moral-machine-
experiment/25d61376-9279-4e89-98c9-1b8a0f84d934
46. The Ethics of
Artificial
Intelligence
‘But given the strong preference for sparing
children, policymakers must be aware of a dual
challenge if they decide not to give a special
status to children:
the challenge of explaining the rationale for
such a decision, and
the challenge of handling the strong backlash
that will inevitably occur the day an
autonomous vehicle sacrifices children in a
dilemma situation.’
p.60
Awad et al ‘The Moral Machine’
NatureVol 563 November 2018
49. The Ethics of
Artificial
Intelligence
Centre for Data Ethics and
Innovation (CDEI)
The Centre for Data Ethics and Innovation
(CDEI) is an advisory body set up by
Government and led by an independent
board of expert members to investigate and
advise on how we maximise the benefits of
data-enabled technologies, including
artificial intelligence (AI).
55. Mark S. Steed MA (Cambridge), MA (Nottingham), MSc (Ashridge-Hult Business
School)
Director of JESS Dubai
Email: mss@jess.sch.ae
Twitter: @JESS_Director
@independenthead
LinkedIn: uk.linkedin.com/in/independenthead
Blog: http://independenthead.blogspot.com
SlideShare: http://www.slideshare.net/independenthead