4005PY Artificial Intelligence
What is Artificial Intelligence?
Machine-based system that can, for a given set of human-
defined objectives, make predictions, recommendations, or
decisions influencing real/ virtual environments. AI
systems are designed to operate with varying levels of
autonomy.
(OECD, 2019)
*can look at Hao 2018a
Examples of AI: Roomba, self-driving cars (ie. Tesla),
voice-systems (ie. Alexa, google), tiktok, smart-home (ie.
Thermostats), facial recognition
*refer to slides for case study
Particularly in Japan – robots – older population
Machine Learning
(Hao, 2018c) Machine-learning algorithm uses statistics to
find patterns In massive amounts of data.
Encompasses lots of things – numbers, words,
images, clicks…it can be digitally stored, fed into a
machine-learning algorithm.
4 types of Machine Learning (WHO, 2021)
, (1) Supervised learning – data used to train the
model are labelled (outcome variable is known).
a. Model infers a function from the data that can be
used for predicting outputs from different inputs.
(2) Unsupervised learning – labelling not included,
instead identifies hidden patterns in the data by the
machine.
(3) Reinforcement learning – machine learning by
trial & error to achieve an objective for which the
machine is ‘rewarded’ or ‘penalised’, depending on
whether inferences reach or hinder achievement of an
objective.
(4) Deep learning/ deep-structured learning – family
of machine learning based on multi-layered models =
progressively extract features from the data
a. Can be supervised, unsupervised, semi-
supervised.
b. Requires large amounts of data to be fed into the
model
What is an Algorithm
Algorithm – process/ set of rules to be followed in
calculations or other problem-solving operation, especially
by a computer (Oxford English Dictionary)
2
, Statistical models – used to develop algorithms; use linear
and logistic regression to build an AI algorithm (Ghandi,
2018)
Real Life applications of AI
(1) Diagnosis
a. AI can be trained w/ old computerized
tomography (CT) scans
i. Of patients who were initially given the all
clear – but then went on to develop cancer.
b. AI is able to learn the subtle signs of cancerous
lesions, which no doctor was able to do
(Svoboda, 2020).
c. AI systems that have been developed can identify
cancer correctly 94% of the time – outperforming
experienced radiologists (65%) (Ardila et al.,
2019)
(2) AI & the Ageing Population
a. Proportion of younger people available to enter
the health & care workforce is not increasing as
fast
i. Struggle to support growing older population
(3) Care Robots
a. Robots carry out a no. of tasks/ to support
human carers.
i. Companionship, therapy, health & activity
monitoring etc. (Kyrarini, 2021)
b. Not routinely used across the world due to costs;
concerns about safety standards, patient privacy
etc. (Kyrarini, 2021)
3
What is Artificial Intelligence?
Machine-based system that can, for a given set of human-
defined objectives, make predictions, recommendations, or
decisions influencing real/ virtual environments. AI
systems are designed to operate with varying levels of
autonomy.
(OECD, 2019)
*can look at Hao 2018a
Examples of AI: Roomba, self-driving cars (ie. Tesla),
voice-systems (ie. Alexa, google), tiktok, smart-home (ie.
Thermostats), facial recognition
*refer to slides for case study
Particularly in Japan – robots – older population
Machine Learning
(Hao, 2018c) Machine-learning algorithm uses statistics to
find patterns In massive amounts of data.
Encompasses lots of things – numbers, words,
images, clicks…it can be digitally stored, fed into a
machine-learning algorithm.
4 types of Machine Learning (WHO, 2021)
, (1) Supervised learning – data used to train the
model are labelled (outcome variable is known).
a. Model infers a function from the data that can be
used for predicting outputs from different inputs.
(2) Unsupervised learning – labelling not included,
instead identifies hidden patterns in the data by the
machine.
(3) Reinforcement learning – machine learning by
trial & error to achieve an objective for which the
machine is ‘rewarded’ or ‘penalised’, depending on
whether inferences reach or hinder achievement of an
objective.
(4) Deep learning/ deep-structured learning – family
of machine learning based on multi-layered models =
progressively extract features from the data
a. Can be supervised, unsupervised, semi-
supervised.
b. Requires large amounts of data to be fed into the
model
What is an Algorithm
Algorithm – process/ set of rules to be followed in
calculations or other problem-solving operation, especially
by a computer (Oxford English Dictionary)
2
, Statistical models – used to develop algorithms; use linear
and logistic regression to build an AI algorithm (Ghandi,
2018)
Real Life applications of AI
(1) Diagnosis
a. AI can be trained w/ old computerized
tomography (CT) scans
i. Of patients who were initially given the all
clear – but then went on to develop cancer.
b. AI is able to learn the subtle signs of cancerous
lesions, which no doctor was able to do
(Svoboda, 2020).
c. AI systems that have been developed can identify
cancer correctly 94% of the time – outperforming
experienced radiologists (65%) (Ardila et al.,
2019)
(2) AI & the Ageing Population
a. Proportion of younger people available to enter
the health & care workforce is not increasing as
fast
i. Struggle to support growing older population
(3) Care Robots
a. Robots carry out a no. of tasks/ to support
human carers.
i. Companionship, therapy, health & activity
monitoring etc. (Kyrarini, 2021)
b. Not routinely used across the world due to costs;
concerns about safety standards, patient privacy
etc. (Kyrarini, 2021)
3
