Ben Byford
AI ethics | freelance web and games designer
An Introduction to
Artifical Intelligence
& Machine Learning
Day 1
What is and isn’t AI
What is ML
Kinds of data & importance or good data
Considering ML uses, harnessing it in your projects
The Ethics of AI
Who's in the room?
What is your knowledge of AI / ML?
What is AI?
(discuss)
Clever stats
Learning programs that try to optimise towards a given goal
*(they are not bestowed with intelligence
and don't have any inner intents)
Why even do AI?
(discuss)
automated cars, health care robots, health imaging, preventive health, mental health, personalised medicine, image recognition, pose detection, face detection, computer games, architecture, trading, spam dectection, sex robots, arts, recruiting, automated life, artifical model simluation for engines, translation, automated logistics
Some of these things are totally new and
are enabled by AI
AI is both a broad discipline
&
a goal of creating human like intelligent agents
Considered research areas within AI:
-
Symbolic knowledge graphs / expert systems
-
Automata
-
Genetic algorithms
-
Whole brain simulation
-
Machine learning
A lot of the time we’re actually talking about ML
ML is a subset of AI which is a range of algorithms that learn from data
Data science is not sexy - 80% 20%.
20% is the fun model making bit
80% is asking the right questions, collecting / finding data, cleaning data, preparing data, communicating findings, repeat...
The sexy bit: types of Machine learning algos
(yes people do say algos, no we're not friends)
Supervised learning
Unsupervised Learning
Reinforcement Learning
Semi-supervised Learning
(person-in-the-loop learning)
The sexy bit: types of Machine learning algos
(yes people do say algos, no we're not friends)
Supervised learning
Linear regression
Nearest neighbour
Support Vector Machines (SVM)
Decision trees / Random forests
Neural networks
Unsupervised Learning
k-means clustering
Association Rules
Reinforcement Learning
Q-Learning
Temporal Difference (TD)
Deep Adversarial Networks
Semi-supervised Learning
(person-in-the-loop learning)
Non-exhaustive list
https://www.youtube.com/watch?v=gn4nRCC9TwQ
Lunch
Types of data
Unstructured
Structured
webcam emoji
(demo)
Transfer learning
GANS
(Generative adversarial network)
(Brief 10mins in pairs)
How does it work?
How could the algorithm be better?
How could the application be better?
(Brief 10mins in pairs)
How could you use Machine Learning
to create your own applications,
artworks or experiences?
AI Ethics
Welcome to my world!
“It is impossible to work in information technology without also engaging in
social engineering.”
Jaron Lanier
Where are these decision getting made?
Govenments
Business
Design
Technology
Data
Govenments
- AI in Warfare
- Regulation of malpractise
- Best practise / investing in research – sending signals
- How can the technology be misused by bad actors?
- Tracking bad actors and security
- Democratisation or tools and access
Business
- Personal data
- Using other companies algorithms
- Company acquisition
- Team diversity
- What does success look like?
- Monopolies of algorithmic power
- Impact on currently held freedoms e.g free speech, democracy
- Who is accountable, what is the visibility of that account?
- Secrecy
Design
- How to implement and where?
- Unintended consequences
- Can the users interpret the system?
- Is it exploitative?
- What can you do with false positives and negatives?
- Is the system design dogmatic?
- Discriminatory?
- What behaviours is the system promoting?
Technology
- How can the algorithm fail?
- Does the system change over time and does it need to be repeatedly tested for changes?
- Is the output explainability? Does it need to be?
Data
- What bias do you have in the data?
- Discriminatory?
- Do you have enough data?
- Is the data reliable?
- Is it appropriate for the domain use?
- Consent of data usage?
- Inference of personal data / personhood
- What score is good for your domain
Ethical agency (philosophy questions)
- To whom and when these obligations are directed
- To whom and when are rejecting obligations permissable?
Super-intelligence / Singularity
- Value alignment
- Off switch problem
- Hard problem of conciousness / robot rights
(5 mins discuss)
What excites you
and
what troubles you about AI?
Thanks!
questions?
AI / ML workshop
By Ben Byford
AI / ML workshop
Workshop to introduce AI / ML, theory, uses and ethics
