Summary
Summary AI in Action Radboud University IBC
- Course
- Institution
This is a clear summary of all lectures of the AI in action course. I got an 8,5 with this summary.
[Show more]
Seller
Follow
lauralisa
Reviews received
Content preview
Lecture 1- IntroductionDefinition of AI
AI system= A system that has a capacity usually associated with human
intelligence.
There are many different definitions and directions of AI and AI research. Examples:
- Creating thinking systems
- Understanding intelligence by synthesizing it
AI Today
We currently live in the AI revolution, because it can be used more.
The AI revolution is possible, because computers have become more powerful and
more data has become available (for example with rise of the WWW).
Before the AI revolution, we were in the ‘AI winter’= Researchers knew potential of
AI, but didn’t have the tools to move forward.
Forms of AI: Voice assistants, self-serving vacuum cleaner, job recommendations, machine
translation, text to speech programmes or targeted advertising.
The role of algorithms in AI
AI is based on algorithms= A set of instructions.
Symbolic AI
The AI system performs operations on symbols (like
words)
These operations are understandable to humans.
Decision trees are a form of symbolic AI
The symbols are concepts represented as words
Each layer presents an if-then decision= Implication
Sub-symbolic AI
The AI system performs operations on numbers. The operations take the form of
equations (which are represented as numbers).
These operations are not understandable to humans.
The operations are able to capture patterns in the data a human would not really
see when analysing the data by hand.
In this way, sub-symbolic AI is supposed to move towards human’s unconscious
thought processes, but this leads to less understandability about the process
Data and machine learning are used
Important when a domain can’t be captured through symbols.
AI and machine learning are not the same. Machine learning is a type of AI.
,What is AI?
Generalized AI Specific AI
A system with: A system with:
-No restrictions on topics of its knowledge. Can -Restricted to a specific domain. Like job
answer any question. recommendations.
-Can do anything a human can do. -Use its knowledge to adapt and achieve
-Not realistic, more from movies. specific goals
-Also called General AI or Strong AI -Realistic, more from real world.
-Also called Narrow AI or Weak AI
AI in the news
‘Amazon’s Alexa assistant told a child to put a coin in a socket’. Voice assistants use
both symbolic and sub-symbolic AI.
No current AI is close to understanding the everyday physical world. We now have an
approximation to intelligence, not the real thing. Therefore, it will never really be
trustworthy. We will need some fundamental advances before we can get to AI we
can trust.
Machine learning: AI that learns from data
A machine is said to have artificial intelligence if it can interpret data, potentially
learn from the data and use that knowledge to adapt and achieve specific goals.
Today much of the AI is based on machine learning, specifically on a sub-category of
machine learning called deep learning. Learning here means to improve itself.
The machine learning algorithm learns from data that has a label= Labelled data.
Robot teacher which is teaching the AI the distinction between images that contain
bees and images with threes. It shows the images with the label so many times that
the AI starts to recognize pictures with threes and bees itself.
Labelled data= Data that the machine learning system learns from, which is also
called training data. It sees data with strawberries and marshmallows labelled as
tasty and play-doh and shampoo labelled as gross.
RODEO: Facets of AI systems in action
RODEO is an acronym that captures the facets important when discussing an AI system:
Rationale= The reasons for which the AI exists
Includes the purposes of the system:
- What goal does the system accomplish for its users?
- How does the system serve the people who operate it?
Any other involved controlling principles, like belief, opinion, practice or phenomena
(including the long-term or high-level objectives of the system and how it is funded)
Ethical issues related to the rationale
Example: Voice Assistants. SIRI is there to help the client and to make money for apple. Also,
it collects voice data from you, which is used to improve its own voice.
Belief underlying SIRI: We cannot ask SIRI certain questions (like how should I plan my life)
and therefore we should strive for broader capabilities
Operations= How the AI system works and continues to work
, Computer system the AI runs on
The resources it needs
Types of algorithms it uses
How the quality of the system is tested
How quality is maintained during operations
Ethical aspects including how and when the system will be taken out of use
Example: Roomba independent vacuum cleaner. You have a Roomba, what happens when it
moves to a new house, do people trip over it, how often do you have to empty it? All the stuff
that makes the Roomba keeps operating is important.
Data= The data that is used to train the AI system
Who produced the data?
Who owns the data?
How is quality ensured?
How is the data kept up-to-date?
For symbolic AI describe the knowledge base
Ethical issues related to data
Targeted advertising runs on our data. It is based on our clicks. It is updated with new clicks
every time. Data may be privacy sensitive.
Roomba: The data is mapping out your house and constantly updates this. This is privacy
sensitive as well.
Enrichment= Additional information that is added to the data inside the AI system
There are 2 types of enrichment:
1. Gathering labels needed to train the system. They can be human-generated or
collected.
2. Using an AI system to generate labels that then feeds another system. For example,
labelling an image as cat to feed a pet recommendation system.
Ethical issues related to enrichment.
Machine translation: In translation systems you can correct translations or if you close
Google after the translation, the system assumes the translation was correct.
Output= The decision, label or other product produced by the AI system
The interface used to present results
Explanation that is produced which explains why the AI system made the decision
The reliability of the output: is it good enough to serve its purpose?
Ethical issues including trustworthiness and bias
In targeted advertising why are two ads shown together? Were they used together or
indepently?
What AI is not
AI is not magic
AI is not new= Has been there for longer in different forms
AI is not general= It doesn’t know something about everything
AI is not humanlike= It makes mistakes a human wouldn’t make
AI is not wise= It learns from pattern and data but has no common sense, like what is
good and bad
, AI is not independent of human knowledge= It doesn’t know what humans don’t
already know
AI is not impartial= AI is biased
AI is not always transparent= We get the decision ‘it’s a cat’, but we don’t know how
the system came to that decision
AI is not always necessary= You may not need a movie recommender system
AI is not inevitable= Humans are making the decisions to use AI systems, we don’t
have to use them
AI is not restricted to commercial applications= It can also be used for cultural
heritage archives
AI is not independent of human needs= A robot may have been made because a
human wanted a friend. Companies create AI to get more market power etc.
Lecture 2- Trustworthy and ethical AI
Introduction
AI makes decisions that affect our lives. It creates our news feeds, diagnoses our diseases,
evaluates loan application. How do you know that AI makes trustworthy decisions?
Automobile, automation, autonomy
Automobile= Self-moving or self-movable.
Automation= Taking a process that requires human action and making it self-
regulating.
Autonomous= When something exists separately and has power to govern or
control itself.
Future of AI
AI provides society with benefits: Speech transcription, GPS navigation, trip planning, email
spam filters, language translation, credit card fraud alerts etc.
But there is more for the future.
Future of AI in health: Assisting physicians in diagnosis, suggesting treatments,
discovering new medicine, monitoring the health of elderly.
Future of AI in science: Scientific modelling and data analysis will increasingly rely on
AI. Improving models of climate change and population growth. Supporting studies of
ecology and food science.
Future of AI in replacing humans: 2 perspectives
1. AI is smarter than humans= There are problems that are too complex for humans to
solve. AI can assist in the solution.
2. AI is more robust= There are jobs that are boring and dangerous, for which humans
could be replaced by AI.
Repetitive factory tasks, self-driving trucks, robots for harvesting fruits etc.
Automation has replaced humans before, such as with washing machines, rickshaw driver
with a motorcycle. We don’t have to intervene on the process anymore.
What will we sacrifice to gain the benefits of AI?
Andrew Ng is an AI expert who said ‘AI is the new electricity’, meaning it will change
everything.
There is a difference between electricity and AI:
The benefits of buying summaries with Stuvia:
Guaranteed quality through customer reviews
Stuvia customers have reviewed more than 700,000 summaries. This how you know that you are buying the best documents.
Quick and easy check-out
You can quickly pay through credit card or Stuvia-credit for the summaries. There is no membership needed.
Focus on what matters
Your fellow students write the study notes themselves, which is why the documents are always reliable and up-to-date. This ensures you quickly get to the core!
Frequently asked questions
What do I get when I buy this document?
You get a PDF, available immediately after your purchase. The purchased document is accessible anytime, anywhere and indefinitely through your profile.
Satisfaction guarantee: how does it work?
Our satisfaction guarantee ensures that you always find a study document that suits you well. You fill out a form, and our customer service team takes care of the rest.
Who am I buying these notes from?
Stuvia is a marketplace, so you are not buying this document from us, but from seller lauralisa. Stuvia facilitates payment to the seller.
Will I be stuck with a subscription?
No, you only buy these notes for $6.96. You're not tied to anything after your purchase.
Can Stuvia be trusted?
4.6 stars on Google & Trustpilot (+1000 reviews)
67474 documents were sold in the last 30 days
Founded in 2010, the go-to place to buy study notes for 14 years now