Artificial Intelligence
Week 1................................................................................................................................. 2
Week 2................................................................................................................................. 6
Week 3................................................................................................................................. 9
Week 4............................................................................................................................... 12
Week 5............................................................................................................................... 15
Week 6............................................................................................................................... 19
Pseudocodes...................................................................................................................... 23
Recap and Q&A............................................................................................................... 25
Brightspace Practice Exam......................................................................................................... 27
This summary was created in block 2 of the academic year 2023-2024. I attended all the
lectures to make this summary, using the slides and transcribed what the professor said to make it
detailed enough to understand concepts like feedforward networks or optimal search.
I also added a chapter in which I copied/pasted all the pseudocodes that were used in the
lectures as that may come in handy during the exams. Lastly, some random notes of the last
lectures are added in the Recap and Q&A chapter, relating to parts that the professor mentioned
or questions that were asked at that moment.
At the end, I added the practice questions that were put on brightspace and what I believe
are the correct answers to these.
, Week 1
Introduction, History of AI
4 approaches to AI:
● Thinking humanly
● Thinking rationally
● Acting humanly
● Acting rationally
What is artificial intelligence and its purpose? For this, it’s important to look at cognitive
psychology first. Cognitive psychology is the study of the computations that make it possible to
perceive, reason, and act.
Artificial intelligence can be seen as a branch of philosophy, computer science, or more
and mostly draws from the disciplines of psychology and computer science to put emphasis on
the computation of perception. It studies how to build programs that enable computers to do what
the mind can do.
The problem of psychology is that it can be seen as an inverse problem; we analyze
behavior and reason retrospectively about the causes of it. However, such limitations are limited
and sometimes impossible to make. AI, on the other hand, uses forward modeling. This can be
seen as designing a simple system and seeing later how it behaves. Because we designed it, we
know how it works.
Forward modeling involves thinking algorithmically; using a set of rules that
unambiguously define a sequence of operations and hyper specific commands. In those cases, as
long as you follow the rules precisely, you can do it by hand. It thus matches input and output
through a series of processes. These can also be seen as functions, which are a list of instructions
that take input values (arguments) that apply some computations to create an output value.
Similarly, humans use input and output; respectively sensory perceptions and behavior.
, Descartes’ theory was defined as Cartesian dualism. This raised the question of how the
physical brain gives rise to the mental mind. On the other hand, materialists believed that all
mental states were caused by physical states. If we accepted materialism to be true, we could
assume that we could always recreate mental states from physical states. Throughout history, this
has brought many attempts to recreate systems of computations.
In the 1940s, McCulloch and Pitt worked with three principles; basic physiology,
propositional logic, and Turing’s theory of computation. From those times, it was believed that
any computable function could be computed by a network of artificial neurons. All logical
operators could be implemented by simple neural networks.
Turing believed that AI were weak, but intelligent if they were indistinguishable from
humans in conversations. Complex grammatical structures and realistic world knowledge would,
according to him, prove intelligence.
John Seaerle believes that true understanding requires the actual chemistry and physical
properties of the human brain. He came up with the Chinese room experiment. The Chinese
Room argument imagines a person inside a closed room in China. Someone on the outside could
ask questions and communicate with the person inside in Chinese. The person inside could use a
book to translate what is said to him. This would allow him to answer every possible question
that he is asked. This is basically what computers do. To an outsider, it may seem like something
inside understands Chinese, even if this is done through a book or by a computer. This is an
argument for Searle that computers do have true intelligence or sentience; they can only simulate
it through functions. Turin had an opposite point of view, as he believed that as long as
behavioral output is indistinguishable from a human, a machine is smart. Searle’s concept is also
known as weak AI. Strong AI is the concept that an appropriately programmed computer with
the right inputs and outputs would have a mind in the same sense as human beings have minds. It
thus suggests that an intelligent system can actually think.
It is important to distinguish between weak AI and strong AI. In the paper by Turing, he
introduced what was called the imitation game. This game implied that a machine is intelligent if
we cannot distinguish it from humans in a conversation. This is still quite impossible to achieve
for a computer. The reason for this is partly due to the complex grammatical structures and the
realistic world knowledge that humans have. It is also important to distinguish between the two
approaches of symbolic AI and connectionist AI.
Symbolic AI
● A subfield of AI
● Focuses on the processing and manipulation of symbols or concepts, rather than
numerical data.
● Intelligence is symbols and relations between them
Week 1................................................................................................................................. 2
Week 2................................................................................................................................. 6
Week 3................................................................................................................................. 9
Week 4............................................................................................................................... 12
Week 5............................................................................................................................... 15
Week 6............................................................................................................................... 19
Pseudocodes...................................................................................................................... 23
Recap and Q&A............................................................................................................... 25
Brightspace Practice Exam......................................................................................................... 27
This summary was created in block 2 of the academic year 2023-2024. I attended all the
lectures to make this summary, using the slides and transcribed what the professor said to make it
detailed enough to understand concepts like feedforward networks or optimal search.
I also added a chapter in which I copied/pasted all the pseudocodes that were used in the
lectures as that may come in handy during the exams. Lastly, some random notes of the last
lectures are added in the Recap and Q&A chapter, relating to parts that the professor mentioned
or questions that were asked at that moment.
At the end, I added the practice questions that were put on brightspace and what I believe
are the correct answers to these.
, Week 1
Introduction, History of AI
4 approaches to AI:
● Thinking humanly
● Thinking rationally
● Acting humanly
● Acting rationally
What is artificial intelligence and its purpose? For this, it’s important to look at cognitive
psychology first. Cognitive psychology is the study of the computations that make it possible to
perceive, reason, and act.
Artificial intelligence can be seen as a branch of philosophy, computer science, or more
and mostly draws from the disciplines of psychology and computer science to put emphasis on
the computation of perception. It studies how to build programs that enable computers to do what
the mind can do.
The problem of psychology is that it can be seen as an inverse problem; we analyze
behavior and reason retrospectively about the causes of it. However, such limitations are limited
and sometimes impossible to make. AI, on the other hand, uses forward modeling. This can be
seen as designing a simple system and seeing later how it behaves. Because we designed it, we
know how it works.
Forward modeling involves thinking algorithmically; using a set of rules that
unambiguously define a sequence of operations and hyper specific commands. In those cases, as
long as you follow the rules precisely, you can do it by hand. It thus matches input and output
through a series of processes. These can also be seen as functions, which are a list of instructions
that take input values (arguments) that apply some computations to create an output value.
Similarly, humans use input and output; respectively sensory perceptions and behavior.
, Descartes’ theory was defined as Cartesian dualism. This raised the question of how the
physical brain gives rise to the mental mind. On the other hand, materialists believed that all
mental states were caused by physical states. If we accepted materialism to be true, we could
assume that we could always recreate mental states from physical states. Throughout history, this
has brought many attempts to recreate systems of computations.
In the 1940s, McCulloch and Pitt worked with three principles; basic physiology,
propositional logic, and Turing’s theory of computation. From those times, it was believed that
any computable function could be computed by a network of artificial neurons. All logical
operators could be implemented by simple neural networks.
Turing believed that AI were weak, but intelligent if they were indistinguishable from
humans in conversations. Complex grammatical structures and realistic world knowledge would,
according to him, prove intelligence.
John Seaerle believes that true understanding requires the actual chemistry and physical
properties of the human brain. He came up with the Chinese room experiment. The Chinese
Room argument imagines a person inside a closed room in China. Someone on the outside could
ask questions and communicate with the person inside in Chinese. The person inside could use a
book to translate what is said to him. This would allow him to answer every possible question
that he is asked. This is basically what computers do. To an outsider, it may seem like something
inside understands Chinese, even if this is done through a book or by a computer. This is an
argument for Searle that computers do have true intelligence or sentience; they can only simulate
it through functions. Turin had an opposite point of view, as he believed that as long as
behavioral output is indistinguishable from a human, a machine is smart. Searle’s concept is also
known as weak AI. Strong AI is the concept that an appropriately programmed computer with
the right inputs and outputs would have a mind in the same sense as human beings have minds. It
thus suggests that an intelligent system can actually think.
It is important to distinguish between weak AI and strong AI. In the paper by Turing, he
introduced what was called the imitation game. This game implied that a machine is intelligent if
we cannot distinguish it from humans in a conversation. This is still quite impossible to achieve
for a computer. The reason for this is partly due to the complex grammatical structures and the
realistic world knowledge that humans have. It is also important to distinguish between the two
approaches of symbolic AI and connectionist AI.
Symbolic AI
● A subfield of AI
● Focuses on the processing and manipulation of symbols or concepts, rather than
numerical data.
● Intelligence is symbols and relations between them
