Dit document bevat een uitgebreide samenvatting van alle colleges van het vak Artificial Intelligence. Moeilijke begrippen zijn uitgelegd en plaatjes/figuren zijn toegevoegd om de uitleg te ondersteunen.
What is AI:
- The branch of computer science that studies how to build or program computers to enable them to
do what minds can do
* = Making computations that make it possible to perceive, reason and act.
- AI puts greater emphasis on:
* Computations (more than psychology)
* Perception, reasoning and action (more than computer science)
The inverse problem:
- Psychology: uses a set of observations → try to infer the processes that produce the observations
we see.
* Problem: these inferences are limited & sometimes impossible to make
- AI: uses forward modelling = designing a (simple) system, and see how it behaves
* This is where AI and computational psychology meet
History of AI:
Weak vs strong Symbolic AI
1e neural
network
computer
Coined the term AI
Weak vs. Strong AI:
- Weak AI: No matter how intelligent AI seems, it is not true intelligence. There are no mental states.
* Turing test: Imitation game → a machine is intelligent if we cannot distinguish it from a
human in conversation.
* Searle: only brains cause minds. Chinese room argument:
- All questions in Chinese can be answered (by a person who does not speek Chinese)
in Chinese because of a big book that contains every possible question.
- Rule-based manipulation of symbols does not constitute intelligence: the inhabitant
of the Chinese room does not understand Chinese.
- Strong AI: Believe that intelligent systems can actually think → “The appropriately programmed
computer with the right inputs and outputs would thereby have a mind in exactly the same sense
human beings have minds.”
* Assumption: the human mind is an information processing system, and that thinking is a
form of computing.
* Problems: Does an accurately enough simulated human mind have all the same properties
as an actual human mind?
,Symbolic vs. Connectionist AI:
- Symbolic AI: Human thinking is a kind of symbol manipulation (IF (A > B) AND (B > C) THEN (A > C).
* Intelligence is thought of as symbols and the relation between them
* ELIZA: an early natural language processor. Purpose: mimic a psychotherapist. Response
public: computers can have conversations.
- How does it work: ELIZA looks for keywords → uses database of rules to construct
new sentences using these keywords. No keywords?, then “I see, please go on”.
* STRIPS: an automated planner. Realization of goals by dividing task in subgoals.
- Problem: Sussman anomaly → Sometimes divided tasks do not work in order to
accomplish end goal.
* Expert system MYCIN: emulates the decision-making ability of a human expert.
- Simple if-then rules with certainty factors
- Better accuracy than physicians, never used due to ethical and legal difficulties
* Symbolic AI does not suffice
* Criticism: it is unclear how processes like pattern recognition would work in a purely
symbolic way. Symbolic representation doesn’t seem necessary for many behaviors.
- Connectionist AI: artificial neural networks.
* Early PDP work: Model of human memory = memory is not stored in neurons, but in the
connections between them (excitatory & inhibitory connections) & human memory is content-
addressable.
* Biologically inspired: computation is massively parallel, this is efficient.
* Lesion tolerant
* Capable of generalization: ANNs are capable of learning, and are able to generalize rules to
novel input
* Principles:
- Neurons output a signal based on their input signal.
- Mental states are represented as N-dimensional vectors of numeric activation
values over neural network units.
- Memory is created by modifying the connection strength (weight) between units.
* Solve complex, non-linear or chaotic classification problems
* No a priori assumption about problem space or statistical distribution
* ANNs can compute anu computable function
* Pattern recognition
Deep networks: adding more layers to the ANN → adds to dimensionality of classification
Machine learning:
- Supervised learning: External knowledgeable supervisor present the system with correctly labelled
training data
- Unsupervised learning: Discover hidden structure in data without labelled data
- Reinforcement learning: Learning from a feedback signal
- Classification: Determining group membership based on input data & then comparing to rules to
assess the object to a classification.
- Regression: Predict outcome data based on input data
, Algorithms = A set of rules that unambiguously defines a sequence of operations
Expert systems: Represent an expert’s understanding of a subject
- Consists out of 2 main components:
* The knowledge base = represents facts about the world and how concepts are related.
- Often using predicate logic (leukaemia is a disease). Also includes more complex
relationships (leukaemia is an abnormality of blood)
* Inference engine = how to reason with the knowledge base. We need to be able to
manipulate the symbols. There are facts and rules → from that we draw conclusions. Use the
predicate logic to connect symbols.
- Complex rules: more control over which rule to apply is necessary.
* 3 ways to control the use of knowledge when solving problems using a knowledge base:
1. Forward chaining = Works sequentially from given statements to a deduced
conclusion. E.g. robot can perceive basic features, has 15 rules in its knowledge base, robot will draw
a conclusion when all antecedents for a certain rule are satisfied. (slide 20).
2. Backward chaining = Starts at the opposite end of the logical process. Starts by
forming a (random of pre-informed) hypothesis and using if-then rules to work backward toward
hypothesis-supporting assertions. E.g. robot can perceive basic features, has 15 rules in its
knowledge bas, robot needs to check whether the hypothesis is true by checking if all the
antecedents off a certain rule are satisfied. If this is the case, than the hypothesis is accepted.
3. Control knowledge.
* How decide which chaining direction to use?
- Backward chaining: helpful when not all facts are known yet
- Forward chaining: helpful when you want to know everything you can form a set of
facts.
Sudoku solving → 2 algorithms:
- Naïve brute force: fill in random digits and then check if the solution is a valid Sudoku grid. If it isn’t
we try again.
* Problem: combinatorial explosion
- Backtracking: works by iteratively generating possible candidate solutions. When it determines that
the candidate solution cannot lead to a complete solution, it abandons the candidate solution. It
then goes back (backtracks) to the previous choice point and creates a new candidate solution.
* When reaching a dead end you take 1 step back and try again with a new solution.
* Assuming a solution exists, the backtracking algorithm guarantees a valid result.
* For an average Sudoku, backtracking takes about 38.000 iterations.
Basic and optimal search:
- Search = the process of looking for a sequence of actions that reaches the goal.
* Input is a problem
* Output is a solution in the form of an action sequence
- Basic/uninformed/blind search = when strategies have no additional information
about states beyond that provided in the problem definition.
- Basic search: simple example is there a path from A to B?
* We can look at all possible paths and display them in a search tree. But again:
combinatorial explosion. So we want a search method that minimizes the number of examined
paths.
Voordelen van het kopen van samenvattingen bij Stuvia op een rij:
Verzekerd van kwaliteit door reviews
Stuvia-klanten hebben meer dan 700.000 samenvattingen beoordeeld. Zo weet je zeker dat je de beste documenten koopt!
Snel en makkelijk kopen
Je betaalt supersnel en eenmalig met iDeal, creditcard of Stuvia-tegoed voor de samenvatting. Zonder lidmaatschap.
Focus op de essentie
Samenvattingen worden geschreven voor en door anderen. Daarom zijn de samenvattingen altijd betrouwbaar en actueel. Zo kom je snel tot de kern!
Veelgestelde vragen
Wat krijg ik als ik dit document koop?
Je krijgt een PDF, die direct beschikbaar is na je aankoop. Het gekochte document is altijd, overal en oneindig toegankelijk via je profiel.
Tevredenheidsgarantie: hoe werkt dat?
Onze tevredenheidsgarantie zorgt ervoor dat je altijd een studiedocument vindt dat goed bij je past. Je vult een formulier in en onze klantenservice regelt de rest.
Van wie koop ik deze samenvatting?
Stuvia is een marktplaats, je koop dit document dus niet van ons, maar van verkoper Millyy. Stuvia faciliteert de betaling aan de verkoper.
Zit ik meteen vast aan een abonnement?
Nee, je koopt alleen deze samenvatting voor €5,89. Je zit daarna nergens aan vast.
