previous exam questions (for exam 1/3) Flashcards
taken from the pinned posts of the Discord
Consider the overall ”picture” of AI which will be represented as sets. In total we have a set for DL, ML and AI. Which of the following 6 statements are true?
- Machine Learning is the direct subset of AI
- Machine Learning is the direct subset of Deep Learning
- Deep Learning is the direct subset of Machine Learning
- Machine Learning is the direct superset of Deep Learning
- Deep Learning is the direct superset of AI
- AI is the direct superset of Deep Learning
1, 3 and 4 are correct
True or False: AI is about making computer based humans.
False
True or False: The Turing Test is about to beat humans in chess.
False
What is Machine Learning?
- a subsector of Artificial Intelligence
- Computer programs that automatically improve with experience
- makes use of algorithms and statistics to analyse and draw inferences from patterns in data
When can a machine be considered as intelligent?
When it can pass the Turing Test
What is true for an algorithm?
- is a step by step procedure
- different algorithms can lead to the same result
- a good algorithm solves the task in less time
What is the Turing test about? Choose the correct statements.
- If the judge has successfully passed the test, it can be considered intelligent.
- There is a human and a computer that pretends to be a human.
- There is a woman and a man who pretends to be a woman.
- The test is about defining a long and controversial list of necessary prerequisites for intelligence.
- The judge has to find out who the human is.
- The result should be that the computer is considered to be intelligent.
2, 5 and 6 are correct
Who showed how to write logic in the form of analytical equations?
George Boole
Which British mathematician is often credited as being the key founder of AI?
Alan Turing
Who defined the first chess program in Plankalkul?
Konrad Zuse
Map the following important contributions in the history of Artificial Intelligence to their associated names.
- showed how to write logic in the form of analytical equations.
- designed at around 1495 a humanoid mechanical knight.
- was the first to recognise that the machine had applications beyond pure calculation, and to have published the first algorithm intended to be carried out by such a machine.
- defined the first chess program in Plankalkü.
- George Boole
- Leonardo daVinci
- Ada Lovelace
- Konrad Zuse
True or False: The first ideas about AI systems where in the 20th century.
False
What was the first mythical automaton with artificial intelligence?
Talos (built by Hephaistos, killed by Jason of the argonauts)
Who said: ”all B’s are A, All C’s are B therfore all C’s are A’s”
Aristotle
When was the ”Birth” of AI?
1956
Which of the following disciplines are main contributers to AI?
- Mathematics
- Logic
- Psychology
- Neuroscience
- Information Technology
- Social Sciences
2, 3, 4 are correct
True or False: Alan Turing coined the term ”Artificial Intelligence” in 1956 at the Dartmouth Conference.
False.
He wasn’t there, but the term was coined at that conference.
Which of the following were present at the Dartmouth conference and then dominated the field for the next 20 years?
- John McCarthy
- George Boole
- Herbert Simon
- Frank Rosenblatt
- Alan Turing
- Allan Newell
1, 3 and 6 are correct
Map the following statements to the correct person.
- is one of the founding fathers of AI.
- has invented a Lingua Charististica.
- has invented the first digital computer.
- has published ”The Mathematical Analysis of Logic”
- Marvin Minsky
- Leibniz
- Konrad Zuse
- George Boole
Name the following calculation machines based on their capabilities.
- could calculate mathematical tables using the method of finite differences.
- could perform the four basic arithmetic operations.
- could perform addition
- could pe programmed to the type of operation to perform.
- Babbage’s Difference Engine
- Leibnitz’ Reckoning Machine
- Pascaline
- Babbage’s Analytical Engine
Why brought Deep Learning unexpected breakthroughs in diverse areas? Choose the correct answer/s.
- It was the first successfull knowledge-intensive system.
- It came up with fundamentally new ideas and techniques.
- Smaller models and less training data were needed.
- Much more training data, bigger models and computational power were available.
4 is correct
Which of the following machines are NOT covered in the ancienct history of artificial intelligence?
- Talos the giant intelligent bronze robot
- James Watt’s steam engine
- The chess playing turk
- clay Golem from Prague
James Watt’s steam engine
Match the following years correctly with the associated definitions!
- Z3, the first digital computer
- Dartmouth Conference
- Eniac, the first Turing complete digital computer
- first neural network computer
- 1941
- 1956
- 1945
- 1951
Why is ”local search” not the optimal strategy for sudoku solving?
no guarantee for completeness
True or False: In search algorithms, a heuristic often denotes a function that estimates the quality of a given state?
True
True or False: State representation describes the current state of the solving process.
True
What is NOT true about Heuristics?
- knowledge that is helpful for solving a problem
- can also go wrong
- guarantees a result that satisfies our problem
- mostly a function that estimates the quality of a state
3 is not true
What is the main problem with Hill-Climbing search?
maybe reaches a local optima
Match the following optimizations for the Hill Climbing Search:
- Stochastic Hill Climbing
- Beam Search
- Simulated Annealing Search
- Random Restart Hill Climbing
a. selects the successor nodes randomly but the nodes are weight
b. makes several iterations with different starting positions
c. combination of hill climbing and random walk
d. keeps track of k-states rather than just one
1a
2d
3c
4b
What is one principle of solving problems like the towers of hanoi?
Divide-and-Conquer
True or False: Simple Exhaustive Search uses either heuristics or constraints.
False
What is one negative characteristic of depth-first search?
- depth-first search never finds the worst solution
- depth-first search could be exponentional
- depth-first search can never finds the perfect solution
- depth-first search is deprecated
2 is correct
Map the following contributions in the history of Al to their originators.
- is one of the participants in the Dartmouth workshop in 1956
- has built the first digital computer
- has proposed a mathematical calculus for formal logic
- has proposed a first formal language for representing scientific knowledge
a. Gottfried Wilhelm Leibniz
b. Marvin Minsky
c. George Boole
d. Konrad Zuse
1b
2d
3c
4a
In a game, each side can make 4 possible moves in each position. If a machine makes a complete look-ahead search for 3 plies (half moves). How many possible leaf positions will it consider? Enter a number as an answer.
4 to the power of 3 = 64
Which of the following techniques have we considered to be particularly useful for different steps when solving a Sudoku puzzle?
- check which value can be filled in a selected square
- try out one of several possible values and possibly come back later if it doesn’t work out
- use the value that has been filled in for reducing the set of possible values in neighboring squares
- selecting the next square that could be filled in
a. constraint satisfaction
b. heuristics
c. constraint propagation
d. backtracking search
1a
2d
3c
4b
True or False: Machine Learning is an essential part of every AI system because an agent that does not learn cannot be considered to be an AI.
False
True or False: At the heart of genetic algorithms is an operator called “cross-over” which combines parts of two solutions into a new solution.
True
Which of the following general heuristics for constraint solving does NOT aim at selecting a variable?
- Degree Heuristics
- Minimum Remaining Values Heuristic
- Least Constraining Value Heuristic
- Min-Conflicts Heuristic
3 is not aiming at selecting a variable
Map the following important contributions in the history of AI to their associated names.
- suggested an algebra for logic
- defined general patterns for logical reasoning
- proved that any logical system powerful enough to formalize mathematics is either incomplete or inconsistent
- constructed the first programmable computer
a. Kurt Gödel
b. Aristotle
c. Edward Feigenbaum
d. George Boole
e. Charles Babbage
f. Ada Lovelace
1d
2b
3a
4e
True or False: Deep Learning is a sub-discipline of Machine Learning.
True
Which of the following statements is false?
- An AI beat the best chess player in 2017.
- Until the present day, an AI hasn’t been able to beat the best GO player.
- Until the present day, an AI hasn’t been able to beat the best chess player.
- An AI beat the best GO player in 2017.
1, 2 and 3 are false.
True or False: According to John McCarthy AI has to confine itself to biologically observable methods since it is related to using computers to understand human intelligence.
False
The “Imitation Game” by Alan Turing focusses on which one of the following definitions of AI?
- system that acts rationally
- system that acts like humans
- system that thinks rationally
- system that thinks like humans
2 is correct
Which programming language did John McCarthy invent?
- Java
- C++
- Python
- Lisp
4 is correct
Match the following names of contributors to the history of AI to their associated work.
- Ada Lovelace
- Claude Shannon
- Konrad Zuse
- Hephaistos
a. first digital computer
b. algorithm for playing chess
c. crafting Talos and other artifical devices
d. programming for Analytical Engine
1d
2b
3a
4c
What is the difference (from the perspective of AI research) between rational thinking and high-level human-like thinking (i.e. that researched by symbolic AI field)?
- Human thinking often makes intuitive and not entirely explainable decisions, which
rational thinking based entirely on rules of logic cannot comprehend - The main goal of think-like-human machines is to pass the Turing test. Rationaly thinking machines maximize predefined utility in given enviroment instead.
- Rationally thinking AIs usually need precise logical reasoning and knowledge representation while human-like model neurons and their nets.
- There is no distinct difference.
- and 3. are correct
True or False: Deep Learning has been particularly successful in pattern recognition.
True
Which three of the following people were 2019 Turing Awardees?
- Yoshua Bengio
- Geoffrey Hinton
- Yann LeCun
- Tim Berners-Lee
- Alan Kay
- Jack Dongarra
1, 2 an 3 are correct
Map the following important contributions to the topic of what Artificial Intelligence is.
- Described what Machine Learning is
- The Turing Award was named after
him - One of the first people who defined
what Artificial Intelligence is. - Was the first person who formalized
algorithms
a. Tom Mitchell
b. Al-Chwarizmi
c. John McCarthy
d. Alan Turing
1a
2d
3c
4b
What is NOT a problem of Systems that think rationally?
1. Not all intelligence is mediated by logic behavior
2. Practical constraints
3. to define comprehensible commands
4. It is hard to formalize knowledge exactly
3 is not a problem of systems that think rationally
True of False: The basic idea of the Turing Test is to compare computers with undeniably intelligent beings (humans), instead of defining a long and controversial list of necessary prerequisites for intelligence.
True
Which statements about the ”Turing test” are correct?
- Requires the cooperation of the most important components of AI: Knowledge, reasoning, language comprehension, learning,…
- The Turing Test is not suitable for mathematical analysis.
- The idea came from Fred Lois Ture (1914-1972).
- The Turing Test is reproducible.
- Definitions of necessary prerequisites for intelligence are necessary.
- The imitation game is an operational test of intelligence for computing machinery
1, 2 and 6 are correct.
True or False: Deep Learning is a form of machine learning that makes use of artificial neural networks.
True
True or False: Machine Learning is the science ”concerned with the question of how to construct computer programs that automatically improve with experience”. (Tom Mitchell)
True
Which statements about ”Algorithms” are correct?
- An algorithm is a step-by-step procedure for finding a solution to a class of problems
- Algorithms do not terminate in finite time.
- Algorithms are not deteministic
- Algorithms are deterministic
- An algorithm is an intuitiv procedure for finding a solution to a class of problems.
- Algorithms terminate in finite time.
1, 4 and 6 are correct
True or False: The Turing Test is meant to evaluate if a model can think rationally.
False
True or False: An algorithm has only two main characteristics. These are that it is deterministic and that it terminates in finite time.
False
Map the concepts with their definitions:
- ”It is the science and engineering of
making intelligent machines … and
does not have to confine itself to me-
thods that are biologically observa-
ble.” -John McCarthy, Stanford(1956) - It is a field that bring together com-
putational models from AI and experi-
mental techniques from psychology to
consutruct precise and testable theo-
ries of the mind. - It is the science ”concerned with the
question of how to construct compu-
ter programs that automatically im-
prove with experience.” -Tom Mitchell
(1997) CMU - A form of machine learning that makes
use of artificial neural networks
a. Machine Learning
b. Deep Learning
c. Cognitive Science
d. Artificial Intelligence
1d
2c
3a
4b
Map the inventors to their invents.
1. Rabbi Loew
2. Kempelen
3. Hephaestos
4. Leonardo da Vinci
a. The Golem of clay
b. Talos, a giant intelligent bronze robot
c. The Chess-Playing Turk
d. humanoid mechanical knight
1a
2c
3b
4d
Which machine/automat was NOT built in the 19th century?
1. Babbage’s Difference Engine
2. Leibnitz’ reckoning machine
3. Euphonia (Joseph Faber’s Amazing Talking Machine)
4. The Automaton Trumpeteer
2 was not built in the 19th century
True or False: Ada Lovelace was the first computer programmer in 1840s
True
Hephaisto built artificial…
- tripods
- brazen heads
- waiters
- talking machines (Euphonia)
- bellows
- humanoid mechanical knights
1, 3 and 5 are correct
True or False: Rene Descartes (1596-1650) proposed that animals are machines, because their bodies (heart, lungs, muscles, etc.) can be explained by physics. He also thought that the soul made the difference between humans and animals.
True
Who showed how to write logic in the form of analytical equations?
1 Reverend George Boole
2 Aristotle
3 Joseph Faber
4 Leibnitz
1
Complete the sentences on the subject of robots.
- Talos, a giant intellligent bronce robot was built by …
- The origin of the word ”Robot” is from …
- The Three (later four) Laws of Robotics are from …
- The autonomous robots, that look like turtles, was constructed by …
a. Grey Walter.
b. Hephaisto.
c. science fiction writer (e.g. I Robot) Isaac Asimov.
d. author Karel Capek who produced the play R.U.R. (Rossum’s Universal Robots).
1b
2d
3c
4a
True or False: Kurt Gödel’s paper, ”On Formally Undecidable Propositions”from 1931, in which he proved that it is in priniciple not possible that a sufficiently complex axiomatization is both complete and consistent, is a setback for logic reasoning.
True
For Artificial Neural Networks found Frank Rosenblatt (1928-1971) some Inspiration from the brain:
- learning happens by changing the strength of connections (synapses)
- For learning it is necessary that the strenght of connections (synapses) will be the same.
- many small discontiguous units (neurons)
- behavior of the whole is is equal to the sum of the parts
- behavior of the whole is more than the sum of the parts
- many small interconnected units (neurons
1, 5 and 6 are correct
About the history of the first chess programs. Assign correctly!
- chess algorithm computed with paper and pencil; first recorded man-machine (chess) game
- defined the first chess program
- algorithm for playing chess foresees strategies that are still used today. (Type-A Strategy; Type-B Strategy)
- the first lost from a Chess world champion versus a chess programm (Deep Blue)
a. 1950; Claude Shannon
b. 1951; Alan Turing
c. 1997; Garry Kasparov
d. 1942-45; Konrad Zuse
1b
2d
3a
4c
True or False: Simon and Newell’s all predictions about AI (1958) came more or less true within 10 years.
False
Progress of AI in the beginning was slower than (unrealistic) expectations. What was NOT the reason?
- Fundamental limitations on techniques and representations
- Lack of scalability
- Difficulty of knowledge representation
- Creation of Lisp (second-oldest high-level language)
4 was not a reason
Deep Learning brought unexpected and unprecedented breakthroughs in diverse areas such as computer vision, language understanding, game playing, etc. Why did this start only a few years ago?
- We have the computationalpower for doing training deep networks
- Readily available programming libraries make it much easier
- The models are smaller
- There was too little demand for such programmes
- Introducing fundamentally new ideas was neccessary
- We have much more data for training deep networks
1, 2 and 6 are correct
True or False: Konrad Zuse’s Z3 was the first digital computer.
True
A heuristic is …?
- a function describing the position of the search space
- a function estimating the quality of a state
- an (e.g. exponential) function describing the complexity (e.g. O(n2)) of a search algo-
rithm - a binary function checking whether or not the problem is solved
2 is correct
What are operations done between generations in genetic algorithms?
1 Divide and Conquer
2 Cross Over
3 Mutation
4 Exhaustive search
5 Selection
6 Local Search
2, 3 and 5 are correct
Which steps belong to which algorithms?
- try to find solutions by looking at near
elements - applying random changes
- trying to improve the current state
step by step using an heuristic - combine multiple partial solutions into
one
a. Local Search
b. Greedy Hill-Climbing Search
c. Divide and Conquer
d. Genetic algorithms
1a
2d
3b
4c
Which of the following statements about stochastic search are true?
- It usually fails completely when it encounters local maximum.
- It is non-deterministic.
- It is sometimes called greedy search.
- It is inspired by evolution of biological species.
- It chooses its steps with probability corresponding to their attractivity.
- It belongs to heuristic algorithms.
2 and 5 are correct
Map phases of genetic algorithms to their description.
- Fitness
- Cross-over
- Mutation
- Selection
a. Evaluate usefulness of generation states.
b. Perform small and random changes of the newly generated states.
c. Create new states by combining selected pairs.
d. Select states to reproduce with probability corresponding to their fitness.
1a
2c
3b
4d
There are many specialized tricks to solve a Sudoku, but the most follow a general pattern. Match the sentences correctly.
- check all possibilities for the selected square(s)
- propagate the information to other squares
- select a square (or a set of squares)
- repeat these steps until the solution is complete
a. Constraint Propagation
b. Constraint Satisfaction
c. Search
d. Heuristics
1b
2a
3d
4c
For defining a search problem, we need
1 - A state description and a goal description
2 - A state function and a goal function
3 - A state function and a goal description
4 - A state description and a goal function
4
True or False: A state description is an abstract representation of the current state of a puzzle. The set of all such descriptions is called the state space.
True
Which statements about ”Local Search” are true?
1 - The Approach is to keep a single ”current” state (or a fixed number of them) and to try to improve it by maximizing a heuristic evaluation.
2 - It uses a lot of memory.
3 - It needs much time to find solutions in large or infinite state spaces.
4 - It guarantees completeness (that a solution is always found) and optimality (that the best/shortest solution is found).
5 - They often quickly finds solutions in large or infinite state spaces.
6 - There are no guarantees for completeness (that a solution is always found) or optimality (that the best/shortest solution is found).
1, 5 and 6 are correct
True or False: Choosing a good state representation may already be crucial for finding a good solution.
True
Which statements about ”Heuristic” are true?
1 - In search algorithms, a heuristic always denotes a function that compute the quality of a given state.
2 - It denotes a rule and it should to be note that heuristics may never go wrong!
3 - In search algorithms, a heuristic often denotes a function that estimates the quality of a given state.
4 - The word is derived from the russian word ”heurestika” and translates as ”always valid”.
5 - The word is derived from the greek word ”heurisko” and translates as ”I find”.
6 - It informally denotes a ”rule of thumb“ and it should to be note that heuristics may also go wrong!
3, 5 and 6 are correct
True or False: The Algorithm from the Beam Search is…
- Start with randomly generated states.
- At each iteration, all the successors of the current state are generated.
- Select the k best successors from the complete list and repeat.
True
Match the statements to the correct area of Hill-climbing- search
- States may be refined in multiple ways; → similarity along various dimensions
- select the successor node ramdomly; better nodes have a higher probability of being selected
- Different initial positions result in different local optima; → make several iterations with different starting positions
- location: states; elevation: heuristic value (objective function); Assumption: states have some sort of (linear) order; continuity regarding small state changes
a. State Space Landscape
b. Stochastic Hill-Climbing
c. Random Restart Hill-Climbing
d. Multi-Dimensional State-Landscape
1d
2b
3c
4a
If the algorithm will stop as soon as is at the top of a hill, this will e named ”Local Optima” and
is a main problem. Of which Search variant?
1 Simulated Annealing Search
2 Beam Search
3 Every of the three named Search variants.
4 Hill-Climbing Search
4
True or False: Information from different beams is combined so the Beam search is different from k parallel hill-climbing searches, but the effectiveness suffers from lack of diversity of the k states.
True
Simulated Annealing Search is …
1 . . . modelled after mutation of DNA and comperable to a stochastic hill-climbing step.
2 . . . a combination of hill-climbing and random walk.
3 . . . only used to find the right temperature in metallurgy.
4 . . . the same idea as in Stochastic Beam Search but uses ”sexual” reproduction.
2
True or False: In the Genetic Algorithms there will be the next generation produced by selection, cross-over, and mutation.
True
True or False: The Genetic Programming is popularized by David Fogel.
True
”Divide-and-Conquer” is a general problem solving strategy with many applications. Match the sentences correctly!
- Decompose the problem into simpler partial problems
- Solve each of these simpler problems by possibly solving them again by division
- Solve each of these simpler problems
- Combine the partial solutions to a …
a. recursive programming
b. DIVIDE
c. CONQUER
d. complete solution
1b
2a
3c
4d
What can happen, if the analysis of the solutions of a recursive solution rise to a simple non-recursive solution?
1 - Much less memory is needed and the length of the solution is greatly reduced.
2 - Much less memory is needed, but the solution length cannot be reduced.
3 - Nothing changes.
4 - The same memory is needed, but the length of the solution is reduced.
2
Which of the following techniques have we considered to be particularly useful for different steps when solving a Sudoku puzzle?
- check which values can be filled in a selected square
- use the value that has been filled in for reducing the set of possible values in neighboring squares
- selecting the next square that could be filled in
- try out one of several possible values and possibly come back later if it doesn’t work out
a. heuristics
b. backtracking search
c. constraint propagation
d. constraint satisfaction
1d
2c
3a
4b
Which three of the following people did write early chess programs?
- John McCarthy
- Arthur Samuel
- Claude Shannon
- Konrad Zuse
- Marvin Minsky
- Alan Turing
3, 4 and 6 are correct
True or False: Minimax search can, in principle, solve any deterministic 2-person zero-sum perfect information game.
True
True or False: the constraint graph can be used to identify to which variables constraints have to propagate.
True
True or False: Beam search is the same as k parallel hill-climbing searches.
False
Choose what those terms refer to:
- Method to check whether a state description satisfies the goal:
- The abstract representation of the current state of a puzzle:
- Trying a potential solution and revising it later:
- The set of all possible descriptions:
a. backtracking
b. state space
c. state description
d. goal function
1d
2c
3a
4b