COMP 3190
1. What makes something Intelligent?: respond flexibly to environment, learn from previous
situations, achieve goals, communicate or share ideas, creativity, self awareness.
2. What is the Turing Test?: Edwardian party game, human and machine are connected to judge, both
try to convince judge that they are human.
3. What is a reverse turing test?: CAPTCHA (completely automated public turing test to tell computers
and humans apart)
4. What is Searle and the Chinese Room?: A room where chinese characters are pushed in one
side, and english response is pushed out the other side. Searle argues that this requires no understanding, and
therefore no deep intelligence.
5. Goals of AI: Create intelligent artificial entities;
Create systems and approaches that would generally be considered intelligent if those problems were solved by
humans;
Find ways of making computers more effective at solving problems, understanding of nature of intelligence;
Find practical solutions to hard problems, find solutions to NP-Complete problems.
Function well despite the complexity!
6. What is NP-Complete?: nondeterministic polynomial-time complete, solutions can be verified in a
straightforward way.
7. What does it mean for a system to be rational?: do the right thing in any situation it finds
itself in, subject to the limits of knowledge.
8. What are the 4 dependencies of rationality?: Performance measure used;
What an entity knows about the immediate world around it;
Non-immediate knowledge of the world;
Repertoire of actions we have available
9. What are the two levels of intelligent processing?: 1. Reasoning consciously
2. Reasoning subconsciously
(recognizing immediately, defined at the 100ms mark). Referred to as Symbolic and Sub symbolic levels.
10. Physical symbol system hypothesis: a physical symbol system has the necessary and sufficient
means for general intelligent action.
11. How is subsymbolic AI represented? what problems are there with this
approach?: Nodes and links, difficult to maintain at high levels
1/9
, COMP 3190
12. What 3 things does conventional computing have?: Objects to work on;
Operations to perform;
Control strategy (solve the problem yourself, convert that into a algorithm.)
13. Where does conventional computing fail?: Impossible to come up with conventional algo-
rithm that would cover a previously unseen case.
14. What is declaritive programming?: Let it develop its own algorithm based on the tools and
objects we provide. Intelligence is shown by avoiding lengthy search. Solution for problems is developed dynamically
by the system
15. Which is faster: Declaritive or conventional programming?: Conventional, but that's
because we do all the thinking ahead of time!
16. what is a scaling problem in regards to generality?: Solutions for small scale problems
(weak technique) will fail as problems become more complex.
17. What are the 3 principles of symbolic AI?: Representing a problem;
Searching through solution space;
Avoiding search through appropriate application of symbolic knowledge
18. what is a well formed formula?: A valid logical sentence.
19. What is planning?: Consider possibilities systematically until it comes up with a sequence that lets it do
what is needed to achieve its goal.
20. What are reactive agents?: Agents that react to their environment purely, no ability to remember
previous situations or learn. (creatures like crickets), good enough for many environments
21. What are reactive agents with state?: Reactive agents that can do some learning, they can recall
what it has seen before and use that for its choice. Insect level intelligence.
22. What is a problem state? What is a goal state? What is a plan?: Problems
configuration at a specific point in time;
Desired problem state, not necessarily end state;
Sequence of operations that takes us from initial state to the goal state.
23. What is a problem space?: abstract representation of the states associated with a particular problem,
must be generated as we explore, stored or maintained so we can search it as we generate.
24. What are the benefits of depth-first search? Disadvantages?: Good: minimal
storage space, O(bd) (branching factor, depth), useful if tree is pretty uniform and there are more than one paths to
goal state.
2/9
1. What makes something Intelligent?: respond flexibly to environment, learn from previous
situations, achieve goals, communicate or share ideas, creativity, self awareness.
2. What is the Turing Test?: Edwardian party game, human and machine are connected to judge, both
try to convince judge that they are human.
3. What is a reverse turing test?: CAPTCHA (completely automated public turing test to tell computers
and humans apart)
4. What is Searle and the Chinese Room?: A room where chinese characters are pushed in one
side, and english response is pushed out the other side. Searle argues that this requires no understanding, and
therefore no deep intelligence.
5. Goals of AI: Create intelligent artificial entities;
Create systems and approaches that would generally be considered intelligent if those problems were solved by
humans;
Find ways of making computers more effective at solving problems, understanding of nature of intelligence;
Find practical solutions to hard problems, find solutions to NP-Complete problems.
Function well despite the complexity!
6. What is NP-Complete?: nondeterministic polynomial-time complete, solutions can be verified in a
straightforward way.
7. What does it mean for a system to be rational?: do the right thing in any situation it finds
itself in, subject to the limits of knowledge.
8. What are the 4 dependencies of rationality?: Performance measure used;
What an entity knows about the immediate world around it;
Non-immediate knowledge of the world;
Repertoire of actions we have available
9. What are the two levels of intelligent processing?: 1. Reasoning consciously
2. Reasoning subconsciously
(recognizing immediately, defined at the 100ms mark). Referred to as Symbolic and Sub symbolic levels.
10. Physical symbol system hypothesis: a physical symbol system has the necessary and sufficient
means for general intelligent action.
11. How is subsymbolic AI represented? what problems are there with this
approach?: Nodes and links, difficult to maintain at high levels
1/9
, COMP 3190
12. What 3 things does conventional computing have?: Objects to work on;
Operations to perform;
Control strategy (solve the problem yourself, convert that into a algorithm.)
13. Where does conventional computing fail?: Impossible to come up with conventional algo-
rithm that would cover a previously unseen case.
14. What is declaritive programming?: Let it develop its own algorithm based on the tools and
objects we provide. Intelligence is shown by avoiding lengthy search. Solution for problems is developed dynamically
by the system
15. Which is faster: Declaritive or conventional programming?: Conventional, but that's
because we do all the thinking ahead of time!
16. what is a scaling problem in regards to generality?: Solutions for small scale problems
(weak technique) will fail as problems become more complex.
17. What are the 3 principles of symbolic AI?: Representing a problem;
Searching through solution space;
Avoiding search through appropriate application of symbolic knowledge
18. what is a well formed formula?: A valid logical sentence.
19. What is planning?: Consider possibilities systematically until it comes up with a sequence that lets it do
what is needed to achieve its goal.
20. What are reactive agents?: Agents that react to their environment purely, no ability to remember
previous situations or learn. (creatures like crickets), good enough for many environments
21. What are reactive agents with state?: Reactive agents that can do some learning, they can recall
what it has seen before and use that for its choice. Insect level intelligence.
22. What is a problem state? What is a goal state? What is a plan?: Problems
configuration at a specific point in time;
Desired problem state, not necessarily end state;
Sequence of operations that takes us from initial state to the goal state.
23. What is a problem space?: abstract representation of the states associated with a particular problem,
must be generated as we explore, stored or maintained so we can search it as we generate.
24. What are the benefits of depth-first search? Disadvantages?: Good: minimal
storage space, O(bd) (branching factor, depth), useful if tree is pretty uniform and there are more than one paths to
goal state.
2/9
