Introduction to Artificial Intelligence Questions and Answers

Q: Which environment is fully observable, deterministic, static and discrete?

Chess without a clock presents the whole board to both players, has fixed rules, waits for the player to move and has finitely many positions, which is why it was studied so early and so successfully. Driving fails every criterion, poker is partially observable and stochastic, and a vacuum cleaner is partially observable and continuous. Classifying a familiar task against the dimensions is the quickest way to fix them in mind.

Q: What was later seen as the limitation of chess as a research problem?

Programs that reached world championship strength did so with search and evaluation techniques that transferred poorly to perception, language and physical action, so the milestone proved less significant than had been hoped. The rules were easy, the milestone was reached, and hardware sufficed. A benchmark whose solution does not generalise is a recurring disappointment in this field, and it is why benchmark choice attracts so much scrutiny.

Q: What is the significance of a system defeating a human champion at a game?

Such results are genuine achievements within their domain but narrow, since the system cannot transfer its competence to anything else. Interpreting them as evidence of general intelligence is the commonest error in reporting on the field. Nor do they establish any similarity of mechanism, and most such games remain unsolved in the game theoretic sense. Reading a result for exactly what it demonstrates, and no more, is the discipline that discussion of this subject most requires.

Q: What is meant by an agent's autonomy?

An agent is autonomous to the degree that its behaviour is determined by what it has learned rather than entirely by what its designer built in, which matters because a purely preprogrammed agent fails whenever the designer's assumptions do. Endurance and unsupervised operation are practical measures often loosely called autonomy but the technical sense concerns the origin of the behaviour. Distinguishing designed from learned competence explains why learning is treated as an architectural addition rather than an application.

Q: Why should an agent have some autonomy in the technical sense?

Any fixed set of rules embodies assumptions that will eventually be violated, and an agent able to learn from its own experience can adapt where a preprogrammed one simply fails. Autonomy costs more rather than less to achieve, regulation generally pushes the other way, and a performance measure is more rather than less necessary for a learning agent. Building in the capacity to cope with what was not foreseen is what distinguishes a robust system from a merely complete one.

Q: What does the phrase acting humanly describe as a goal for the field?

Acting humanly is the Turing test programme, judged entirely by whether the behaviour passes for human, while thinking humanly is the cognitive modelling programme judged by whether the process matches. Deference and interface design are separate concerns. Keeping the four quadrants apart matters because each implies a different standard of evidence: a psychological experiment for one, a behavioural test for another and a mathematical argument for the rational quadrants.

Q: What does thinking rationally describe as a goal for the field?

The thinking rationally quadrant is the logicist programme, in which correct inference from true premises is the criterion, and it gave the field its logic and theorem proving traditions. Acting rationally, the fourth quadrant, is about choosing actions well and is broader, since acting well sometimes requires acting before a proof is available. Human agreement and speed are separate criteria. The gap between correct inference and effective action is exactly why the field moved from the third quadrant to the fourth.

Q: Why did the field move from thinking rationally to acting rationally?

Proving the right thing to do may be intractable, and in many situations no proof is available at all, yet an agent must still act, so the criterion shifted to doing the best it can with what it knows. Probability and decision theory replaced logic as the central formalism for uncertain domains. Logic is consistent, computational cost varies, and human reasoning is a separate question. A criterion that can always be applied is more useful than one that is sometimes unattainable.

Q: What is a knowledge base in a symbolic system?

A knowledge base holds sentences in a representation language, and an inference procedure derives new sentences from them, so what the system knows is separate from how it reasons. That separation is what allows knowledge to be added without rewriting the reasoner. Training data, decision logs and algorithm libraries are different artefacts. Keeping declarative knowledge apart from procedural machinery is the founding architectural idea of the knowledge based tradition.

81.

Which environment is fully observable, deterministic, static and discrete?

A.Chess without a clock
B.Driving a car through a city in ordinary traffic
C.Playing poker against several other players
D.A robot vacuum cleaner operating in a house

Answer: A

Chess without a clock presents the whole board to both players, has fixed rules, waits for the player to move and has finitely many positions, which is why it was studied so early and so successfully. Driving fails every criterion, poker is partially observable and stochastic, and a vacuum cleaner is partially observable and continuous. Classifying a familiar task against the dimensions is the quickest way to fix them in mind.

Discuss in Forum

82.

Why was chess a favoured early research problem?

A.It has commercial applications that attracted early funding
B.Its environment is simple to specify while the task is genuinely hard
C.It requires the same commonsense knowledge as everyday tasks
D.It is fully solvable by exhaustive search on modest hardware

Answer: B

The rules are trivially stated and the state is fully visible, yet the branching factor makes exhaustive search impossible, so the game isolates the search and evaluation problem from every other difficulty. That made it an ideal laboratory. It requires no commonsense knowledge, which was later seen as a limitation of the choice, and it is not exhaustively solvable. A problem that isolates one difficulty cleanly is worth more as a testbed than a realistic one that mixes several.

Discuss in Forum

83.

What was later seen as the limitation of chess as a research problem?

A.No program was ever able to defeat a strong human player
B.The game required more computing power than was available
C.Success there did not transfer to messier real world tasks
D.The rules of chess proved too complicated to encode correctly

Answer: C

Programs that reached world championship strength did so with search and evaluation techniques that transferred poorly to perception, language and physical action, so the milestone proved less significant than had been hoped. The rules were easy, the milestone was reached, and hardware sufficed. A benchmark whose solution does not generalise is a recurring disappointment in this field, and it is why benchmark choice attracts so much scrutiny.

Discuss in Forum

84.

What is the significance of a system defeating a human champion at a game?

A.It proves that general artificial intelligence has been achieved
B.It shows that the system reasons in the same way a person does
C.It establishes that the game has been completely solved
D.It demonstrates capability in that domain and little beyond it

Answer: D

Such results are genuine achievements within their domain but narrow, since the system cannot transfer its competence to anything else. Interpreting them as evidence of general intelligence is the commonest error in reporting on the field. Nor do they establish any similarity of mechanism, and most such games remain unsolved in the game theoretic sense. Reading a result for exactly what it demonstrates, and no more, is the discipline that discussion of this subject most requires.

Discuss in Forum

85.

What is meant by an agent's autonomy?

A.The extent to which it relies on its own experience rather than prior knowledge
B.The distance it can travel without any human intervention whatever
C.The length of time it can operate before its power supply is exhausted
D.The number of decisions it takes without consulting a human operator

Answer: A

An agent is autonomous to the degree that its behaviour is determined by what it has learned rather than entirely by what its designer built in, which matters because a purely preprogrammed agent fails whenever the designer's assumptions do. Endurance and unsupervised operation are practical measures often loosely called autonomy but the technical sense concerns the origin of the behaviour. Distinguishing designed from learned competence explains why learning is treated as an architectural addition rather than an application.

Discuss in Forum

86.

Why should an agent have some autonomy in the technical sense?

A.An autonomous agent needs no performance measure to guide it
B.A designer cannot anticipate every situation the agent will meet
C.Autonomous agents are cheaper to build than preprogrammed ones
D.Regulations require systems to operate without human supervision

Answer: B

Any fixed set of rules embodies assumptions that will eventually be violated, and an agent able to learn from its own experience can adapt where a preprogrammed one simply fails. Autonomy costs more rather than less to achieve, regulation generally pushes the other way, and a performance measure is more rather than less necessary for a learning agent. Building in the capacity to cope with what was not foreseen is what distinguishes a robust system from a merely complete one.

Discuss in Forum

87.

What does the phrase acting humanly describe as a goal for the field?

A.Building systems that always defer to human judgement
B.Designing interfaces that people find natural to use
C.Producing behaviour indistinguishable from that of a person
D.Reproducing the internal reasoning steps a person would take

Answer: C

Acting humanly is the Turing test programme, judged entirely by whether the behaviour passes for human, while thinking humanly is the cognitive modelling programme judged by whether the process matches. Deference and interface design are separate concerns. Keeping the four quadrants apart matters because each implies a different standard of evidence: a psychological experiment for one, a behavioural test for another and a mathematical argument for the rational quadrants.

Discuss in Forum

88.

What does thinking rationally describe as a goal for the field?

A.Producing conclusions that a thoughtful person would agree with
B.Selecting actions that maximise an agent's expected utility
C.Reasoning quickly enough to respond in real time to events
D.Deriving conclusions by correct logical inference from premises

Answer: D

The thinking rationally quadrant is the logicist programme, in which correct inference from true premises is the criterion, and it gave the field its logic and theorem proving traditions. Acting rationally, the fourth quadrant, is about choosing actions well and is broader, since acting well sometimes requires acting before a proof is available. Human agreement and speed are separate criteria. The gap between correct inference and effective action is exactly why the field moved from the third quadrant to the fourth.

Discuss in Forum

89.

Why did the field move from thinking rationally to acting rationally?

A.Correct inference is often too slow or impossible under uncertainty
B.Logic was found to contain contradictions that could not be resolved
C.Acting rationally requires less computing power in every case
D.Human beings were shown never to reason logically at all

Answer: A

Proving the right thing to do may be intractable, and in many situations no proof is available at all, yet an agent must still act, so the criterion shifted to doing the best it can with what it knows. Probability and decision theory replaced logic as the central formalism for uncertain domains. Logic is consistent, computational cost varies, and human reasoning is a separate question. A criterion that can always be applied is more useful than one that is sometimes unattainable.

Discuss in Forum

90.

What is a knowledge base in a symbolic system?

A.A library of algorithms the system can select between at run time
B.A store of sentences about the world in a formal language
C.A collection of training examples used to fit a model's parameters
D.A database of past decisions the system has taken and their outcomes

Answer: B

A knowledge base holds sentences in a representation language, and an inference procedure derives new sentences from them, so what the system knows is separate from how it reasons. That separation is what allows knowledge to be added without rewriting the reasoner. Training data, decision logs and algorithm libraries are different artefacts. Keeping declarative knowledge apart from procedural machinery is the founding architectural idea of the knowledge based tradition.

Discuss in Forum

Loading…

Topics

Other Subjects

Tags: Introduction to Artificial Intelligence Questions and Answers || Introduction to Artificial Intelligence MCQ Questions and Answers || Introduction to Artificial Intelligence GK Questions and Answers || Introduction to Artificial Intelligence GK MCQ Questions || Introduction to Artificial Intelligence Multiple Choice Questions and Answers || Introduction to Artificial Intelligence GK || GK on Introduction to Artificial Intelligence || Artificial Intelligence Questions and Answers || Artificial Intelligence MCQ Questions and Answers || Artificial Intelligence GK Questions and Answers || GK on Artificial Intelligence

Introduction to Artificial Intelligence Questions and Answers

Topics

Other Subjects

About Us

Stay informed about all important Dates & Notifications