Intelligent Agents Questions and Answers

Q: What is metareasoning?

An agent with limited time must decide how long to deliberate before acting, and that decision is itself a computation whose cost must be accounted. Anytime algorithms, which produce a usable answer at any point and improve with time, are the practical expression of it. Reasoning about other agents, meta-level languages and verification are separate topics. A system that allocates its own computational effort rationally is the logical completion of bounded rationality.

Q: What is an anytime algorithm?

An anytime algorithm produces an approximate answer quickly and refines it, so an agent under time pressure can take whatever quality is available when the deadline arrives. Iterative deepening search and iterative refinement planners are examples. Constant running time, arbitrary start points and unpredictability are different properties. Making the quality of an answer a function of the time spent, rather than requiring completion, is exactly what a real time agent needs.

Q: What is a multi-agent system?

When several agents act in the same environment, each must consider that the others are also choosing actions, which brings in game theory, communication and questions of cooperation and competition. Parallelism within one agent, shared embodiment and hierarchy are different arrangements. The essential feature is that another entity is optimising something of its own, which makes the environment neither static nor merely stochastic but strategic.

Q: Why is a multi-agent environment not simply a stochastic one?

Modelling another agent as noise ignores that it is choosing deliberately, and often choosing in response to what this agent does, which random variation never does. A strategic opponent will exploit a predictable policy. Rate of change, definitional exclusions and observability are not the distinction. The difference between a world that is merely uncertain and one that contains another optimiser is what game theory exists to address.

Q: What is a cooperative multi-agent setting?

Cooperation is defined by the objectives, not the architecture or the protocol: agents cooperate when their performance measures align so that helping another helps oneself. Common design, shared language and turn taking may accompany cooperation but none constitutes it, and competitive agents may share all three. Classifying an interaction by whose objective is served is what determines whether coordination or strategy is the appropriate framework.

Q: When are two agents said to be in competition?

Competition arises when the performance measures oppose, as in a zero sum game, so a rational opponent will act to reduce this agent's score. That makes minimax reasoning and worst case assumptions appropriate, where a cooperative setting would call for coordination. Communication restrictions, independent design and turn taking are incidental. Many real settings are partly cooperative and partly competitive, which is harder than either extreme.

Q: What is agent communication used for?

Communication lets agents pool observations, allocate tasks and commit to joint plans, turning a collection of individuals into something that acts coherently. Speech act theory, with informing, requesting and promising as primitives, supplies the usual framework. Program migration, human interrogation and clock synchronisation are distinct capabilities. Treating a message as an action that changes the recipient's state is what places communication inside the agent framework rather than beside it.

Q: Why is a message treated as an action in agent theory?

Because informing another agent alters what it believes and may alter what it does, a message is chosen for its effect exactly as a physical action is, and it can therefore be planned for, deceived with and reasoned about within the same framework. Bandwidth, time and encoding are practical concerns. Bringing communication inside the action framework, rather than treating it as a separate mechanism, is what allows a single planner to reason about both.

Q: What is a software agent or softbot?

A softbot perceives through system calls, file reads and network responses and acts through commands and messages, so its environment is an operating system or a network rather than a room. Web crawlers, trading agents and personal assistants are examples. Simulation, implementation medium and resource footprint are separate matters, and every agent's program is software. Recognising that the agent abstraction applies unchanged to a purely informational environment is what makes it so widely useful.

Q: What environment properties does a web crawling agent face?

A crawler sees only the pages it has fetched, its choices about what to fetch affect what it can reach later, and the web changes while it works. It is also effectively multi-agent, since other crawlers and site operators respond to it. Classifying an unfamiliar task against the standard dimensions is the quickest way to see which techniques will be needed, and here it points at partial observability and sequential decision making rather than simple classification.

21.

What is metareasoning?

A.Reasoning about how much reasoning to do
B.Reasoning about the beliefs of other agents in the environment
C.Reasoning using a meta-level knowledge representation language
D.Reasoning about the correctness of an inference procedure

Answer: A

An agent with limited time must decide how long to deliberate before acting, and that decision is itself a computation whose cost must be accounted. Anytime algorithms, which produce a usable answer at any point and improve with time, are the practical expression of it. Reasoning about other agents, meta-level languages and verification are separate topics. A system that allocates its own computational effort rationally is the logical completion of bounded rationality.

Discuss in Forum

22.

What is an anytime algorithm?

A.One whose running time cannot be predicted in advance
B.One that can be interrupted and still return a usable answer
C.One that runs in constant time regardless of input size
D.One that may be started at any point in the agent's operation

Answer: B

An anytime algorithm produces an approximate answer quickly and refines it, so an agent under time pressure can take whatever quality is available when the deadline arrives. Iterative deepening search and iterative refinement planners are examples. Constant running time, arbitrary start points and unpredictability are different properties. Making the quality of an answer a function of the time spent, rather than requiring completion, is exactly what a real time agent needs.

Discuss in Forum

23.

What is a multi-agent system?

A.A system in which one agent controls several physical bodies
B.A design in which agents are arranged in a strict hierarchy
C.One in which several agents' actions affect a common environment
D.One agent running several programs in parallel on a single processor

Answer: C

When several agents act in the same environment, each must consider that the others are also choosing actions, which brings in game theory, communication and questions of cooperation and competition. Parallelism within one agent, shared embodiment and hierarchy are different arrangements. The essential feature is that another entity is optimising something of its own, which makes the environment neither static nor merely stochastic but strategic.

Discuss in Forum

24.

Why is a multi-agent environment not simply a stochastic one?

A.The other agents change the environment more rapidly than chance would
B.Stochastic environments cannot contain more than one agent by definition
C.Multi-agent environments are always fully observable to every participant
D.The other agents choose actions to further their own objectives

Answer: D

Modelling another agent as noise ignores that it is choosing deliberately, and often choosing in response to what this agent does, which random variation never does. A strategic opponent will exploit a predictable policy. Rate of change, definitional exclusions and observability are not the distinction. The difference between a world that is merely uncertain and one that contains another optimiser is what game theory exists to address.

Discuss in Forum

25.

What is a cooperative multi-agent setting?

A.The agents share a performance measure and gain from coordination
B.The agents are all built by the same designer using one architecture
C.The agents communicate with each other using a common language
D.The agents take turns rather than acting simultaneously

Answer: A

Cooperation is defined by the objectives, not the architecture or the protocol: agents cooperate when their performance measures align so that helping another helps oneself. Common design, shared language and turn taking may accompany cooperation but none constitutes it, and competitive agents may share all three. Classifying an interaction by whose objective is served is what determines whether coordination or strategy is the appropriate framework.

Discuss in Forum

26.

When are two agents said to be in competition?

A.Only one agent is permitted to act at any given moment
B.One agent's gain is another's loss under their performance measures
C.The agents are prevented from communicating with one another
D.The agents were designed independently by different organisations

Answer: B

Competition arises when the performance measures oppose, as in a zero sum game, so a rational opponent will act to reduce this agent's score. That makes minimax reasoning and worst case assumptions appropriate, where a cooperative setting would call for coordination. Communication restrictions, independent design and turn taking are incidental. Many real settings are partly cooperative and partly competitive, which is harder than either extreme.

Discuss in Forum

27.

What is agent communication used for?

A.Allowing a human operator to interrogate the agent's reasoning
B.Synchronising the internal clocks of a distributed agent system
C.Sharing information and coordinating actions between agents
D.Transferring an agent's program from one machine to another

Answer: C

Communication lets agents pool observations, allocate tasks and commit to joint plans, turning a collection of individuals into something that acts coherently. Speech act theory, with informing, requesting and promising as primitives, supplies the usual framework. Program migration, human interrogation and clock synchronisation are distinct capabilities. Treating a message as an action that changes the recipient's state is what places communication inside the agent framework rather than beside it.

Discuss in Forum

28.

Why is a message treated as an action in agent theory?

A.Messages consume communication bandwidth that must be budgeted
B.Sending a message takes time that could be spent on other actions
C.Messages must be encoded in a formal language to be understood
D.It changes the state of the recipient and so has consequences

Answer: D

Because informing another agent alters what it believes and may alter what it does, a message is chosen for its effect exactly as a physical action is, and it can therefore be planned for, deceived with and reasoned about within the same framework. Bandwidth, time and encoding are practical concerns. Bringing communication inside the action framework, rather than treating it as a separate mechanism, is what allows a single planner to reason about both.

Discuss in Forum

29.

What is a software agent or softbot?

A.An agent whose environment is a software system rather than the physical world
B.A simulation of a physical robot used for testing before deployment
C.An agent implemented entirely in software rather than in hardware
D.A lightweight agent designed to run on limited computing resources

Answer: A

A softbot perceives through system calls, file reads and network responses and acts through commands and messages, so its environment is an operating system or a network rather than a room. Web crawlers, trading agents and personal assistants are examples. Simulation, implementation medium and resource footprint are separate matters, and every agent's program is software. Recognising that the agent abstraction applies unchanged to a purely informational environment is what makes it so widely useful.

Discuss in Forum

30.

What environment properties does a web crawling agent face?

A.Partially observable, episodic and deterministic
B.Partially observable, sequential and dynamic
C.Fully observable, episodic and static
D.Fully observable, deterministic and discrete

Answer: B

A crawler sees only the pages it has fetched, its choices about what to fetch affect what it can reach later, and the web changes while it works. It is also effectively multi-agent, since other crawlers and site operators respond to it. Classifying an unfamiliar task against the standard dimensions is the quickest way to see which techniques will be needed, and here it points at partial observability and sequential decision making rather than simple classification.

Discuss in Forum

Loading…

Topics

Other Subjects

Tags: Intelligent Agents Questions and Answers || Intelligent Agents MCQ Questions and Answers || Intelligent Agents GK Questions and Answers || Intelligent Agents GK MCQ Questions || Intelligent Agents Multiple Choice Questions and Answers || Intelligent Agents GK || GK on Intelligent Agents || Artificial Intelligence Questions and Answers || Artificial Intelligence MCQ Questions and Answers || Artificial Intelligence GK Questions and Answers || GK on Artificial Intelligence

Intelligent Agents Questions and Answers

Topics

Other Subjects

About Us

Stay informed about all important Dates & Notifications