CS461 Fundamentals and Unguided Search (Quiz 1)
Which of the following areas does the field of Artificial Intelligence borrow concepts from?
psychology, linguistics, economics, and mathematics
This school of thought holds that systems should only be considered intelligent if they follow the same reasoning process as humans. This is known as
strong AI
single agent
the agent is the only goal-directed actor in the environment
correct
the algorithm can find a valid solution
complete
the algorithm can find every solution
optimal
the algorithm can find the 'best' solution (however 'best' is defined)
informed
the algorithm can guide or direct its search some way other than blindly generating every possible state
optimally efficient
the algorithm finds the solution at least as fast (in big-O form) as any other algorithm
static
the environment will not change while the agent is deliberating
sequential
the history of previous states is also relevant
deterministic
the next problem state depends only on the current state and the action chosen
known
the outcomes of each action (or their probabilities) are available to the agent the agent is aware of the "laws of physics" for the problem it is to solve
stochastic
the transition to the next state depends on a random element in addition to the agent's action
multi-agent
there are other agents also working in the environment
continuous
time is treated as a constantly-changing flow
discrete
time, space, or both are considered in distinct unit-sized steps
The school of thought that intelligent systems should be judged by whether they achieve correct results, even if the method used to reach those results is much different than human cognition. This is known as
weak AI
What is the difference between algorithms and heuristics?
Algorithms may require too much time or storage to be practical; heuristics often get a 'good enough' answer, and do it quickly
The Turing Test is based on whether a computer system can
act humanly enough to fool a human
nondeterministic
all possible outcomes are known, but not their probabilities
observable
all relevant aspects of the environment can be perceived
The key feature of a greedy algorithm is that it
chooses the 'best' option based on purely local information
Natural language processing--for example, determining the contextual changes in definitions or links between direct and indirect references:
is much more difficult than we might have originally expected
partially observable
it is not possible to determine all relevant information about the environmet
What are characteristics of problems well-suited to AI?
none of these is correct
episodic
only the current state, not its history, matters
competitive
other agents may interfere or prevent action by this agent
