Introduction Flashcards
What does it mean to “act humanly?”
This pertains to the Turing Test approach of AI. Teaching AI to mimic human behaviour.
What is the Turing Test?
A thought experiment that would sidestep philosophical vagueness of the question, “can machines think?” A computer passes the test if a human interrogator, after posing some written questions, cannot tell whether the written response comes from a person or from a computer.
What is the Winograd Schema Challenge?
A test of machine intelligence designed to be an improvement on the Turing test. It is a multiple-choice test that employs questions to be answered by applying the resolution of anaphora. the machine must identify the antecedent of an ambigious pronoun in a statement.
What is an example of the Winograd Schema Challenge?
“The trophy doesn’t fit in the suitcase because it’s too large.” - “it’s” refers to the trophy is this sentence.
“The trophy doesn’t fit in the suitcase because it’s too small.” - “it’s” refers to the suitcase in this sentence.
What is the Chinese Room Argument?
Imagine a person who doesn’t speak chinese is standing in a room, translating Chinese to English following a set of instructions. Does the person in the room understand Chinese? It argues that syntax is not semantics.
What does it mean to “think humanly.”
The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, learning etc.
What interesting findings were there from the hermaphrodite nematode study?
Scientists recreated the nematode brain (302 neurons, 132 muscles and 26 non-muscle end organs).
What does it mean to think rationally?
The solving of problems through logic-based thinking.
What is a syllogism?
A logical structure that applies deductive reasoning to arrive at a conclusion given true premises.
What does it mean to “act rationally.”
This refers to an agent acting in an environment - the creation and study of intelligent agents.
What are the three criteria for an agent acting intelligently?
- Actions are appropriate for it’s goals and circumstances
- It’s flexible to changing environments and goals
- It learns from experience
What are two schools of though on AI?
- Weak AI - the belief that machines can be made to act as if they were intelligent.
- Strong AI - the belief that machines are actually thinking.
What are the two goals of AI?
- Scientific goal: to understand the principles that make intelligent behaviour possible in natural or artificial systems.
- Engineering goal: design useful, intelligent artifacts
What is an agent?
An agent is an entity that perceives its environment through sensors and acts on its environment through effectors/actuators. An agent can be seen as a mapping between percept sequences and actions.
What can be said about an agent that relies less on built-in knowledge?
The agent is more autonomous.
What is a percept?
The input that an intelligent agent is perceiving at any given moment.
What is a rational agent?
A rational agent is:
- the performance measure defining the agent’s degree of success
- the percept sequence, the sequence of all things perceived by the agent
- the agent’s knowledge of the environment
- the actions that the agent can perform
For each possible percept sequence, an ideal rational agent does whatever is possible to maximize its performance, based on the percept sequence and its built-in knowledge.
What is the task environment?
PEAS - Performance measure, Environment, Actuators and Sensors
What is the simulated vs Situated environment?
Simulated: a separate program is used to simulate an environment, feed precepts to agents, evaluate performance.
Situated: the agent acts directly ont he actual environment the agent has a physical body in the world.
What is embodiment?
The robots have bodies and experience the world directly - their actions are part of the dynamics with the world, and actions have immediate feedback on the robot’s own sensations.
What are static vs dynamic environments?
Static: The environment does not change while the agent is thinking
Dynamic: the environment may change while the agent is thinking
What are discrete vs continuous environments?
Discrete: only a finite (or countable) number of discrete percepts/actions
Continuous: states, percepts, or actions can vary continuously.
What are fully observable and partially observable environments?
Fully observable: agent percept contains all relevant information about the world.
Partially Observable: some relevant information is hidden from the agent.
What is deterministic vs Stochastic environments?
Deterministic: the current state uniquely determines the next state.
Stochastic: there is some random element involved
What are episodic vs Sequential environments?
episodic: Every action by the agent is evaluated independently
Sequential: the agent is evaluated based on a long sequence of actions
What are unknown vs known environments?
Unknown: the rules of the game or physics/dynamics of the environment are partially/fully unknown to the agent.
Known: the rules of the game, or physics/dynamics of the environment are known to the agent.
What is a single-agent vs multi-agent environment?
Single agent: there is only one agent operating in the environment
Multi-agent: there are more than one agent’s operating in the environment
What is a reactive agent?
An agent that chooses the next action based solely on what they currently percieve, using a policy or set of rules which are simple to apply.
What are the limitations of reactive agents?
Reactive agents have no memory or state, they are unable to base decisions on previous observations. They may repeat the same sequence of actions over and over.
What is a model-based agent?
Model-based agents are the most effective way to handle partial observability. The agent keeps track of the current state of the world, using an internal model. It then chooses an action in the same way as the reflex agent.
What are the limitations of a model-based agent?
These agents will perform poorly when the task requires searching several moves ahead, complex tasks requiring many individual steps and logical reasoning to achieve goals.
What is a planning agent?
A planning agent will construct plans to achieve it’s goals, and then execute them. Has a world model and a planning model which feeds into the actions.
What is a goal-based agent?
An AI agent that is designed to achieve a specific goal. A planning, goal-based agent is more flexible because the knowledge that supports its decisions is represented explicitly and can be modified.
What is a utility-based agent?
A model-based agent uses a model of the world, along with utility function that measures it preferences among states of the world. IT chooses the action that leads to the best expected utility, where expected utility is computed by averaging over all possible outcome states, weighted by the probability of the outcome.
What is a game-playing agent?
An agent with an opponent model, world model and planning model.
What is a learning agent?
Learning is not a separate module, but rather a set of techniques for improving the existing modules.
